{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 3120,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.4296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8497.0,
      "completions/mean_length": 7881.193359375,
      "completions/mean_terminated_length": 1474.9691162109375,
      "completions/min_length": 379.0,
      "completions/min_terminated_length": 379.0,
      "entropy": 0.14000436663627625,
      "epoch": 0.0016025641025641025,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 568.662109375,
      "learning_rate": 1e-06,
      "loss": 0.274,
      "num_tokens": 4883523.0,
      "reward": 0.28620433807373047,
      "reward_std": 0.19634175300598145,
      "rewards/progression_diversity/mean": -0.09636208415031433,
      "rewards/progression_diversity/std": 0.1318667083978653,
      "rewards/symbolic_reward_accuracy/mean": 0.265625,
      "rewards/symbolic_reward_accuracy/std": 0.44209739565849304,
      "rewards/symbolic_reward_partial_score/mean": 0.5490233898162842,
      "rewards/symbolic_reward_partial_score/std": 0.4077247977256775,
      "rewards/tag_count_reward/mean": -0.369140625,
      "rewards/tag_count_reward/std": 0.4830440282821655,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9890698790550232,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 412.0,
      "sampling/sampling_logp_difference/mean": 11.238412857055664,
      "step": 1
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.15292271971702576,
      "epoch": 0.003205128205128205,
      "grad_norm": 1453.4033203125,
      "learning_rate": 1e-06,
      "loss": 0.2567,
      "step": 2
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.1447555311024189,
      "epoch": 0.004807692307692308,
      "grad_norm": 155.9544219970703,
      "learning_rate": 1e-06,
      "loss": 0.2182,
      "step": 3
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.3828125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.15076642483472824,
      "epoch": 0.00641025641025641,
      "grad_norm": 180.8573455810547,
      "learning_rate": 1e-06,
      "loss": 0.2647,
      "step": 4
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.47265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8724.0,
      "completions/mean_length": 8531.75,
      "completions/mean_terminated_length": 1493.807373046875,
      "completions/min_length": 354.0,
      "completions/min_terminated_length": 354.0,
      "entropy": 0.141545832157135,
      "epoch": 0.008012820512820512,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 267.66961669921875,
      "learning_rate": 1e-06,
      "loss": 0.1422,
      "num_tokens": 10076899.0,
      "reward": 0.27696967124938965,
      "reward_std": 0.21476207673549652,
      "rewards/progression_diversity/mean": -0.10186256468296051,
      "rewards/progression_diversity/std": 0.1328817903995514,
      "rewards/symbolic_reward_accuracy/mean": 0.267578125,
      "rewards/symbolic_reward_accuracy/std": 0.4431293308734894,
      "rewards/symbolic_reward_partial_score/mean": 0.5268880128860474,
      "rewards/symbolic_reward_partial_score/std": 0.4123326241970062,
      "rewards/tag_count_reward/mean": -0.40625,
      "rewards/tag_count_reward/std": 0.49161264300346375,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.984929621219635,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 412.0,
      "sampling/sampling_logp_difference/mean": 11.871856689453125,
      "step": 5
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.14717822521924973,
      "epoch": 0.009615384615384616,
      "grad_norm": 116.89068603515625,
      "learning_rate": 1e-06,
      "loss": 0.2298,
      "step": 6
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.13329345732927322,
      "epoch": 0.011217948717948718,
      "grad_norm": 971.0422973632812,
      "learning_rate": 1e-06,
      "loss": 0.3372,
      "step": 7
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.15128736197948456,
      "epoch": 0.01282051282051282,
      "grad_norm": 416.9343566894531,
      "learning_rate": 1e-06,
      "loss": 0.3162,
      "step": 8
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.44140625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8558.0,
      "completions/mean_length": 8034.72265625,
      "completions/mean_terminated_length": 1437.0419921875,
      "completions/min_length": 345.0,
      "completions/min_terminated_length": 345.0,
      "entropy": 0.12974420934915543,
      "epoch": 0.014423076923076924,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 525.6014404296875,
      "learning_rate": 1e-06,
      "loss": 0.3298,
      "num_tokens": 15093045.0,
      "reward": 0.22926117479801178,
      "reward_std": 0.21232059597969055,
      "rewards/progression_diversity/mean": -0.10122619569301605,
      "rewards/progression_diversity/std": 0.1385258138179779,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.5384114980697632,
      "rewards/symbolic_reward_partial_score/std": 0.38595736026763916,
      "rewards/tag_count_reward/mean": -0.390625,
      "rewards/tag_count_reward/std": 0.48836761713027954,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9804725050926208,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 412.0,
      "sampling/sampling_logp_difference/mean": 12.684000968933105,
      "step": 9
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.1481311321258545,
      "epoch": 0.016025641025641024,
      "grad_norm": 207.61981201171875,
      "learning_rate": 1e-06,
      "loss": 0.2655,
      "step": 10
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.14685294777154922,
      "epoch": 0.017628205128205128,
      "grad_norm": 452.76580810546875,
      "learning_rate": 1e-06,
      "loss": 0.3486,
      "step": 11
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.13027992472052574,
      "epoch": 0.019230769230769232,
      "grad_norm": 58.394657135009766,
      "learning_rate": 1e-06,
      "loss": 0.3177,
      "step": 12
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.384765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4718.0,
      "completions/mean_length": 7190.44921875,
      "completions/mean_terminated_length": 1440.831787109375,
      "completions/min_length": 343.0,
      "completions/min_terminated_length": 343.0,
      "entropy": 0.145263209939003,
      "epoch": 0.020833333333333332,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 442.4850158691406,
      "learning_rate": 1e-06,
      "loss": 0.2593,
      "num_tokens": 19648011.0,
      "reward": 0.234406977891922,
      "reward_std": 0.19648107886314392,
      "rewards/progression_diversity/mean": -0.08762294054031372,
      "rewards/progression_diversity/std": 0.1332884430885315,
      "rewards/symbolic_reward_accuracy/mean": 0.16796875,
      "rewards/symbolic_reward_accuracy/std": 0.374204158782959,
      "rewards/symbolic_reward_partial_score/mean": 0.5577148199081421,
      "rewards/symbolic_reward_partial_score/std": 0.37337514758110046,
      "rewards/tag_count_reward/mean": -0.328125,
      "rewards/tag_count_reward/std": 0.4699897766113281,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.983834981918335,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 414.0,
      "sampling/sampling_logp_difference/mean": 12.08355712890625,
      "step": 13
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.14843080937862396,
      "epoch": 0.022435897435897436,
      "grad_norm": 94.6893081665039,
      "learning_rate": 1e-06,
      "loss": 0.2235,
      "step": 14
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.16313014179468155,
      "epoch": 0.02403846153846154,
      "grad_norm": 207.52377319335938,
      "learning_rate": 1e-06,
      "loss": 0.2741,
      "step": 15
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.15617449581623077,
      "epoch": 0.02564102564102564,
      "grad_norm": 304.16375732421875,
      "learning_rate": 1e-06,
      "loss": 0.3583,
      "step": 16
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.349609375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8593.0,
      "completions/mean_length": 6609.98046875,
      "completions/mean_terminated_length": 1356.078125,
      "completions/min_length": 289.0,
      "completions/min_terminated_length": 289.0,
      "entropy": 0.20762895047664642,
      "epoch": 0.027243589743589744,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 390.9715576171875,
      "learning_rate": 1e-06,
      "loss": 0.2115,
      "num_tokens": 23865953.0,
      "reward": 0.32011666893959045,
      "reward_std": 0.19630393385887146,
      "rewards/progression_diversity/mean": -0.08647972345352173,
      "rewards/progression_diversity/std": 0.1337437778711319,
      "rewards/symbolic_reward_accuracy/mean": 0.28515625,
      "rewards/symbolic_reward_accuracy/std": 0.45193037390708923,
      "rewards/symbolic_reward_partial_score/mean": 0.6044433116912842,
      "rewards/symbolic_reward_partial_score/std": 0.38585761189460754,
      "rewards/tag_count_reward/mean": -0.314453125,
      "rewards/tag_count_reward/std": 0.4647517800331116,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.987565815448761,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 416.0,
      "sampling/sampling_logp_difference/mean": 11.790186882019043,
      "step": 17
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5859375,
      "entropy": 0.16485845297574997,
      "epoch": 0.028846153846153848,
      "grad_norm": 20.569774627685547,
      "learning_rate": 1e-06,
      "loss": 0.3498,
      "step": 18
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.14989716559648514,
      "epoch": 0.030448717948717948,
      "grad_norm": 29.013126373291016,
      "learning_rate": 1e-06,
      "loss": 0.2945,
      "step": 19
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.16761308163404465,
      "epoch": 0.03205128205128205,
      "grad_norm": 17.99753761291504,
      "learning_rate": 1e-06,
      "loss": 0.2861,
      "step": 20
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.396484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 15963.0,
      "completions/mean_length": 7455.453125,
      "completions/mean_terminated_length": 1589.7735595703125,
      "completions/min_length": 342.0,
      "completions/min_terminated_length": 342.0,
      "entropy": 0.1430811509490013,
      "epoch": 0.03365384615384615,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 338.8027648925781,
      "learning_rate": 1e-06,
      "loss": 0.3612,
      "num_tokens": 28525033.0,
      "reward": 0.2473517507314682,
      "reward_std": 0.21615710854530334,
      "rewards/progression_diversity/mean": -0.09685845673084259,
      "rewards/progression_diversity/std": 0.13814808428287506,
      "rewards/symbolic_reward_accuracy/mean": 0.193359375,
      "rewards/symbolic_reward_accuracy/std": 0.39531853795051575,
      "rewards/symbolic_reward_partial_score/mean": 0.5562499761581421,
      "rewards/symbolic_reward_partial_score/std": 0.38326844573020935,
      "rewards/tag_count_reward/mean": -0.345703125,
      "rewards/tag_count_reward/std": 0.4760620892047882,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9782972931861877,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 416.0,
      "sampling/sampling_logp_difference/mean": 13.17431640625,
      "step": 21
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.15020054578781128,
      "epoch": 0.035256410256410256,
      "grad_norm": 72.29728698730469,
      "learning_rate": 1e-06,
      "loss": 0.2999,
      "step": 22
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.14304012060165405,
      "epoch": 0.03685897435897436,
      "grad_norm": 74.60310363769531,
      "learning_rate": 1e-06,
      "loss": 0.2514,
      "step": 23
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.1455524042248726,
      "epoch": 0.038461538461538464,
      "grad_norm": 23.46770668029785,
      "learning_rate": 1e-06,
      "loss": 0.2941,
      "step": 24
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.431640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8141.0,
      "completions/mean_length": 7897.568359375,
      "completions/mean_terminated_length": 1452.54638671875,
      "completions/min_length": 358.0,
      "completions/min_terminated_length": 358.0,
      "entropy": 0.15165143460035324,
      "epoch": 0.04006410256410257,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 739.3831176757812,
      "learning_rate": 1e-06,
      "loss": 0.2292,
      "num_tokens": 33441724.0,
      "reward": 0.2494029402732849,
      "reward_std": 0.1849609613418579,
      "rewards/progression_diversity/mean": -0.10511670261621475,
      "rewards/progression_diversity/std": 0.14027173817157745,
      "rewards/symbolic_reward_accuracy/mean": 0.208984375,
      "rewards/symbolic_reward_accuracy/std": 0.40698084235191345,
      "rewards/symbolic_reward_partial_score/mean": 0.5431803464889526,
      "rewards/symbolic_reward_partial_score/std": 0.38958677649497986,
      "rewards/tag_count_reward/mean": -0.37890625,
      "rewards/tag_count_reward/std": 0.4855891764163971,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9824032783508301,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 416.0,
      "sampling/sampling_logp_difference/mean": 12.627696990966797,
      "step": 25
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.14603624492883682,
      "epoch": 0.041666666666666664,
      "grad_norm": 473.4931335449219,
      "learning_rate": 1e-06,
      "loss": 0.2678,
      "step": 26
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.3515625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.13400625064969063,
      "epoch": 0.04326923076923077,
      "grad_norm": 485.1720886230469,
      "learning_rate": 1e-06,
      "loss": 0.3149,
      "step": 27
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.14134199172258377,
      "epoch": 0.04487179487179487,
      "grad_norm": 265.03302001953125,
      "learning_rate": 1e-06,
      "loss": 0.2932,
      "step": 28
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5220.0,
      "completions/mean_length": 8731.86328125,
      "completions/mean_terminated_length": 1543.492431640625,
      "completions/min_length": 366.0,
      "completions/min_terminated_length": 366.0,
      "entropy": 0.12327419593930244,
      "epoch": 0.046474358974358976,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 270.61407470703125,
      "learning_rate": 1e-06,
      "loss": 0.2511,
      "num_tokens": 38830534.0,
      "reward": 0.20257076621055603,
      "reward_std": 0.1862163543701172,
      "rewards/progression_diversity/mean": -0.11499577015638351,
      "rewards/progression_diversity/std": 0.14146001636981964,
      "rewards/symbolic_reward_accuracy/mean": 0.166015625,
      "rewards/symbolic_reward_accuracy/std": 0.3724585771560669,
      "rewards/symbolic_reward_partial_score/mean": 0.49026691913604736,
      "rewards/symbolic_reward_partial_score/std": 0.3924698829650879,
      "rewards/tag_count_reward/mean": -0.4296875,
      "rewards/tag_count_reward/std": 0.4955156147480011,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.978987991809845,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 416.0,
      "sampling/sampling_logp_difference/mean": 13.129805564880371,
      "step": 29
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.13143375515937805,
      "epoch": 0.04807692307692308,
      "grad_norm": 348.0553283691406,
      "learning_rate": 1e-06,
      "loss": 0.2894,
      "step": 30
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.1378588154911995,
      "epoch": 0.049679487179487176,
      "grad_norm": 297.4830322265625,
      "learning_rate": 1e-06,
      "loss": 0.2873,
      "step": 31
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.4765625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.11010397225618362,
      "epoch": 0.05128205128205128,
      "grad_norm": 120.72930145263672,
      "learning_rate": 1e-06,
      "loss": 0.3805,
      "step": 32
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.412109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 10000.0,
      "completions/mean_length": 7616.541015625,
      "completions/mean_terminated_length": 1470.5814208984375,
      "completions/min_length": 363.0,
      "completions/min_terminated_length": 363.0,
      "entropy": 0.15361008048057556,
      "epoch": 0.052884615384615384,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 288.5027160644531,
      "learning_rate": 1e-06,
      "loss": 0.2605,
      "num_tokens": 43573467.0,
      "reward": 0.26951634883880615,
      "reward_std": 0.16951517760753632,
      "rewards/progression_diversity/mean": -0.10158812254667282,
      "rewards/progression_diversity/std": 0.14149615168571472,
      "rewards/symbolic_reward_accuracy/mean": 0.236328125,
      "rewards/symbolic_reward_accuracy/std": 0.42524150013923645,
      "rewards/symbolic_reward_partial_score/mean": 0.549560546875,
      "rewards/symbolic_reward_partial_score/std": 0.393799364566803,
      "rewards/tag_count_reward/mean": -0.361328125,
      "rewards/tag_count_reward/std": 0.48085519671440125,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9846019744873047,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 420.0,
      "sampling/sampling_logp_difference/mean": 12.345019340515137,
      "step": 33
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.1322014182806015,
      "epoch": 0.05448717948717949,
      "grad_norm": 719.1787719726562,
      "learning_rate": 1e-06,
      "loss": 0.2974,
      "step": 34
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.1471644788980484,
      "epoch": 0.05608974358974359,
      "grad_norm": 506.037109375,
      "learning_rate": 1e-06,
      "loss": 0.2677,
      "step": 35
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.144605353474617,
      "epoch": 0.057692307692307696,
      "grad_norm": 172.5960693359375,
      "learning_rate": 1e-06,
      "loss": 0.2405,
      "step": 36
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.376953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8070.0,
      "completions/mean_length": 7113.50390625,
      "completions/mean_terminated_length": 1504.7083740234375,
      "completions/min_length": 377.0,
      "completions/min_terminated_length": 377.0,
      "entropy": 0.15522907674312592,
      "epoch": 0.05929487179487179,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 546.6926879882812,
      "learning_rate": 1e-06,
      "loss": 0.314,
      "num_tokens": 48071261.0,
      "reward": 0.29389289021492004,
      "reward_std": 0.20416641235351562,
      "rewards/progression_diversity/mean": -0.09996913373470306,
      "rewards/progression_diversity/std": 0.1435985416173935,
      "rewards/symbolic_reward_accuracy/mean": 0.25390625,
      "rewards/symbolic_reward_accuracy/std": 0.43567025661468506,
      "rewards/symbolic_reward_partial_score/mean": 0.5851888656616211,
      "rewards/symbolic_reward_partial_score/std": 0.38423076272010803,
      "rewards/tag_count_reward/mean": -0.330078125,
      "rewards/tag_count_reward/std": 0.47070086002349854,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9879682660102844,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 420.0,
      "sampling/sampling_logp_difference/mean": 12.075263023376465,
      "step": 37
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.1799706146121025,
      "epoch": 0.060897435897435896,
      "grad_norm": 217.79710388183594,
      "learning_rate": 1e-06,
      "loss": 0.2229,
      "step": 38
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.1559307500720024,
      "epoch": 0.0625,
      "grad_norm": 349.0166931152344,
      "learning_rate": 1e-06,
      "loss": 0.2719,
      "step": 39
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.15247832983732224,
      "epoch": 0.0641025641025641,
      "grad_norm": 172.07949829101562,
      "learning_rate": 1e-06,
      "loss": 0.319,
      "step": 40
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.427734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4535.0,
      "completions/mean_length": 7818.34375,
      "completions/mean_terminated_length": 1416.0272216796875,
      "completions/min_length": 317.0,
      "completions/min_terminated_length": 317.0,
      "entropy": 0.13341034203767776,
      "epoch": 0.06570512820512821,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 447.7645568847656,
      "learning_rate": 1e-06,
      "loss": 0.273,
      "num_tokens": 53003245.0,
      "reward": 0.218738853931427,
      "reward_std": 0.2060059905052185,
      "rewards/progression_diversity/mean": -0.1119544506072998,
      "rewards/progression_diversity/std": 0.1507621854543686,
      "rewards/symbolic_reward_accuracy/mean": 0.171875,
      "rewards/symbolic_reward_accuracy/std": 0.3776407241821289,
      "rewards/symbolic_reward_partial_score/mean": 0.5076009035110474,
      "rewards/symbolic_reward_partial_score/std": 0.3918800354003906,
      "rewards/tag_count_reward/mean": -0.35546875,
      "rewards/tag_count_reward/std": 0.47912323474884033,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9806472659111023,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 420.0,
      "sampling/sampling_logp_difference/mean": 13.268798828125,
      "step": 41
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.14531062543392181,
      "epoch": 0.0673076923076923,
      "grad_norm": 1331.5009765625,
      "learning_rate": 1e-06,
      "loss": 0.3872,
      "step": 42
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.1627102643251419,
      "epoch": 0.06891025641025642,
      "grad_norm": 724.1676025390625,
      "learning_rate": 1e-06,
      "loss": 0.3012,
      "step": 43
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.16679170727729797,
      "epoch": 0.07051282051282051,
      "grad_norm": 689.5390625,
      "learning_rate": 1e-06,
      "loss": 0.3947,
      "step": 44
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.369140625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4762.0,
      "completions/mean_length": 6941.892578125,
      "completions/mean_terminated_length": 1416.9442138671875,
      "completions/min_length": 364.0,
      "completions/min_terminated_length": 364.0,
      "entropy": 0.1360173150897026,
      "epoch": 0.07211538461538461,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 262.25537109375,
      "learning_rate": 1e-06,
      "loss": 0.3696,
      "num_tokens": 57407846.0,
      "reward": 0.32967936992645264,
      "reward_std": 0.2605833411216736,
      "rewards/progression_diversity/mean": -0.10432835668325424,
      "rewards/progression_diversity/std": 0.15316592156887054,
      "rewards/symbolic_reward_accuracy/mean": 0.302734375,
      "rewards/symbolic_reward_accuracy/std": 0.45989060401916504,
      "rewards/symbolic_reward_partial_score/mean": 0.6082682609558105,
      "rewards/symbolic_reward_partial_score/std": 0.3930491805076599,
      "rewards/tag_count_reward/mean": -0.333984375,
      "rewards/tag_count_reward/std": 0.47209542989730835,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9809261560440063,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 422.0,
      "sampling/sampling_logp_difference/mean": 13.200485229492188,
      "step": 45
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.1455419659614563,
      "epoch": 0.07371794871794872,
      "grad_norm": 1122.5438232421875,
      "learning_rate": 1e-06,
      "loss": 0.2685,
      "step": 46
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.17444385588169098,
      "epoch": 0.07532051282051282,
      "grad_norm": 311.899169921875,
      "learning_rate": 1e-06,
      "loss": 0.2814,
      "step": 47
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.17196480184793472,
      "epoch": 0.07692307692307693,
      "grad_norm": 131.08229064941406,
      "learning_rate": 1e-06,
      "loss": 0.2203,
      "step": 48
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.41796875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4885.0,
      "completions/mean_length": 7645.94921875,
      "completions/mean_terminated_length": 1370.97314453125,
      "completions/min_length": 364.0,
      "completions/min_terminated_length": 364.0,
      "entropy": 0.14967508614063263,
      "epoch": 0.07852564102564102,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1408.8106689453125,
      "learning_rate": 1e-06,
      "loss": 0.2864,
      "num_tokens": 62251772.0,
      "reward": 0.27286070585250854,
      "reward_std": 0.2150975465774536,
      "rewards/progression_diversity/mean": -0.11773853003978729,
      "rewards/progression_diversity/std": 0.15623939037322998,
      "rewards/symbolic_reward_accuracy/mean": 0.251953125,
      "rewards/symbolic_reward_accuracy/std": 0.43455907702445984,
      "rewards/symbolic_reward_partial_score/mean": 0.5332520008087158,
      "rewards/symbolic_reward_partial_score/std": 0.40305641293525696,
      "rewards/tag_count_reward/mean": -0.37109375,
      "rewards/tag_count_reward/std": 0.4835699498653412,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9803931713104248,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 424.0,
      "sampling/sampling_logp_difference/mean": 13.439233779907227,
      "step": 49
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.1299237497150898,
      "epoch": 0.08012820512820513,
      "grad_norm": 226.3897705078125,
      "learning_rate": 1e-06,
      "loss": 0.3439,
      "step": 50
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.14567925781011581,
      "epoch": 0.08173076923076923,
      "grad_norm": 225.25894165039062,
      "learning_rate": 1e-06,
      "loss": 0.227,
      "step": 51
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.16078957170248032,
      "epoch": 0.08333333333333333,
      "grad_norm": 259.5469665527344,
      "learning_rate": 1e-06,
      "loss": 0.2553,
      "step": 52
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.384765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 14739.0,
      "completions/mean_length": 7208.03515625,
      "completions/mean_terminated_length": 1469.4158935546875,
      "completions/min_length": 357.0,
      "completions/min_terminated_length": 357.0,
      "entropy": 0.16834494471549988,
      "epoch": 0.08493589743589744,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 538.5927734375,
      "learning_rate": 1e-06,
      "loss": 0.2477,
      "num_tokens": 66784174.0,
      "reward": 0.3087325990200043,
      "reward_std": 0.2367071658372879,
      "rewards/progression_diversity/mean": -0.1125800758600235,
      "rewards/progression_diversity/std": 0.15995118021965027,
      "rewards/symbolic_reward_accuracy/mean": 0.2890625,
      "rewards/symbolic_reward_accuracy/std": 0.45377036929130554,
      "rewards/symbolic_reward_partial_score/mean": 0.5712727904319763,
      "rewards/symbolic_reward_partial_score/std": 0.4000099301338196,
      "rewards/tag_count_reward/mean": -0.349609375,
      "rewards/tag_count_reward/std": 0.47731292247772217,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9840480089187622,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 428.0,
      "sampling/sampling_logp_difference/mean": 12.845325469970703,
      "step": 53
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.15644784271717072,
      "epoch": 0.08653846153846154,
      "grad_norm": 271.4656677246094,
      "learning_rate": 1e-06,
      "loss": 0.2644,
      "step": 54
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.1287785843014717,
      "epoch": 0.08814102564102565,
      "grad_norm": 238.35011291503906,
      "learning_rate": 1e-06,
      "loss": 0.2856,
      "step": 55
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.15913260728120804,
      "epoch": 0.08974358974358974,
      "grad_norm": 207.0217742919922,
      "learning_rate": 1e-06,
      "loss": 0.3107,
      "step": 56
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.44140625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6438.0,
      "completions/mean_length": 8034.2265625,
      "completions/mean_terminated_length": 1436.15380859375,
      "completions/min_length": 296.0,
      "completions/min_terminated_length": 296.0,
      "entropy": 0.14069371670484543,
      "epoch": 0.09134615384615384,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 261.5517578125,
      "learning_rate": 1e-06,
      "loss": 0.2341,
      "num_tokens": 71809458.0,
      "reward": 0.28543901443481445,
      "reward_std": 0.24921724200248718,
      "rewards/progression_diversity/mean": -0.13432222604751587,
      "rewards/progression_diversity/std": 0.17075546085834503,
      "rewards/symbolic_reward_accuracy/mean": 0.2734375,
      "rewards/symbolic_reward_accuracy/std": 0.4461594223976135,
      "rewards/symbolic_reward_partial_score/mean": 0.5334147214889526,
      "rewards/symbolic_reward_partial_score/std": 0.41190940141677856,
      "rewards/tag_count_reward/mean": -0.373046875,
      "rewards/tag_count_reward/std": 0.48408737778663635,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9736584424972534,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 432.0,
      "sampling/sampling_logp_difference/mean": 14.813698768615723,
      "step": 57
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.1330208107829094,
      "epoch": 0.09294871794871795,
      "grad_norm": 234.9654541015625,
      "learning_rate": 1e-06,
      "loss": 0.2835,
      "step": 58
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.4140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.13570880889892578,
      "epoch": 0.09455128205128205,
      "grad_norm": 195.11659240722656,
      "learning_rate": 1e-06,
      "loss": 0.2919,
      "step": 59
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.453125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6171875,
      "entropy": 0.12034280598163605,
      "epoch": 0.09615384615384616,
      "grad_norm": 103.8945083618164,
      "learning_rate": 1e-06,
      "loss": 0.3713,
      "step": 60
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 16376.0,
      "completions/mean_length": 7816.056640625,
      "completions/mean_terminated_length": 1563.773681640625,
      "completions/min_length": 268.0,
      "completions/min_terminated_length": 268.0,
      "entropy": 0.1452302411198616,
      "epoch": 0.09775641025641026,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 524.85205078125,
      "learning_rate": 1e-06,
      "loss": 0.203,
      "num_tokens": 76649167.0,
      "reward": 0.2318515032529831,
      "reward_std": 0.18006575107574463,
      "rewards/progression_diversity/mean": -0.13223427534103394,
      "rewards/progression_diversity/std": 0.17053887248039246,
      "rewards/symbolic_reward_accuracy/mean": 0.189453125,
      "rewards/symbolic_reward_accuracy/std": 0.3922513723373413,
      "rewards/symbolic_reward_partial_score/mean": 0.5187825560569763,
      "rewards/symbolic_reward_partial_score/std": 0.3892955183982849,
      "rewards/tag_count_reward/mean": -0.361328125,
      "rewards/tag_count_reward/std": 0.48085519671440125,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9811015129089355,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 434.0,
      "sampling/sampling_logp_difference/mean": 13.738810539245605,
      "step": 61
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.1534811109304428,
      "epoch": 0.09935897435897435,
      "grad_norm": 12.576883316040039,
      "learning_rate": 1e-06,
      "loss": 0.276,
      "step": 62
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.14197808876633644,
      "epoch": 0.10096153846153846,
      "grad_norm": 69.41390991210938,
      "learning_rate": 1e-06,
      "loss": 0.3,
      "step": 63
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.15332899242639542,
      "epoch": 0.10256410256410256,
      "grad_norm": 46.41984939575195,
      "learning_rate": 1e-06,
      "loss": 0.3122,
      "step": 64
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4021.0,
      "completions/mean_length": 7709.984375,
      "completions/mean_terminated_length": 1380.29736328125,
      "completions/min_length": 336.0,
      "completions/min_terminated_length": 336.0,
      "entropy": 0.11576578766107559,
      "epoch": 0.10416666666666667,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 849.912353515625,
      "learning_rate": 1e-06,
      "loss": 0.3962,
      "num_tokens": 81463175.0,
      "reward": 0.2805628180503845,
      "reward_std": 0.22776535153388977,
      "rewards/progression_diversity/mean": -0.13366013765335083,
      "rewards/progression_diversity/std": 0.1738085299730301,
      "rewards/symbolic_reward_accuracy/mean": 0.267578125,
      "rewards/symbolic_reward_accuracy/std": 0.4431293308734894,
      "rewards/symbolic_reward_partial_score/mean": 0.530810534954071,
      "rewards/symbolic_reward_partial_score/std": 0.40938085317611694,
      "rewards/tag_count_reward/mean": -0.37890625,
      "rewards/tag_count_reward/std": 0.4855891764163971,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9786410331726074,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 436.0,
      "sampling/sampling_logp_difference/mean": 14.30560302734375,
      "step": 65
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.15666405856609344,
      "epoch": 0.10576923076923077,
      "grad_norm": 67.19806671142578,
      "learning_rate": 1e-06,
      "loss": 0.2378,
      "step": 66
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6171875,
      "entropy": 0.14105742424726486,
      "epoch": 0.10737179487179487,
      "grad_norm": 82.8458023071289,
      "learning_rate": 1e-06,
      "loss": 0.3127,
      "step": 67
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.14606218039989471,
      "epoch": 0.10897435897435898,
      "grad_norm": 113.2381362915039,
      "learning_rate": 1e-06,
      "loss": 0.2326,
      "step": 68
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.4140625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5696.0,
      "completions/mean_length": 7631.79296875,
      "completions/mean_terminated_length": 1446.9000244140625,
      "completions/min_length": 475.0,
      "completions/min_terminated_length": 475.0,
      "entropy": 0.15287064760923386,
      "epoch": 0.11057692307692307,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 262.7812194824219,
      "learning_rate": 1e-06,
      "loss": 0.2887,
      "num_tokens": 86220381.0,
      "reward": 0.29364585876464844,
      "reward_std": 0.19495844841003418,
      "rewards/progression_diversity/mean": -0.13199639320373535,
      "rewards/progression_diversity/std": 0.17195309698581696,
      "rewards/symbolic_reward_accuracy/mean": 0.265625,
      "rewards/symbolic_reward_accuracy/std": 0.44209739565849304,
      "rewards/symbolic_reward_partial_score/mean": 0.5769693851470947,
      "rewards/symbolic_reward_partial_score/std": 0.39537864923477173,
      "rewards/tag_count_reward/mean": -0.375,
      "rewards/tag_count_reward/std": 0.4845963716506958,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9830174446105957,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 436.0,
      "sampling/sampling_logp_difference/mean": 13.659445762634277,
      "step": 69
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.1318165846168995,
      "epoch": 0.11217948717948718,
      "grad_norm": 550.3621826171875,
      "learning_rate": 1e-06,
      "loss": 0.3745,
      "step": 70
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.1556747555732727,
      "epoch": 0.11378205128205128,
      "grad_norm": 144.26145935058594,
      "learning_rate": 1e-06,
      "loss": 0.2036,
      "step": 71
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.3671875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.1447812095284462,
      "epoch": 0.11538461538461539,
      "grad_norm": 35.39904022216797,
      "learning_rate": 1e-06,
      "loss": 0.2419,
      "step": 72
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.431640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8566.0,
      "completions/mean_length": 7925.490234375,
      "completions/mean_terminated_length": 1501.673583984375,
      "completions/min_length": 330.0,
      "completions/min_terminated_length": 330.0,
      "entropy": 0.1336100473999977,
      "epoch": 0.11698717948717949,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 890.0448608398438,
      "learning_rate": 1e-06,
      "loss": 0.3014,
      "num_tokens": 91139272.0,
      "reward": 0.23745985329151154,
      "reward_std": 0.2408495545387268,
      "rewards/progression_diversity/mean": -0.14415252208709717,
      "rewards/progression_diversity/std": 0.18349310755729675,
      "rewards/symbolic_reward_accuracy/mean": 0.1953125,
      "rewards/symbolic_reward_accuracy/std": 0.3968288004398346,
      "rewards/symbolic_reward_partial_score/mean": 0.5307128429412842,
      "rewards/symbolic_reward_partial_score/std": 0.38882943987846375,
      "rewards/tag_count_reward/mean": -0.375,
      "rewards/tag_count_reward/std": 0.4845963716506958,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9764099717140198,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 438.0,
      "sampling/sampling_logp_difference/mean": 14.951321601867676,
      "step": 73
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.13262945413589478,
      "epoch": 0.11858974358974358,
      "grad_norm": 236.77122497558594,
      "learning_rate": 1e-06,
      "loss": 0.3062,
      "step": 74
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.4375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6015625,
      "entropy": 0.12846176326274872,
      "epoch": 0.1201923076923077,
      "grad_norm": 424.5876159667969,
      "learning_rate": 1e-06,
      "loss": 0.3248,
      "step": 75
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.1426731012761593,
      "epoch": 0.12179487179487179,
      "grad_norm": 137.90576171875,
      "learning_rate": 1e-06,
      "loss": 0.2592,
      "step": 76
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.400390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6630.0,
      "completions/mean_length": 7431.486328125,
      "completions/mean_terminated_length": 1453.4234619140625,
      "completions/min_length": 318.0,
      "completions/min_terminated_length": 318.0,
      "entropy": 0.1482650637626648,
      "epoch": 0.1233974358974359,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 503.6921691894531,
      "learning_rate": 1e-06,
      "loss": 0.309,
      "num_tokens": 95835905.0,
      "reward": 0.31015902757644653,
      "reward_std": 0.23790916800498962,
      "rewards/progression_diversity/mean": -0.12814107537269592,
      "rewards/progression_diversity/std": 0.17228002846240997,
      "rewards/symbolic_reward_accuracy/mean": 0.294921875,
      "rewards/symbolic_reward_accuracy/std": 0.4564536213874817,
      "rewards/symbolic_reward_partial_score/mean": 0.5648274421691895,
      "rewards/symbolic_reward_partial_score/std": 0.4122779667377472,
      "rewards/tag_count_reward/mean": -0.349609375,
      "rewards/tag_count_reward/std": 0.47731292247772217,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9831830263137817,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 440.0,
      "sampling/sampling_logp_difference/mean": 13.701435089111328,
      "step": 77
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.15608947724103928,
      "epoch": 0.125,
      "grad_norm": 502.0856628417969,
      "learning_rate": 1e-06,
      "loss": 0.2134,
      "step": 78
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.1563669890165329,
      "epoch": 0.1266025641025641,
      "grad_norm": 40.49258804321289,
      "learning_rate": 1e-06,
      "loss": 0.2021,
      "step": 79
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6640625,
      "entropy": 0.13854920864105225,
      "epoch": 0.1282051282051282,
      "grad_norm": 24.90146255493164,
      "learning_rate": 1e-06,
      "loss": 0.3625,
      "step": 80
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.376953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8946.0,
      "completions/mean_length": 7095.765625,
      "completions/mean_terminated_length": 1476.2381591796875,
      "completions/min_length": 340.0,
      "completions/min_terminated_length": 340.0,
      "entropy": 0.1516287624835968,
      "epoch": 0.12980769230769232,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1240.288818359375,
      "learning_rate": 1e-06,
      "loss": 0.3117,
      "num_tokens": 100263337.0,
      "reward": 0.2255856692790985,
      "reward_std": 0.19480521976947784,
      "rewards/progression_diversity/mean": -0.12746772170066833,
      "rewards/progression_diversity/std": 0.17732733488082886,
      "rewards/symbolic_reward_accuracy/mean": 0.154296875,
      "rewards/symbolic_reward_accuracy/std": 0.36158639192581177,
      "rewards/symbolic_reward_partial_score/mean": 0.5641438961029053,
      "rewards/symbolic_reward_partial_score/std": 0.36627063155174255,
      "rewards/tag_count_reward/mean": -0.349609375,
      "rewards/tag_count_reward/std": 0.47731292247772217,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9782713651657104,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 440.0,
      "sampling/sampling_logp_difference/mean": 14.690707206726074,
      "step": 81
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.17118209600448608,
      "epoch": 0.13141025641025642,
      "grad_norm": 143.5312042236328,
      "learning_rate": 1e-06,
      "loss": 0.2574,
      "step": 82
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.14265629649162292,
      "epoch": 0.1330128205128205,
      "grad_norm": 2867.10009765625,
      "learning_rate": 1e-06,
      "loss": 0.4231,
      "step": 83
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5859375,
      "entropy": 0.15464875847101212,
      "epoch": 0.1346153846153846,
      "grad_norm": 54.51691436767578,
      "learning_rate": 1e-06,
      "loss": 0.2703,
      "step": 84
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.48046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3804.0,
      "completions/mean_length": 8625.69921875,
      "completions/mean_terminated_length": 1450.7293701171875,
      "completions/min_length": 362.0,
      "completions/min_terminated_length": 362.0,
      "entropy": 0.13001390919089317,
      "epoch": 0.1362179487179487,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 455.9178161621094,
      "learning_rate": 1e-06,
      "loss": 0.222,
      "num_tokens": 105565695.0,
      "reward": 0.1887899488210678,
      "reward_std": 0.2059934437274933,
      "rewards/progression_diversity/mean": -0.16299909353256226,
      "rewards/progression_diversity/std": 0.18564143776893616,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.47913411259651184,
      "rewards/symbolic_reward_partial_score/std": 0.38510748744010925,
      "rewards/tag_count_reward/mean": -0.447265625,
      "rewards/tag_count_reward/std": 0.4976975917816162,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9722630977630615,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 442.0,
      "sampling/sampling_logp_difference/mean": 15.671956062316895,
      "step": 85
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.1201215460896492,
      "epoch": 0.13782051282051283,
      "grad_norm": 307.8591003417969,
      "learning_rate": 1e-06,
      "loss": 0.3693,
      "step": 86
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.12688108533620834,
      "epoch": 0.13942307692307693,
      "grad_norm": 181.98202514648438,
      "learning_rate": 1e-06,
      "loss": 0.3682,
      "step": 87
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.4140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.12507511675357819,
      "epoch": 0.14102564102564102,
      "grad_norm": 60.04415512084961,
      "learning_rate": 1e-06,
      "loss": 0.3179,
      "step": 88
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.3828125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4394.0,
      "completions/mean_length": 7166.556640625,
      "completions/mean_terminated_length": 1449.4083251953125,
      "completions/min_length": 429.0,
      "completions/min_terminated_length": 429.0,
      "entropy": 0.16381006687879562,
      "epoch": 0.14262820512820512,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 344.05682373046875,
      "learning_rate": 1e-06,
      "loss": 0.2595,
      "num_tokens": 109989372.0,
      "reward": 0.28582632541656494,
      "reward_std": 0.21509969234466553,
      "rewards/progression_diversity/mean": -0.12635371088981628,
      "rewards/progression_diversity/std": 0.17559568583965302,
      "rewards/symbolic_reward_accuracy/mean": 0.255859375,
      "rewards/symbolic_reward_accuracy/std": 0.43676990270614624,
      "rewards/symbolic_reward_partial_score/mean": 0.5617838501930237,
      "rewards/symbolic_reward_partial_score/std": 0.3981553614139557,
      "rewards/tag_count_reward/mean": -0.349609375,
      "rewards/tag_count_reward/std": 0.47731292247772217,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9799978733062744,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 444.0,
      "sampling/sampling_logp_difference/mean": 14.372432708740234,
      "step": 89
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.14315053820610046,
      "epoch": 0.14423076923076922,
      "grad_norm": 354.8428955078125,
      "learning_rate": 1e-06,
      "loss": 0.3071,
      "step": 90
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.16275375336408615,
      "epoch": 0.14583333333333334,
      "grad_norm": 645.2577514648438,
      "learning_rate": 1e-06,
      "loss": 0.3115,
      "step": 91
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.16187621653079987,
      "epoch": 0.14743589743589744,
      "grad_norm": 0.025196930393576622,
      "learning_rate": 1e-06,
      "loss": 0.2797,
      "step": 92
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.333984375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4512.0,
      "completions/mean_length": 6437.927734375,
      "completions/mean_terminated_length": 1450.307861328125,
      "completions/min_length": 304.0,
      "completions/min_terminated_length": 304.0,
      "entropy": 0.18978476524353027,
      "epoch": 0.14903846153846154,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 520.9977416992188,
      "learning_rate": 1e-06,
      "loss": 0.1873,
      "num_tokens": 114084839.0,
      "reward": 0.3511761724948883,
      "reward_std": 0.209753155708313,
      "rewards/progression_diversity/mean": -0.11236479878425598,
      "rewards/progression_diversity/std": 0.17498160898685455,
      "rewards/symbolic_reward_accuracy/mean": 0.326171875,
      "rewards/symbolic_reward_accuracy/std": 0.4692695140838623,
      "rewards/symbolic_reward_partial_score/mean": 0.6183431148529053,
      "rewards/symbolic_reward_partial_score/std": 0.38944968581199646,
      "rewards/tag_count_reward/mean": -0.2890625,
      "rewards/tag_count_reward/std": 0.45377036929130554,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9891281127929688,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 444.0,
      "sampling/sampling_logp_difference/mean": 12.81410026550293,
      "step": 93
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6171875,
      "entropy": 0.16329990327358246,
      "epoch": 0.15064102564102563,
      "grad_norm": 115.51496887207031,
      "learning_rate": 1e-06,
      "loss": 0.305,
      "step": 94
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.17047469317913055,
      "epoch": 0.15224358974358973,
      "grad_norm": 48.81769561767578,
      "learning_rate": 1e-06,
      "loss": 0.235,
      "step": 95
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.328125,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6328125,
      "entropy": 0.16202182322740555,
      "epoch": 0.15384615384615385,
      "grad_norm": 21.413990020751953,
      "learning_rate": 1e-06,
      "loss": 0.2707,
      "step": 96
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.45703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4237.0,
      "completions/mean_length": 8286.236328125,
      "completions/mean_terminated_length": 1470.1331787109375,
      "completions/min_length": 301.0,
      "completions/min_terminated_length": 301.0,
      "entropy": 0.12126778811216354,
      "epoch": 0.15544871794871795,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1298.6553955078125,
      "learning_rate": 1e-06,
      "loss": 0.3518,
      "num_tokens": 119384688.0,
      "reward": 0.15234431624412537,
      "reward_std": 0.18723168969154358,
      "rewards/progression_diversity/mean": -0.1576594114303589,
      "rewards/progression_diversity/std": 0.18727099895477295,
      "rewards/symbolic_reward_accuracy/mean": 0.099609375,
      "rewards/symbolic_reward_accuracy/std": 0.29977133870124817,
      "rewards/symbolic_reward_partial_score/mean": 0.44340819120407104,
      "rewards/symbolic_reward_partial_score/std": 0.3761260211467743,
      "rewards/tag_count_reward/mean": -0.388671875,
      "rewards/tag_count_reward/std": 0.4879252314567566,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9738816022872925,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 446.0,
      "sampling/sampling_logp_difference/mean": 15.43978500366211,
      "step": 97
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.1406988874077797,
      "epoch": 0.15705128205128205,
      "grad_norm": 550.3086547851562,
      "learning_rate": 1e-06,
      "loss": 0.2552,
      "step": 98
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6328125,
      "entropy": 0.13929974660277367,
      "epoch": 0.15865384615384615,
      "grad_norm": 161.8490447998047,
      "learning_rate": 1e-06,
      "loss": 0.2264,
      "step": 99
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.4609375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.71875,
      "entropy": 0.11794158071279526,
      "epoch": 0.16025641025641027,
      "grad_norm": 22.501441955566406,
      "learning_rate": 1e-06,
      "loss": 0.4061,
      "step": 100
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.392578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8333.0,
      "completions/mean_length": 7378.84375,
      "completions/mean_terminated_length": 1558.791015625,
      "completions/min_length": 426.0,
      "completions/min_terminated_length": 426.0,
      "entropy": 0.14221202209591866,
      "epoch": 0.16185897435897437,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 3660.10693359375,
      "learning_rate": 1e-06,
      "loss": 0.409,
      "num_tokens": 124050112.0,
      "reward": 0.18008412420749664,
      "reward_std": 0.16677263379096985,
      "rewards/progression_diversity/mean": -0.14051449298858643,
      "rewards/progression_diversity/std": 0.18938302993774414,
      "rewards/symbolic_reward_accuracy/mean": 0.115234375,
      "rewards/symbolic_reward_accuracy/std": 0.3196168541908264,
      "rewards/symbolic_reward_partial_score/mean": 0.4884277284145355,
      "rewards/symbolic_reward_partial_score/std": 0.37571844458580017,
      "rewards/tag_count_reward/mean": -0.341796875,
      "rewards/tag_count_reward/std": 0.4747757613658905,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.976874589920044,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 448.0,
      "sampling/sampling_logp_difference/mean": 15.273395538330078,
      "step": 101
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.1293311044573784,
      "epoch": 0.16346153846153846,
      "grad_norm": 400.09881591796875,
      "learning_rate": 1e-06,
      "loss": 0.3765,
      "step": 102
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.1685006469488144,
      "epoch": 0.16506410256410256,
      "grad_norm": 77.7186050415039,
      "learning_rate": 1e-06,
      "loss": 0.2434,
      "step": 103
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.359375,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6796875,
      "entropy": 0.16083669662475586,
      "epoch": 0.16666666666666666,
      "grad_norm": 67.05364990234375,
      "learning_rate": 1e-06,
      "loss": 0.2542,
      "step": 104
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.330078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8690.0,
      "completions/mean_length": 6384.462890625,
      "completions/mean_terminated_length": 1457.5772705078125,
      "completions/min_length": 347.0,
      "completions/min_terminated_length": 347.0,
      "entropy": 0.17285646498203278,
      "epoch": 0.16826923076923078,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 1558.9949951171875,
      "learning_rate": 1e-06,
      "loss": 0.1558,
      "num_tokens": 128130573.0,
      "reward": 0.2611575722694397,
      "reward_std": 0.21417993307113647,
      "rewards/progression_diversity/mean": -0.116178959608078,
      "rewards/progression_diversity/std": 0.17685534060001373,
      "rewards/symbolic_reward_accuracy/mean": 0.197265625,
      "rewards/symbolic_reward_accuracy/std": 0.3983237147331238,
      "rewards/symbolic_reward_partial_score/mean": 0.5794758796691895,
      "rewards/symbolic_reward_partial_score/std": 0.3689756989479065,
      "rewards/tag_count_reward/mean": -0.298828125,
      "rewards/tag_count_reward/std": 0.45819199085235596,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9883815050125122,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 452.0,
      "sampling/sampling_logp_difference/mean": 13.416872024536133,
      "step": 105
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.1663920134305954,
      "epoch": 0.16987179487179488,
      "grad_norm": 319.74737548828125,
      "learning_rate": 1e-06,
      "loss": 0.2595,
      "step": 106
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.18017150461673737,
      "epoch": 0.17147435897435898,
      "grad_norm": 447.3372497558594,
      "learning_rate": 1e-06,
      "loss": 0.313,
      "step": 107
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.182367742061615,
      "epoch": 0.17307692307692307,
      "grad_norm": 80.31652069091797,
      "learning_rate": 1e-06,
      "loss": 0.2265,
      "step": 108
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.400390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5179.0,
      "completions/mean_length": 7449.263671875,
      "completions/mean_terminated_length": 1483.0716552734375,
      "completions/min_length": 316.0,
      "completions/min_terminated_length": 316.0,
      "entropy": 0.1524367406964302,
      "epoch": 0.17467948717948717,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 3725.33349609375,
      "learning_rate": 1e-06,
      "loss": 0.3077,
      "num_tokens": 132834852.0,
      "reward": 0.24695852398872375,
      "reward_std": 0.17919716238975525,
      "rewards/progression_diversity/mean": -0.14106225967407227,
      "rewards/progression_diversity/std": 0.19025002419948578,
      "rewards/symbolic_reward_accuracy/mean": 0.20703125,
      "rewards/symbolic_reward_accuracy/std": 0.40557438135147095,
      "rewards/symbolic_reward_partial_score/mean": 0.5316731333732605,
      "rewards/symbolic_reward_partial_score/std": 0.3761518597602844,
      "rewards/tag_count_reward/mean": -0.353515625,
      "rewards/tag_count_reward/std": 0.47852855920791626,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9842573404312134,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 456.0,
      "sampling/sampling_logp_difference/mean": 14.218891143798828,
      "step": 109
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.14695103839039803,
      "epoch": 0.1762820512820513,
      "grad_norm": 219.70684814453125,
      "learning_rate": 1e-06,
      "loss": 0.2567,
      "step": 110
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.16737890988588333,
      "epoch": 0.1778846153846154,
      "grad_norm": 367.99285888671875,
      "learning_rate": 1e-06,
      "loss": 0.2145,
      "step": 111
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.1623668521642685,
      "epoch": 0.1794871794871795,
      "grad_norm": 359.1845703125,
      "learning_rate": 1e-06,
      "loss": 0.3114,
      "step": 112
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.298828125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5083.0,
      "completions/mean_length": 5895.5625,
      "completions/mean_terminated_length": 1425.559814453125,
      "completions/min_length": 363.0,
      "completions/min_terminated_length": 363.0,
      "entropy": 0.17023801803588867,
      "epoch": 0.18108974358974358,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 908.6353149414062,
      "learning_rate": 1e-06,
      "loss": 0.2684,
      "num_tokens": 136701844.0,
      "reward": 0.3186452388763428,
      "reward_std": 0.20103424787521362,
      "rewards/progression_diversity/mean": -0.1164332926273346,
      "rewards/progression_diversity/std": 0.18981561064720154,
      "rewards/symbolic_reward_accuracy/mean": 0.27734375,
      "rewards/symbolic_reward_accuracy/std": 0.4481254518032074,
      "rewards/symbolic_reward_partial_score/mean": 0.5979329347610474,
      "rewards/symbolic_reward_partial_score/std": 0.38331446051597595,
      "rewards/tag_count_reward/mean": -0.259765625,
      "rewards/tag_count_reward/std": 0.4389347732067108,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9887030124664307,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 456.0,
      "sampling/sampling_logp_difference/mean": 13.705371856689453,
      "step": 113
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.19520440697669983,
      "epoch": 0.18269230769230768,
      "grad_norm": 1410.1068115234375,
      "learning_rate": 1e-06,
      "loss": 0.2648,
      "step": 114
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.20632527768611908,
      "epoch": 0.1842948717948718,
      "grad_norm": 189.16241455078125,
      "learning_rate": 1e-06,
      "loss": 0.2618,
      "step": 115
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5859375,
      "entropy": 0.18539946526288986,
      "epoch": 0.1858974358974359,
      "grad_norm": 1.3296606540679932,
      "learning_rate": 1e-06,
      "loss": 0.2421,
      "step": 116
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 13409.0,
      "completions/mean_length": 6352.482421875,
      "completions/mean_terminated_length": 1453.369140625,
      "completions/min_length": 322.0,
      "completions/min_terminated_length": 322.0,
      "entropy": 0.17278361320495605,
      "epoch": 0.1875,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 269.99188232421875,
      "learning_rate": 1e-06,
      "loss": 0.2609,
      "num_tokens": 140781531.0,
      "reward": 0.28726375102996826,
      "reward_std": 0.20572300255298615,
      "rewards/progression_diversity/mean": -0.1266537606716156,
      "rewards/progression_diversity/std": 0.19276343286037445,
      "rewards/symbolic_reward_accuracy/mean": 0.236328125,
      "rewards/symbolic_reward_accuracy/std": 0.42524150013923645,
      "rewards/symbolic_reward_partial_score/mean": 0.5861165523529053,
      "rewards/symbolic_reward_partial_score/std": 0.37256380915641785,
      "rewards/tag_count_reward/mean": -0.291015625,
      "rewards/tag_count_reward/std": 0.45467492938041687,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9863637089729309,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 460.0,
      "sampling/sampling_logp_difference/mean": 14.210054397583008,
      "step": 117
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.1698644682765007,
      "epoch": 0.1891025641025641,
      "grad_norm": 174.22406005859375,
      "learning_rate": 1e-06,
      "loss": 0.2695,
      "step": 118
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.17513630539178848,
      "epoch": 0.1907051282051282,
      "grad_norm": 80.73851013183594,
      "learning_rate": 1e-06,
      "loss": 0.2233,
      "step": 119
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.1907574087381363,
      "epoch": 0.19230769230769232,
      "grad_norm": 10.501818656921387,
      "learning_rate": 1e-06,
      "loss": 0.2534,
      "step": 120
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.28515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 12543.0,
      "completions/mean_length": 5693.0859375,
      "completions/mean_terminated_length": 1428.404296875,
      "completions/min_length": 362.0,
      "completions/min_terminated_length": 362.0,
      "entropy": 0.20927876979112625,
      "epoch": 0.19391025641025642,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1835.8458251953125,
      "learning_rate": 1e-06,
      "loss": 0.2083,
      "num_tokens": 144458791.0,
      "reward": 0.3568437099456787,
      "reward_std": 0.23056712746620178,
      "rewards/progression_diversity/mean": -0.11690312623977661,
      "rewards/progression_diversity/std": 0.19476282596588135,
      "rewards/symbolic_reward_accuracy/mean": 0.326171875,
      "rewards/symbolic_reward_accuracy/std": 0.4692695140838623,
      "rewards/symbolic_reward_partial_score/mean": 0.6289225220680237,
      "rewards/symbolic_reward_partial_score/std": 0.3799564838409424,
      "rewards/tag_count_reward/mean": -0.263671875,
      "rewards/tag_count_reward/std": 0.4410543739795685,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9867503643035889,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 462.0,
      "sampling/sampling_logp_difference/mean": 14.30959415435791,
      "step": 121
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5859375,
      "entropy": 0.19147824496030807,
      "epoch": 0.1955128205128205,
      "grad_norm": 149.43600463867188,
      "learning_rate": 1e-06,
      "loss": 0.2888,
      "step": 122
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.19553864002227783,
      "epoch": 0.1971153846153846,
      "grad_norm": 161.30584716796875,
      "learning_rate": 1e-06,
      "loss": 0.2042,
      "step": 123
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.1698673516511917,
      "epoch": 0.1987179487179487,
      "grad_norm": 62.17693328857422,
      "learning_rate": 1e-06,
      "loss": 0.3143,
      "step": 124
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.349609375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5310.0,
      "completions/mean_length": 6637.630859375,
      "completions/mean_terminated_length": 1398.5916748046875,
      "completions/min_length": 349.0,
      "completions/min_terminated_length": 349.0,
      "entropy": 0.15772393345832825,
      "epoch": 0.20032051282051283,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 361.6252136230469,
      "learning_rate": 1e-06,
      "loss": 0.2752,
      "num_tokens": 148787658.0,
      "reward": 0.2472916841506958,
      "reward_std": 0.20401528477668762,
      "rewards/progression_diversity/mean": -0.13557906448841095,
      "rewards/progression_diversity/std": 0.1981465220451355,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.5566895008087158,
      "rewards/symbolic_reward_partial_score/std": 0.3679920732975006,
      "rewards/tag_count_reward/mean": -0.30859375,
      "rewards/tag_count_reward/std": 0.4623647928237915,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9800022840499878,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 462.0,
      "sampling/sampling_logp_difference/mean": 15.337078094482422,
      "step": 125
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.17300068587064743,
      "epoch": 0.20192307692307693,
      "grad_norm": 195.64173889160156,
      "learning_rate": 1e-06,
      "loss": 0.2278,
      "step": 126
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.14945074170827866,
      "epoch": 0.20352564102564102,
      "grad_norm": 388.9164733886719,
      "learning_rate": 1e-06,
      "loss": 0.3515,
      "step": 127
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.3515625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.640625,
      "entropy": 0.1672022044658661,
      "epoch": 0.20512820512820512,
      "grad_norm": 10.722322463989258,
      "learning_rate": 1e-06,
      "loss": 0.2494,
      "step": 128
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.341796875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4674.0,
      "completions/mean_length": 6536.28125,
      "completions/mean_terminated_length": 1422.480712890625,
      "completions/min_length": 355.0,
      "completions/min_terminated_length": 355.0,
      "entropy": 0.16759486496448517,
      "epoch": 0.20673076923076922,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 282.8157043457031,
      "learning_rate": 1e-06,
      "loss": 0.2464,
      "num_tokens": 153045402.0,
      "reward": 0.29630646109580994,
      "reward_std": 0.23296579718589783,
      "rewards/progression_diversity/mean": -0.1359575092792511,
      "rewards/progression_diversity/std": 0.20307843387126923,
      "rewards/symbolic_reward_accuracy/mean": 0.265625,
      "rewards/symbolic_reward_accuracy/std": 0.44209739565849304,
      "rewards/symbolic_reward_partial_score/mean": 0.555371105670929,
      "rewards/symbolic_reward_partial_score/std": 0.40097343921661377,
      "rewards/tag_count_reward/mean": -0.283203125,
      "rewards/tag_count_reward/std": 0.4509948492050171,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9835497140884399,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 464.0,
      "sampling/sampling_logp_difference/mean": 14.896586418151855,
      "step": 129
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6328125,
      "entropy": 0.1613958775997162,
      "epoch": 0.20833333333333334,
      "grad_norm": 273.4793701171875,
      "learning_rate": 1e-06,
      "loss": 0.2878,
      "step": 130
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.17489013820886612,
      "epoch": 0.20993589743589744,
      "grad_norm": 54.437557220458984,
      "learning_rate": 1e-06,
      "loss": 0.2721,
      "step": 131
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.16963838040828705,
      "epoch": 0.21153846153846154,
      "grad_norm": 174.172119140625,
      "learning_rate": 1e-06,
      "loss": 0.2341,
      "step": 132
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.365234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 14442.0,
      "completions/mean_length": 7063.861328125,
      "completions/mean_terminated_length": 1701.1968994140625,
      "completions/min_length": 464.0,
      "completions/min_terminated_length": 464.0,
      "entropy": 0.14898524433374405,
      "epoch": 0.21314102564102563,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1326.01611328125,
      "learning_rate": 1e-06,
      "loss": 0.3208,
      "num_tokens": 157607635.0,
      "reward": 0.22771768271923065,
      "reward_std": 0.22894077003002167,
      "rewards/progression_diversity/mean": -0.14180660247802734,
      "rewards/progression_diversity/std": 0.19994327425956726,
      "rewards/symbolic_reward_accuracy/mean": 0.173828125,
      "rewards/symbolic_reward_accuracy/std": 0.3793322443962097,
      "rewards/symbolic_reward_partial_score/mean": 0.5255045294761658,
      "rewards/symbolic_reward_partial_score/std": 0.3672705292701721,
      "rewards/tag_count_reward/mean": -0.328125,
      "rewards/tag_count_reward/std": 0.4699897766113281,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9785855412483215,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 464.0,
      "sampling/sampling_logp_difference/mean": 15.50029182434082,
      "step": 133
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.14145556092262268,
      "epoch": 0.21474358974358973,
      "grad_norm": 567.6675415039062,
      "learning_rate": 1e-06,
      "loss": 0.3571,
      "step": 134
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.15257297456264496,
      "epoch": 0.21634615384615385,
      "grad_norm": 3.565154552459717,
      "learning_rate": 1e-06,
      "loss": 0.2647,
      "step": 135
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.16691310703754425,
      "epoch": 0.21794871794871795,
      "grad_norm": 91.63186645507812,
      "learning_rate": 1e-06,
      "loss": 0.2332,
      "step": 136
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.349609375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3840.0,
      "completions/mean_length": 6634.4609375,
      "completions/mean_terminated_length": 1393.7177734375,
      "completions/min_length": 375.0,
      "completions/min_terminated_length": 375.0,
      "entropy": 0.16496124863624573,
      "epoch": 0.21955128205128205,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 526.1981201171875,
      "learning_rate": 1e-06,
      "loss": 0.289,
      "num_tokens": 161836447.0,
      "reward": 0.20310071110725403,
      "reward_std": 0.19233137369155884,
      "rewards/progression_diversity/mean": -0.14451980590820312,
      "rewards/progression_diversity/std": 0.2100125253200531,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.5359863042831421,
      "rewards/symbolic_reward_partial_score/std": 0.3669736087322235,
      "rewards/tag_count_reward/mean": -0.3125,
      "rewards/tag_count_reward/std": 0.4639657139778137,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9792190790176392,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 468.0,
      "sampling/sampling_logp_difference/mean": 15.816644668579102,
      "step": 137
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.16982516646385193,
      "epoch": 0.22115384615384615,
      "grad_norm": 759.0693359375,
      "learning_rate": 1e-06,
      "loss": 0.4166,
      "step": 138
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6015625,
      "entropy": 0.16247816383838654,
      "epoch": 0.22275641025641027,
      "grad_norm": 522.9348754882812,
      "learning_rate": 1e-06,
      "loss": 0.2726,
      "step": 139
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.640625,
      "entropy": 0.1550752818584442,
      "epoch": 0.22435897435897437,
      "grad_norm": 4.066892147064209,
      "learning_rate": 1e-06,
      "loss": 0.3117,
      "step": 140
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.29296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7450.0,
      "completions/mean_length": 5863.76171875,
      "completions/mean_terminated_length": 1504.5469970703125,
      "completions/min_length": 438.0,
      "completions/min_terminated_length": 438.0,
      "entropy": 0.1805713176727295,
      "epoch": 0.22596153846153846,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 949.1548461914062,
      "learning_rate": 1e-06,
      "loss": 0.2557,
      "num_tokens": 165682309.0,
      "reward": 0.20798048377037048,
      "reward_std": 0.1601024717092514,
      "rewards/progression_diversity/mean": -0.11650343239307404,
      "rewards/progression_diversity/std": 0.19056853652000427,
      "rewards/symbolic_reward_accuracy/mean": 0.1171875,
      "rewards/symbolic_reward_accuracy/std": 0.32195815443992615,
      "rewards/symbolic_reward_partial_score/mean": 0.5506673455238342,
      "rewards/symbolic_reward_partial_score/std": 0.35328954458236694,
      "rewards/tag_count_reward/mean": -0.263671875,
      "rewards/tag_count_reward/std": 0.4410543739795685,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9906647205352783,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 468.0,
      "sampling/sampling_logp_difference/mean": 13.65820598602295,
      "step": 141
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.19785359501838684,
      "epoch": 0.22756410256410256,
      "grad_norm": 413.5582580566406,
      "learning_rate": 1e-06,
      "loss": 0.3065,
      "step": 142
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.625,
      "entropy": 0.1836673691868782,
      "epoch": 0.22916666666666666,
      "grad_norm": 538.4183959960938,
      "learning_rate": 1e-06,
      "loss": 0.2451,
      "step": 143
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.19951891899108887,
      "epoch": 0.23076923076923078,
      "grad_norm": 140.01963806152344,
      "learning_rate": 1e-06,
      "loss": 0.2766,
      "step": 144
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.3046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 13188.0,
      "completions/mean_length": 6076.03515625,
      "completions/mean_terminated_length": 1559.061767578125,
      "completions/min_length": 292.0,
      "completions/min_terminated_length": 292.0,
      "entropy": 0.18072357773780823,
      "epoch": 0.23237179487179488,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 477.57244873046875,
      "learning_rate": 1e-06,
      "loss": 0.2059,
      "num_tokens": 169643863.0,
      "reward": 0.2985851764678955,
      "reward_std": 0.21802590787410736,
      "rewards/progression_diversity/mean": -0.12000055611133575,
      "rewards/progression_diversity/std": 0.19343887269496918,
      "rewards/symbolic_reward_accuracy/mean": 0.248046875,
      "rewards/symbolic_reward_accuracy/std": 0.4323015511035919,
      "rewards/symbolic_reward_partial_score/mean": 0.5884765386581421,
      "rewards/symbolic_reward_partial_score/std": 0.3711862862110138,
      "rewards/tag_count_reward/mean": -0.255859375,
      "rewards/tag_count_reward/std": 0.43676990270614624,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9889644384384155,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 468.0,
      "sampling/sampling_logp_difference/mean": 13.810128211975098,
      "step": 145
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.16814905405044556,
      "epoch": 0.23397435897435898,
      "grad_norm": 629.6782836914062,
      "learning_rate": 1e-06,
      "loss": 0.3065,
      "step": 146
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.17732855677604675,
      "epoch": 0.23557692307692307,
      "grad_norm": 60.79619216918945,
      "learning_rate": 1e-06,
      "loss": 0.2682,
      "step": 147
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.18027547001838684,
      "epoch": 0.23717948717948717,
      "grad_norm": 11.454753875732422,
      "learning_rate": 1e-06,
      "loss": 0.2156,
      "step": 148
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.291015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8313.0,
      "completions/mean_length": 5886.9921875,
      "completions/mean_terminated_length": 1578.3031005859375,
      "completions/min_length": 335.0,
      "completions/min_terminated_length": 335.0,
      "entropy": 0.20559810101985931,
      "epoch": 0.2387820512820513,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 932.6332397460938,
      "learning_rate": 1e-06,
      "loss": 0.1883,
      "num_tokens": 173516387.0,
      "reward": 0.26607826352119446,
      "reward_std": 0.2007283866405487,
      "rewards/progression_diversity/mean": -0.11971336603164673,
      "rewards/progression_diversity/std": 0.19882185757160187,
      "rewards/symbolic_reward_accuracy/mean": 0.201171875,
      "rewards/symbolic_reward_accuracy/std": 0.4012683033943176,
      "rewards/symbolic_reward_partial_score/mean": 0.5738606452941895,
      "rewards/symbolic_reward_partial_score/std": 0.3670341968536377,
      "rewards/tag_count_reward/mean": -0.255859375,
      "rewards/tag_count_reward/std": 0.43676990270614624,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9897212386131287,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 470.0,
      "sampling/sampling_logp_difference/mean": 13.849344253540039,
      "step": 149
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.18233749270439148,
      "epoch": 0.2403846153846154,
      "grad_norm": 219.96221923828125,
      "learning_rate": 1e-06,
      "loss": 0.3033,
      "step": 150
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.17845363169908524,
      "epoch": 0.2419871794871795,
      "grad_norm": 243.94744873046875,
      "learning_rate": 1e-06,
      "loss": 0.2881,
      "step": 151
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5859375,
      "entropy": 0.17516274005174637,
      "epoch": 0.24358974358974358,
      "grad_norm": 35.456787109375,
      "learning_rate": 1e-06,
      "loss": 0.2953,
      "step": 152
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.302734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5891.0,
      "completions/mean_length": 5964.396484375,
      "completions/mean_terminated_length": 1440.47900390625,
      "completions/min_length": 332.0,
      "completions/min_terminated_length": 332.0,
      "entropy": 0.17081477493047714,
      "epoch": 0.24519230769230768,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 492.8033752441406,
      "learning_rate": 1e-06,
      "loss": 0.3071,
      "num_tokens": 177494126.0,
      "reward": 0.2438780814409256,
      "reward_std": 0.16874653100967407,
      "rewards/progression_diversity/mean": -0.12586447596549988,
      "rewards/progression_diversity/std": 0.202118381857872,
      "rewards/symbolic_reward_accuracy/mean": 0.169921875,
      "rewards/symbolic_reward_accuracy/std": 0.3759314715862274,
      "rewards/symbolic_reward_partial_score/mean": 0.5612630844116211,
      "rewards/symbolic_reward_partial_score/std": 0.35993054509162903,
      "rewards/tag_count_reward/mean": -0.251953125,
      "rewards/tag_count_reward/std": 0.43455907702445984,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9910095930099487,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 472.0,
      "sampling/sampling_logp_difference/mean": 13.843090057373047,
      "step": 153
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.19071951508522034,
      "epoch": 0.2467948717948718,
      "grad_norm": 417.1937561035156,
      "learning_rate": 1e-06,
      "loss": 0.2772,
      "step": 154
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.18628785014152527,
      "epoch": 0.2483974358974359,
      "grad_norm": 27.67140007019043,
      "learning_rate": 1e-06,
      "loss": 0.2686,
      "step": 155
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.2085033506155014,
      "epoch": 0.25,
      "grad_norm": 0.0168951116502285,
      "learning_rate": 1e-06,
      "loss": 0.1791,
      "step": 156
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.255859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5164.0,
      "completions/mean_length": 5305.685546875,
      "completions/mean_terminated_length": 1496.6063232421875,
      "completions/min_length": 349.0,
      "completions/min_terminated_length": 349.0,
      "entropy": 0.18942154943943024,
      "epoch": 0.2516025641025641,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 547.8964233398438,
      "learning_rate": 1e-06,
      "loss": 0.2312,
      "num_tokens": 181116349.0,
      "reward": 0.27441728115081787,
      "reward_std": 0.20713432133197784,
      "rewards/progression_diversity/mean": -0.10222004354000092,
      "rewards/progression_diversity/std": 0.18552517890930176,
      "rewards/symbolic_reward_accuracy/mean": 0.189453125,
      "rewards/symbolic_reward_accuracy/std": 0.3922513723373413,
      "rewards/symbolic_reward_partial_score/mean": 0.6082357168197632,
      "rewards/symbolic_reward_partial_score/std": 0.34247255325317383,
      "rewards/tag_count_reward/mean": -0.20703125,
      "rewards/tag_count_reward/std": 0.40557438135147095,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9906232357025146,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 472.0,
      "sampling/sampling_logp_difference/mean": 13.893712043762207,
      "step": 157
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.20169401168823242,
      "epoch": 0.2532051282051282,
      "grad_norm": 378.7434997558594,
      "learning_rate": 1e-06,
      "loss": 0.3347,
      "step": 158
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.20617859065532684,
      "epoch": 0.2548076923076923,
      "grad_norm": 51.51784133911133,
      "learning_rate": 1e-06,
      "loss": 0.2684,
      "step": 159
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.20408597588539124,
      "epoch": 0.2564102564102564,
      "grad_norm": 19.432321548461914,
      "learning_rate": 1e-06,
      "loss": 0.1783,
      "step": 160
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.306640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6080.0,
      "completions/mean_length": 6120.22265625,
      "completions/mean_terminated_length": 1581.0308837890625,
      "completions/min_length": 403.0,
      "completions/min_terminated_length": 403.0,
      "entropy": 0.17949867248535156,
      "epoch": 0.25801282051282054,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 459.5679626464844,
      "learning_rate": 1e-06,
      "loss": 0.2659,
      "num_tokens": 185044895.0,
      "reward": 0.27867066860198975,
      "reward_std": 0.21932919323444366,
      "rewards/progression_diversity/mean": -0.12316776067018509,
      "rewards/progression_diversity/std": 0.19920648634433746,
      "rewards/symbolic_reward_accuracy/mean": 0.2265625,
      "rewards/symbolic_reward_accuracy/std": 0.4190165400505066,
      "rewards/symbolic_reward_partial_score/mean": 0.5664713382720947,
      "rewards/symbolic_reward_partial_score/std": 0.3668426275253296,
      "rewards/tag_count_reward/mean": -0.259765625,
      "rewards/tag_count_reward/std": 0.4389347732067108,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9886577129364014,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 476.0,
      "sampling/sampling_logp_difference/mean": 14.280139923095703,
      "step": 161
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.17472190409898758,
      "epoch": 0.25961538461538464,
      "grad_norm": 1472.97216796875,
      "learning_rate": 1e-06,
      "loss": 0.2911,
      "step": 162
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.17166371643543243,
      "epoch": 0.26121794871794873,
      "grad_norm": 229.58348083496094,
      "learning_rate": 1e-06,
      "loss": 0.2621,
      "step": 163
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.20802847295999527,
      "epoch": 0.26282051282051283,
      "grad_norm": 183.23561096191406,
      "learning_rate": 1e-06,
      "loss": 0.2019,
      "step": 164
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.208984375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6132.0,
      "completions/mean_length": 4608.12890625,
      "completions/mean_terminated_length": 1496.972900390625,
      "completions/min_length": 339.0,
      "completions/min_terminated_length": 339.0,
      "entropy": 0.20654169470071793,
      "epoch": 0.2644230769230769,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 412.5071716308594,
      "learning_rate": 1e-06,
      "loss": 0.2796,
      "num_tokens": 188104865.0,
      "reward": 0.3180442452430725,
      "reward_std": 0.21282601356506348,
      "rewards/progression_diversity/mean": -0.08571553975343704,
      "rewards/progression_diversity/std": 0.17538563907146454,
      "rewards/symbolic_reward_accuracy/mean": 0.24609375,
      "rewards/symbolic_reward_accuracy/std": 0.4311550557613373,
      "rewards/symbolic_reward_partial_score/mean": 0.6339681148529053,
      "rewards/symbolic_reward_partial_score/std": 0.34711363911628723,
      "rewards/tag_count_reward/mean": -0.189453125,
      "rewards/tag_count_reward/std": 0.3922513723373413,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9954339265823364,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 476.0,
      "sampling/sampling_logp_difference/mean": 13.250750541687012,
      "step": 165
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.2085501253604889,
      "epoch": 0.266025641025641,
      "grad_norm": 412.70355224609375,
      "learning_rate": 1e-06,
      "loss": 0.2375,
      "step": 166
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.328125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.21886557340621948,
      "epoch": 0.2676282051282051,
      "grad_norm": 369.0177307128906,
      "learning_rate": 1e-06,
      "loss": 0.1979,
      "step": 167
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.2201170101761818,
      "epoch": 0.2692307692307692,
      "grad_norm": 0.01998511515557766,
      "learning_rate": 1e-06,
      "loss": 0.2208,
      "step": 168
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.24609375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4548.0,
      "completions/mean_length": 5164.029296875,
      "completions/mean_terminated_length": 1501.5517578125,
      "completions/min_length": 311.0,
      "completions/min_terminated_length": 311.0,
      "entropy": 0.19382669776678085,
      "epoch": 0.2708333333333333,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 395.1278991699219,
      "learning_rate": 1e-06,
      "loss": 0.2256,
      "num_tokens": 191630160.0,
      "reward": 0.29971635341644287,
      "reward_std": 0.2082447111606598,
      "rewards/progression_diversity/mean": -0.09819044172763824,
      "rewards/progression_diversity/std": 0.18440598249435425,
      "rewards/symbolic_reward_accuracy/mean": 0.228515625,
      "rewards/symbolic_reward_accuracy/std": 0.4202871024608612,
      "rewards/symbolic_reward_partial_score/mean": 0.614306628704071,
      "rewards/symbolic_reward_partial_score/std": 0.3599141538143158,
      "rewards/tag_count_reward/mean": -0.20703125,
      "rewards/tag_count_reward/std": 0.40557438135147095,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9958318471908569,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 480.0,
      "sampling/sampling_logp_difference/mean": 13.064840316772461,
      "step": 169
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.328125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.18345515429973602,
      "epoch": 0.2724358974358974,
      "grad_norm": 296.6773986816406,
      "learning_rate": 1e-06,
      "loss": 0.2652,
      "step": 170
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3515625,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.65625,
      "entropy": 0.2129678875207901,
      "epoch": 0.27403846153846156,
      "grad_norm": 5.62185001373291,
      "learning_rate": 1e-06,
      "loss": 0.2223,
      "step": 171
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.328125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.2066056728363037,
      "epoch": 0.27564102564102566,
      "grad_norm": 7.7060699462890625,
      "learning_rate": 1e-06,
      "loss": 0.225,
      "step": 172
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.240234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4527.0,
      "completions/mean_length": 5134.7578125,
      "completions/mean_terminated_length": 1577.7994384765625,
      "completions/min_length": 383.0,
      "completions/min_terminated_length": 383.0,
      "entropy": 0.18826917558908463,
      "epoch": 0.27724358974358976,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 683.8331298828125,
      "learning_rate": 1e-06,
      "loss": 0.2989,
      "num_tokens": 195087860.0,
      "reward": 0.2965736985206604,
      "reward_std": 0.2112705409526825,
      "rewards/progression_diversity/mean": -0.09214277565479279,
      "rewards/progression_diversity/std": 0.17536333203315735,
      "rewards/symbolic_reward_accuracy/mean": 0.224609375,
      "rewards/symbolic_reward_accuracy/std": 0.41773295402526855,
      "rewards/symbolic_reward_partial_score/mean": 0.6068847179412842,
      "rewards/symbolic_reward_partial_score/std": 0.3553755581378937,
      "rewards/tag_count_reward/mean": -0.193359375,
      "rewards/tag_count_reward/std": 0.39531853795051575,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0022835731506348,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 480.0,
      "sampling/sampling_logp_difference/mean": 11.84273910522461,
      "step": 173
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.20123252272605896,
      "epoch": 0.27884615384615385,
      "grad_norm": 212.23048400878906,
      "learning_rate": 1e-06,
      "loss": 0.2336,
      "step": 174
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.20349763333797455,
      "epoch": 0.28044871794871795,
      "grad_norm": 150.0238037109375,
      "learning_rate": 1e-06,
      "loss": 0.2346,
      "step": 175
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3515625,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.22359148412942886,
      "epoch": 0.28205128205128205,
      "grad_norm": 88.31288146972656,
      "learning_rate": 1e-06,
      "loss": 0.1458,
      "step": 176
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 12719.0,
      "completions/mean_length": 5528.5,
      "completions/mean_terminated_length": 1602.04248046875,
      "completions/min_length": 312.0,
      "completions/min_terminated_length": 312.0,
      "entropy": 0.19052383303642273,
      "epoch": 0.28365384615384615,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 955.2317504882812,
      "learning_rate": 1e-06,
      "loss": 0.241,
      "num_tokens": 198768004.0,
      "reward": 0.2399653196334839,
      "reward_std": 0.19414269924163818,
      "rewards/progression_diversity/mean": -0.10552017390727997,
      "rewards/progression_diversity/std": 0.18494197726249695,
      "rewards/symbolic_reward_accuracy/mean": 0.16015625,
      "rewards/symbolic_reward_accuracy/std": 0.3671095669269562,
      "rewards/symbolic_reward_partial_score/mean": 0.5618652105331421,
      "rewards/symbolic_reward_partial_score/std": 0.350735604763031,
      "rewards/tag_count_reward/mean": -0.236328125,
      "rewards/tag_count_reward/std": 0.42524150013923645,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.989559531211853,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 482.0,
      "sampling/sampling_logp_difference/mean": 14.195697784423828,
      "step": 177
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6171875,
      "entropy": 0.17670485377311707,
      "epoch": 0.28525641025641024,
      "grad_norm": 969.2748413085938,
      "learning_rate": 1e-06,
      "loss": 0.3582,
      "step": 178
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.18852558732032776,
      "epoch": 0.28685897435897434,
      "grad_norm": 0.01699661649763584,
      "learning_rate": 1e-06,
      "loss": 0.2577,
      "step": 179
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3515625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6015625,
      "entropy": 0.19296500831842422,
      "epoch": 0.28846153846153844,
      "grad_norm": 0.020286090672016144,
      "learning_rate": 1e-06,
      "loss": 0.2662,
      "step": 180
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.25,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5431.0,
      "completions/mean_length": 5255.9375,
      "completions/mean_terminated_length": 1546.5833740234375,
      "completions/min_length": 291.0,
      "completions/min_terminated_length": 291.0,
      "entropy": 0.1944911777973175,
      "epoch": 0.2900641025641026,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 478.534423828125,
      "learning_rate": 1e-06,
      "loss": 0.2062,
      "num_tokens": 202400020.0,
      "reward": 0.23154094815254211,
      "reward_std": 0.18954092264175415,
      "rewards/progression_diversity/mean": -0.09737183153629303,
      "rewards/progression_diversity/std": 0.1798097938299179,
      "rewards/symbolic_reward_accuracy/mean": 0.134765625,
      "rewards/symbolic_reward_accuracy/std": 0.3418070077896118,
      "rewards/symbolic_reward_partial_score/mean": 0.5712727904319763,
      "rewards/symbolic_reward_partial_score/std": 0.3447929322719574,
      "rewards/tag_count_reward/mean": -0.197265625,
      "rewards/tag_count_reward/std": 0.3983237147331238,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9921056032180786,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 482.0,
      "sampling/sampling_logp_difference/mean": 13.62202262878418,
      "step": 181
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.20868180692195892,
      "epoch": 0.2916666666666667,
      "grad_norm": 51.3494758605957,
      "learning_rate": 1e-06,
      "loss": 0.2408,
      "step": 182
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.18569938838481903,
      "epoch": 0.2932692307692308,
      "grad_norm": 55.675235748291016,
      "learning_rate": 1e-06,
      "loss": 0.268,
      "step": 183
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3671875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.1851130947470665,
      "epoch": 0.2948717948717949,
      "grad_norm": 65.24762725830078,
      "learning_rate": 1e-06,
      "loss": 0.3019,
      "step": 184
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5164.0,
      "completions/mean_length": 4260.142578125,
      "completions/mean_terminated_length": 1462.329345703125,
      "completions/min_length": 310.0,
      "completions/min_terminated_length": 310.0,
      "entropy": 0.2189076542854309,
      "epoch": 0.296474358974359,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 344.04833984375,
      "learning_rate": 1e-06,
      "loss": 0.2133,
      "num_tokens": 205497869.0,
      "reward": 0.31361886858940125,
      "reward_std": 0.19572681188583374,
      "rewards/progression_diversity/mean": -0.07756727933883667,
      "rewards/progression_diversity/std": 0.17236009240150452,
      "rewards/symbolic_reward_accuracy/mean": 0.234375,
      "rewards/symbolic_reward_accuracy/std": 0.42402184009552,
      "rewards/symbolic_reward_partial_score/mean": 0.6332682371139526,
      "rewards/symbolic_reward_partial_score/std": 0.3539099097251892,
      "rewards/tag_count_reward/mean": -0.162109375,
      "rewards/tag_count_reward/std": 0.3689115643501282,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0022011995315552,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 484.0,
      "sampling/sampling_logp_difference/mean": 12.16126823425293,
      "step": 185
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.22352929413318634,
      "epoch": 0.2980769230769231,
      "grad_norm": 37.45823669433594,
      "learning_rate": 1e-06,
      "loss": 0.2166,
      "step": 186
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.2236320823431015,
      "epoch": 0.29967948717948717,
      "grad_norm": 4.9470672607421875,
      "learning_rate": 1e-06,
      "loss": 0.1972,
      "step": 187
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.22875476628541946,
      "epoch": 0.30128205128205127,
      "grad_norm": 0.024561388418078423,
      "learning_rate": 1e-06,
      "loss": 0.2051,
      "step": 188
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4226.0,
      "completions/mean_length": 3966.052734375,
      "completions/mean_terminated_length": 1388.742919921875,
      "completions/min_length": 306.0,
      "completions/min_terminated_length": 306.0,
      "entropy": 0.22446440160274506,
      "epoch": 0.30288461538461536,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 2142.739013671875,
      "learning_rate": 1e-06,
      "loss": 0.2362,
      "num_tokens": 208400968.0,
      "reward": 0.2870832681655884,
      "reward_std": 0.18633730709552765,
      "rewards/progression_diversity/mean": -0.07292823493480682,
      "rewards/progression_diversity/std": 0.16945891082286835,
      "rewards/symbolic_reward_accuracy/mean": 0.19140625,
      "rewards/symbolic_reward_accuracy/std": 0.3937928080558777,
      "rewards/symbolic_reward_partial_score/mean": 0.6227864623069763,
      "rewards/symbolic_reward_partial_score/std": 0.33123111724853516,
      "rewards/tag_count_reward/mean": -0.138671875,
      "rewards/tag_count_reward/std": 0.34594178199768066,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0030953884124756,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 484.0,
      "sampling/sampling_logp_difference/mean": 12.393356323242188,
      "step": 189
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.22635683417320251,
      "epoch": 0.30448717948717946,
      "grad_norm": 370.2960205078125,
      "learning_rate": 1e-06,
      "loss": 0.2681,
      "step": 190
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.2461889684200287,
      "epoch": 0.3060897435897436,
      "grad_norm": 0.01940600946545601,
      "learning_rate": 1e-06,
      "loss": 0.1929,
      "step": 191
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.375,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.24533099681138992,
      "epoch": 0.3076923076923077,
      "grad_norm": 0.031975965946912766,
      "learning_rate": 1e-06,
      "loss": 0.1993,
      "step": 192
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.25390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5819.0,
      "completions/mean_length": 5224.373046875,
      "completions/mean_terminated_length": 1426.59423828125,
      "completions/min_length": 297.0,
      "completions/min_terminated_length": 297.0,
      "entropy": 0.1923881694674492,
      "epoch": 0.3092948717948718,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 366.1446838378906,
      "learning_rate": 1e-06,
      "loss": 0.3055,
      "num_tokens": 212054839.0,
      "reward": 0.2073436677455902,
      "reward_std": 0.1643807291984558,
      "rewards/progression_diversity/mean": -0.1074308454990387,
      "rewards/progression_diversity/std": 0.19538284838199615,
      "rewards/symbolic_reward_accuracy/mean": 0.1171875,
      "rewards/symbolic_reward_accuracy/std": 0.32195815443992615,
      "rewards/symbolic_reward_partial_score/mean": 0.5267577767372131,
      "rewards/symbolic_reward_partial_score/std": 0.3376568853855133,
      "rewards/tag_count_reward/mean": -0.19921875,
      "rewards/tag_count_reward/std": 0.39980348944664,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9920334815979004,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 488.0,
      "sampling/sampling_logp_difference/mean": 14.228421211242676,
      "step": 193
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.2093525379896164,
      "epoch": 0.3108974358974359,
      "grad_norm": 143.63833618164062,
      "learning_rate": 1e-06,
      "loss": 0.22,
      "step": 194
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.20566636323928833,
      "epoch": 0.3125,
      "grad_norm": 507.2211608886719,
      "learning_rate": 1e-06,
      "loss": 0.2588,
      "step": 195
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.21372250467538834,
      "epoch": 0.3141025641025641,
      "grad_norm": 837.7913818359375,
      "learning_rate": 1e-06,
      "loss": 0.2089,
      "step": 196
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.22265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4510.0,
      "completions/mean_length": 4631.9921875,
      "completions/mean_terminated_length": 1265.8392333984375,
      "completions/min_length": 345.0,
      "completions/min_terminated_length": 345.0,
      "entropy": 0.22118443250656128,
      "epoch": 0.3157051282051282,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 653.4398193359375,
      "learning_rate": 1e-06,
      "loss": 0.2032,
      "num_tokens": 215352051.0,
      "reward": 0.25279223918914795,
      "reward_std": 0.19195367395877838,
      "rewards/progression_diversity/mean": -0.10407879203557968,
      "rewards/progression_diversity/std": 0.20429863035678864,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.5993651747703552,
      "rewards/symbolic_reward_partial_score/std": 0.34102150797843933,
      "rewards/tag_count_reward/mean": -0.173828125,
      "rewards/tag_count_reward/std": 0.3793322443962097,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9943366050720215,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 488.0,
      "sampling/sampling_logp_difference/mean": 14.024613380432129,
      "step": 197
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.21031095832586288,
      "epoch": 0.3173076923076923,
      "grad_norm": 9.880915641784668,
      "learning_rate": 1e-06,
      "loss": 0.2061,
      "step": 198
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.21502617746591568,
      "epoch": 0.3189102564102564,
      "grad_norm": 159.4617156982422,
      "learning_rate": 1e-06,
      "loss": 0.2279,
      "step": 199
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.2149796113371849,
      "epoch": 0.32051282051282054,
      "grad_norm": 0.021712809801101685,
      "learning_rate": 1e-06,
      "loss": 0.2856,
      "step": 200
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.30859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5767.0,
      "completions/mean_length": 5912.080078125,
      "completions/mean_terminated_length": 1238.17236328125,
      "completions/min_length": 283.0,
      "completions/min_terminated_length": 283.0,
      "entropy": 0.19441968947649002,
      "epoch": 0.32211538461538464,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 380.85797119140625,
      "learning_rate": 1e-06,
      "loss": 0.2383,
      "num_tokens": 219297724.0,
      "reward": 0.19366931915283203,
      "reward_std": 0.19040024280548096,
      "rewards/progression_diversity/mean": -0.148204505443573,
      "rewards/progression_diversity/std": 0.2301713079214096,
      "rewards/symbolic_reward_accuracy/mean": 0.099609375,
      "rewards/symbolic_reward_accuracy/std": 0.29977133870124817,
      "rewards/symbolic_reward_partial_score/mean": 0.528759777545929,
      "rewards/symbolic_reward_partial_score/std": 0.35425814986228943,
      "rewards/tag_count_reward/mean": -0.232421875,
      "rewards/tag_count_reward/std": 0.42278963327407837,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.979872465133667,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 490.0,
      "sampling/sampling_logp_difference/mean": 17.00942611694336,
      "step": 201
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6328125,
      "entropy": 0.15824927389621735,
      "epoch": 0.32371794871794873,
      "grad_norm": 166.4498291015625,
      "learning_rate": 1e-06,
      "loss": 0.3229,
      "step": 202
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.640625,
      "entropy": 0.20293635874986649,
      "epoch": 0.32532051282051283,
      "grad_norm": 0.022664187476038933,
      "learning_rate": 1e-06,
      "loss": 0.2402,
      "step": 203
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.671875,
      "entropy": 0.1808452159166336,
      "epoch": 0.3269230769230769,
      "grad_norm": 0.013569237664341927,
      "learning_rate": 1e-06,
      "loss": 0.3128,
      "step": 204
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.26171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4302.0,
      "completions/mean_length": 5205.919921875,
      "completions/mean_terminated_length": 1243.320068359375,
      "completions/min_length": 258.0,
      "completions/min_terminated_length": 258.0,
      "entropy": 0.2147229164838791,
      "epoch": 0.328525641025641,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 922.9512939453125,
      "learning_rate": 1e-06,
      "loss": 0.2163,
      "num_tokens": 222885587.0,
      "reward": 0.2559148669242859,
      "reward_std": 0.22893401980400085,
      "rewards/progression_diversity/mean": -0.12775209546089172,
      "rewards/progression_diversity/std": 0.2234182357788086,
      "rewards/symbolic_reward_accuracy/mean": 0.17578125,
      "rewards/symbolic_reward_accuracy/std": 0.3810062110424042,
      "rewards/symbolic_reward_partial_score/mean": 0.5825684070587158,
      "rewards/symbolic_reward_partial_score/std": 0.36761772632598877,
      "rewards/tag_count_reward/mean": -0.23046875,
      "rewards/tag_count_reward/std": 0.42154473066329956,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9854786396026611,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 492.0,
      "sampling/sampling_logp_difference/mean": 16.239261627197266,
      "step": 205
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.2067190259695053,
      "epoch": 0.3301282051282051,
      "grad_norm": 220.25572204589844,
      "learning_rate": 1e-06,
      "loss": 0.2657,
      "step": 206
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.22213280200958252,
      "epoch": 0.3317307692307692,
      "grad_norm": 127.26791381835938,
      "learning_rate": 1e-06,
      "loss": 0.219,
      "step": 207
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.20553400367498398,
      "epoch": 0.3333333333333333,
      "grad_norm": 14.982529640197754,
      "learning_rate": 1e-06,
      "loss": 0.2511,
      "step": 208
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.283203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4364.0,
      "completions/mean_length": 5600.46484375,
      "completions/mean_terminated_length": 1339.9400634765625,
      "completions/min_length": 266.0,
      "completions/min_terminated_length": 266.0,
      "entropy": 0.18529720604419708,
      "epoch": 0.3349358974358974,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 476.10418701171875,
      "learning_rate": 1e-06,
      "loss": 0.325,
      "num_tokens": 226612593.0,
      "reward": 0.20574292540550232,
      "reward_std": 0.2026432454586029,
      "rewards/progression_diversity/mean": -0.12785804271697998,
      "rewards/progression_diversity/std": 0.21478815376758575,
      "rewards/symbolic_reward_accuracy/mean": 0.11328125,
      "rewards/symbolic_reward_accuracy/std": 0.3172462284564972,
      "rewards/symbolic_reward_partial_score/mean": 0.5507487058639526,
      "rewards/symbolic_reward_partial_score/std": 0.3405325710773468,
      "rewards/tag_count_reward/mean": -0.26171875,
      "rewards/tag_count_reward/std": 0.44000017642974854,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.983894407749176,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 492.0,
      "sampling/sampling_logp_difference/mean": 16.456192016601562,
      "step": 209
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.19580994546413422,
      "epoch": 0.33653846153846156,
      "grad_norm": 216.3179931640625,
      "learning_rate": 1e-06,
      "loss": 0.2533,
      "step": 210
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.19553975760936737,
      "epoch": 0.33814102564102566,
      "grad_norm": 1221.7452392578125,
      "learning_rate": 1e-06,
      "loss": 0.3216,
      "step": 211
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.20991089940071106,
      "epoch": 0.33974358974358976,
      "grad_norm": 300.2898254394531,
      "learning_rate": 1e-06,
      "loss": 0.2328,
      "step": 212
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5346.0,
      "completions/mean_length": 5424.576171875,
      "completions/mean_terminated_length": 1460.5291748046875,
      "completions/min_length": 420.0,
      "completions/min_terminated_length": 420.0,
      "entropy": 0.1831618994474411,
      "epoch": 0.34134615384615385,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 961.52490234375,
      "learning_rate": 1e-06,
      "loss": 0.3222,
      "num_tokens": 230279560.0,
      "reward": 0.22739171981811523,
      "reward_std": 0.2230047881603241,
      "rewards/progression_diversity/mean": -0.12215697765350342,
      "rewards/progression_diversity/std": 0.21225501596927643,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.5380859375,
      "rewards/symbolic_reward_partial_score/std": 0.3611067831516266,
      "rewards/tag_count_reward/mean": -0.2421875,
      "rewards/tag_count_reward/std": 0.42882615327835083,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.986311137676239,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 494.0,
      "sampling/sampling_logp_difference/mean": 16.12148666381836,
      "step": 213
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.22351820021867752,
      "epoch": 0.34294871794871795,
      "grad_norm": 112.74642181396484,
      "learning_rate": 1e-06,
      "loss": 0.2065,
      "step": 214
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5859375,
      "entropy": 0.20397323369979858,
      "epoch": 0.34455128205128205,
      "grad_norm": 106.15531921386719,
      "learning_rate": 1e-06,
      "loss": 0.2369,
      "step": 215
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6328125,
      "entropy": 0.17963356524705887,
      "epoch": 0.34615384615384615,
      "grad_norm": 2.7653439044952393,
      "learning_rate": 1e-06,
      "loss": 0.2618,
      "step": 216
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.251953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4075.0,
      "completions/mean_length": 5181.009765625,
      "completions/mean_terminated_length": 1407.6788330078125,
      "completions/min_length": 382.0,
      "completions/min_terminated_length": 382.0,
      "entropy": 0.21698909252882004,
      "epoch": 0.34775641025641024,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 2655.067138671875,
      "learning_rate": 1e-06,
      "loss": 0.1933,
      "num_tokens": 233824525.0,
      "reward": 0.2999926805496216,
      "reward_std": 0.2547125220298767,
      "rewards/progression_diversity/mean": -0.12036378681659698,
      "rewards/progression_diversity/std": 0.2145136296749115,
      "rewards/symbolic_reward_accuracy/mean": 0.248046875,
      "rewards/symbolic_reward_accuracy/std": 0.4323015511035919,
      "rewards/symbolic_reward_partial_score/mean": 0.578857421875,
      "rewards/symbolic_reward_partial_score/std": 0.3651575446128845,
      "rewards/tag_count_reward/mean": -0.212890625,
      "rewards/tag_count_reward/std": 0.409751296043396,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9898908734321594,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 496.0,
      "sampling/sampling_logp_difference/mean": 15.950399398803711,
      "step": 217
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.19810165464878082,
      "epoch": 0.34935897435897434,
      "grad_norm": 90.42063903808594,
      "learning_rate": 1e-06,
      "loss": 0.2934,
      "step": 218
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.21933219581842422,
      "epoch": 0.35096153846153844,
      "grad_norm": 158.33421325683594,
      "learning_rate": 1e-06,
      "loss": 0.1813,
      "step": 219
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.22585007548332214,
      "epoch": 0.3525641025641026,
      "grad_norm": 114.27921295166016,
      "learning_rate": 1e-06,
      "loss": 0.2229,
      "step": 220
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.259765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4320.0,
      "completions/mean_length": 5331.865234375,
      "completions/mean_terminated_length": 1453.41162109375,
      "completions/min_length": 309.0,
      "completions/min_terminated_length": 309.0,
      "entropy": 0.20159848034381866,
      "epoch": 0.3541666666666667,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 953.3041381835938,
      "learning_rate": 1e-06,
      "loss": 0.2609,
      "num_tokens": 237447016.0,
      "reward": 0.2347480058670044,
      "reward_std": 0.1780080795288086,
      "rewards/progression_diversity/mean": -0.12334457039833069,
      "rewards/progression_diversity/std": 0.21702897548675537,
      "rewards/symbolic_reward_accuracy/mean": 0.146484375,
      "rewards/symbolic_reward_accuracy/std": 0.35393697023391724,
      "rewards/symbolic_reward_partial_score/mean": 0.5691568851470947,
      "rewards/symbolic_reward_partial_score/std": 0.3390918970108032,
      "rewards/tag_count_reward/mean": -0.2265625,
      "rewards/tag_count_reward/std": 0.4190165400505066,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9891093969345093,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 496.0,
      "sampling/sampling_logp_difference/mean": 16.061904907226562,
      "step": 221
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.20780060440301895,
      "epoch": 0.3557692307692308,
      "grad_norm": 171.24063110351562,
      "learning_rate": 1e-06,
      "loss": 0.2021,
      "step": 222
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.19532399624586105,
      "epoch": 0.3573717948717949,
      "grad_norm": 127.67808532714844,
      "learning_rate": 1e-06,
      "loss": 0.2701,
      "step": 223
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2121635228395462,
      "epoch": 0.358974358974359,
      "grad_norm": 36.59706497192383,
      "learning_rate": 1e-06,
      "loss": 0.19,
      "step": 224
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3714.0,
      "completions/mean_length": 4855.25390625,
      "completions/mean_terminated_length": 1326.0458984375,
      "completions/min_length": 270.0,
      "completions/min_terminated_length": 270.0,
      "entropy": 0.17904973775148392,
      "epoch": 0.3605769230769231,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 378.9632568359375,
      "learning_rate": 1e-06,
      "loss": 0.3551,
      "num_tokens": 240758634.0,
      "reward": 0.26461654901504517,
      "reward_std": 0.2048269361257553,
      "rewards/progression_diversity/mean": -0.11061383038759232,
      "rewards/progression_diversity/std": 0.20703071355819702,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.5862630605697632,
      "rewards/symbolic_reward_partial_score/std": 0.34296321868896484,
      "rewards/tag_count_reward/mean": -0.203125,
      "rewards/tag_count_reward/std": 0.4027182459831238,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9891663789749146,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 496.0,
      "sampling/sampling_logp_difference/mean": 16.460721969604492,
      "step": 225
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.20133854448795319,
      "epoch": 0.36217948717948717,
      "grad_norm": 1110.4901123046875,
      "learning_rate": 1e-06,
      "loss": 0.2465,
      "step": 226
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.22215987741947174,
      "epoch": 0.36378205128205127,
      "grad_norm": 26.63279151916504,
      "learning_rate": 1e-06,
      "loss": 0.2531,
      "step": 227
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.23379182070493698,
      "epoch": 0.36538461538461536,
      "grad_norm": 0.5691369771957397,
      "learning_rate": 1e-06,
      "loss": 0.1706,
      "step": 228
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3607.0,
      "completions/mean_length": 4385.185546875,
      "completions/mean_terminated_length": 1326.664306640625,
      "completions/min_length": 251.0,
      "completions/min_terminated_length": 251.0,
      "entropy": 0.23052329570055008,
      "epoch": 0.36698717948717946,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 360.1962585449219,
      "learning_rate": 1e-06,
      "loss": 0.2534,
      "num_tokens": 243786505.0,
      "reward": 0.3250322937965393,
      "reward_std": 0.22688430547714233,
      "rewards/progression_diversity/mean": -0.09101086109876633,
      "rewards/progression_diversity/std": 0.18869265913963318,
      "rewards/symbolic_reward_accuracy/mean": 0.25390625,
      "rewards/symbolic_reward_accuracy/std": 0.43567025661468506,
      "rewards/symbolic_reward_partial_score/mean": 0.6346517205238342,
      "rewards/symbolic_reward_partial_score/std": 0.33860599994659424,
      "rewards/tag_count_reward/mean": -0.16796875,
      "rewards/tag_count_reward/std": 0.374204158782959,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9971761703491211,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 498.0,
      "sampling/sampling_logp_difference/mean": 14.882648468017578,
      "step": 229
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.2123500406742096,
      "epoch": 0.3685897435897436,
      "grad_norm": 343.66748046875,
      "learning_rate": 1e-06,
      "loss": 0.2335,
      "step": 230
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.24668648838996887,
      "epoch": 0.3701923076923077,
      "grad_norm": 318.5328674316406,
      "learning_rate": 1e-06,
      "loss": 0.1648,
      "step": 231
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.266249381005764,
      "epoch": 0.3717948717948718,
      "grad_norm": 0.021640097722411156,
      "learning_rate": 1e-06,
      "loss": 0.163,
      "step": 232
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.232421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4106.0,
      "completions/mean_length": 4834.337890625,
      "completions/mean_terminated_length": 1337.1119384765625,
      "completions/min_length": 335.0,
      "completions/min_terminated_length": 335.0,
      "entropy": 0.22328076511621475,
      "epoch": 0.3733974358974359,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1194.5416259765625,
      "learning_rate": 1e-06,
      "loss": 0.2147,
      "num_tokens": 247135110.0,
      "reward": 0.28227871656417847,
      "reward_std": 0.21940109133720398,
      "rewards/progression_diversity/mean": -0.10660064220428467,
      "rewards/progression_diversity/std": 0.20069687068462372,
      "rewards/symbolic_reward_accuracy/mean": 0.201171875,
      "rewards/symbolic_reward_accuracy/std": 0.4012683033943176,
      "rewards/symbolic_reward_partial_score/mean": 0.6026855707168579,
      "rewards/symbolic_reward_partial_score/std": 0.3378095328807831,
      "rewards/tag_count_reward/mean": -0.181640625,
      "rewards/tag_count_reward/std": 0.38592514395713806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9878015518188477,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 500.0,
      "sampling/sampling_logp_difference/mean": 16.46212387084961,
      "step": 233
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2334306240081787,
      "epoch": 0.375,
      "grad_norm": 738.572509765625,
      "learning_rate": 1e-06,
      "loss": 0.2937,
      "step": 234
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.21482889354228973,
      "epoch": 0.3766025641025641,
      "grad_norm": 477.08978271484375,
      "learning_rate": 1e-06,
      "loss": 0.274,
      "step": 235
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.21712397038936615,
      "epoch": 0.3782051282051282,
      "grad_norm": 372.0608215332031,
      "learning_rate": 1e-06,
      "loss": 0.33,
      "step": 236
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.25390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3953.0,
      "completions/mean_length": 5198.51953125,
      "completions/mean_terminated_length": 1391.9423828125,
      "completions/min_length": 300.0,
      "completions/min_terminated_length": 300.0,
      "entropy": 0.1952293962240219,
      "epoch": 0.3798076923076923,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 534.4794311523438,
      "learning_rate": 1e-06,
      "loss": 0.223,
      "num_tokens": 250668384.0,
      "reward": 0.20283983647823334,
      "reward_std": 0.18369260430335999,
      "rewards/progression_diversity/mean": -0.1090836450457573,
      "rewards/progression_diversity/std": 0.19674594700336456,
      "rewards/symbolic_reward_accuracy/mean": 0.09765625,
      "rewards/symbolic_reward_accuracy/std": 0.29713961482048035,
      "rewards/symbolic_reward_partial_score/mean": 0.553466796875,
      "rewards/symbolic_reward_partial_score/std": 0.3304186761379242,
      "rewards/tag_count_reward/mean": -0.20703125,
      "rewards/tag_count_reward/std": 0.40557438135147095,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9892090559005737,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 500.0,
      "sampling/sampling_logp_difference/mean": 16.296913146972656,
      "step": 237
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.2330704852938652,
      "epoch": 0.3814102564102564,
      "grad_norm": 31.066238403320312,
      "learning_rate": 1e-06,
      "loss": 0.2238,
      "step": 238
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.21373382210731506,
      "epoch": 0.38301282051282054,
      "grad_norm": 120.66889953613281,
      "learning_rate": 1e-06,
      "loss": 0.2189,
      "step": 239
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.19722618907690048,
      "epoch": 0.38461538461538464,
      "grad_norm": 12.67758560180664,
      "learning_rate": 1e-06,
      "loss": 0.2972,
      "step": 240
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.19921875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4564.0,
      "completions/mean_length": 4309.216796875,
      "completions/mean_terminated_length": 1305.246337890625,
      "completions/min_length": 298.0,
      "completions/min_terminated_length": 298.0,
      "entropy": 0.23315541446208954,
      "epoch": 0.38621794871794873,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 488.22998046875,
      "learning_rate": 1e-06,
      "loss": 0.1621,
      "num_tokens": 253703711.0,
      "reward": 0.26993727684020996,
      "reward_std": 0.21065741777420044,
      "rewards/progression_diversity/mean": -0.09562888741493225,
      "rewards/progression_diversity/std": 0.19961923360824585,
      "rewards/symbolic_reward_accuracy/mean": 0.171875,
      "rewards/symbolic_reward_accuracy/std": 0.3776407241821289,
      "rewards/symbolic_reward_partial_score/mean": 0.6106607913970947,
      "rewards/symbolic_reward_partial_score/std": 0.32578223943710327,
      "rewards/tag_count_reward/mean": -0.154296875,
      "rewards/tag_count_reward/std": 0.36158639192581177,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.994120717048645,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 504.0,
      "sampling/sampling_logp_difference/mean": 15.699442863464355,
      "step": 241
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.22725315392017365,
      "epoch": 0.38782051282051283,
      "grad_norm": 728.95068359375,
      "learning_rate": 1e-06,
      "loss": 0.2635,
      "step": 242
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.24956174194812775,
      "epoch": 0.3894230769230769,
      "grad_norm": 216.6647491455078,
      "learning_rate": 1e-06,
      "loss": 0.162,
      "step": 243
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.22464978694915771,
      "epoch": 0.391025641025641,
      "grad_norm": 351.8695373535156,
      "learning_rate": 1e-06,
      "loss": 0.241,
      "step": 244
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.236328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4847.0,
      "completions/mean_length": 4997.755859375,
      "completions/mean_terminated_length": 1474.135498046875,
      "completions/min_length": 321.0,
      "completions/min_terminated_length": 321.0,
      "entropy": 0.20301750302314758,
      "epoch": 0.3926282051282051,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 481.3808898925781,
      "learning_rate": 1e-06,
      "loss": 0.2076,
      "num_tokens": 257209026.0,
      "reward": 0.2624889016151428,
      "reward_std": 0.19924888014793396,
      "rewards/progression_diversity/mean": -0.10560305416584015,
      "rewards/progression_diversity/std": 0.1964842677116394,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.5640299320220947,
      "rewards/symbolic_reward_partial_score/std": 0.3424655795097351,
      "rewards/tag_count_reward/mean": -0.181640625,
      "rewards/tag_count_reward/std": 0.38592514395713806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9906989336013794,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 504.0,
      "sampling/sampling_logp_difference/mean": 16.07154083251953,
      "step": 245
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.2440878227353096,
      "epoch": 0.3942307692307692,
      "grad_norm": 111.19556427001953,
      "learning_rate": 1e-06,
      "loss": 0.1726,
      "step": 246
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.22275415807962418,
      "epoch": 0.3958333333333333,
      "grad_norm": 237.71017456054688,
      "learning_rate": 1e-06,
      "loss": 0.2368,
      "step": 247
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.3984375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.2082279548048973,
      "epoch": 0.3974358974358974,
      "grad_norm": 15.373961448669434,
      "learning_rate": 1e-06,
      "loss": 0.3212,
      "step": 248
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4378.0,
      "completions/mean_length": 5073.48828125,
      "completions/mean_terminated_length": 1458.78857421875,
      "completions/min_length": 371.0,
      "completions/min_terminated_length": 371.0,
      "entropy": 0.2016061171889305,
      "epoch": 0.39903846153846156,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1202.8836669921875,
      "learning_rate": 1e-06,
      "loss": 0.2794,
      "num_tokens": 260669868.0,
      "reward": 0.18259669840335846,
      "reward_std": 0.16573688387870789,
      "rewards/progression_diversity/mean": -0.1065426617860794,
      "rewards/progression_diversity/std": 0.19588853418827057,
      "rewards/symbolic_reward_accuracy/mean": 0.068359375,
      "rewards/symbolic_reward_accuracy/std": 0.25260838866233826,
      "rewards/symbolic_reward_partial_score/mean": 0.5438476204872131,
      "rewards/symbolic_reward_partial_score/std": 0.32987871766090393,
      "rewards/tag_count_reward/mean": -0.205078125,
      "rewards/tag_count_reward/std": 0.4041535556316376,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9889222383499146,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 504.0,
      "sampling/sampling_logp_difference/mean": 16.378686904907227,
      "step": 249
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.2196366786956787,
      "epoch": 0.40064102564102566,
      "grad_norm": 56.70030975341797,
      "learning_rate": 1e-06,
      "loss": 0.2487,
      "step": 250
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.2119002267718315,
      "epoch": 0.40224358974358976,
      "grad_norm": 3.978065013885498,
      "learning_rate": 1e-06,
      "loss": 0.2374,
      "step": 251
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.21355099976062775,
      "epoch": 0.40384615384615385,
      "grad_norm": 0.02542886696755886,
      "learning_rate": 1e-06,
      "loss": 0.2578,
      "step": 252
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3575.0,
      "completions/mean_length": 4913.615234375,
      "completions/mean_terminated_length": 1402.27294921875,
      "completions/min_length": 301.0,
      "completions/min_terminated_length": 301.0,
      "entropy": 0.22765664756298065,
      "epoch": 0.40544871794871795,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 457.21380615234375,
      "learning_rate": 1e-06,
      "loss": 0.1611,
      "num_tokens": 264010663.0,
      "reward": 0.2566929757595062,
      "reward_std": 0.22699853777885437,
      "rewards/progression_diversity/mean": -0.10218832641839981,
      "rewards/progression_diversity/std": 0.19470685720443726,
      "rewards/symbolic_reward_accuracy/mean": 0.177734375,
      "rewards/symbolic_reward_accuracy/std": 0.3826628625392914,
      "rewards/symbolic_reward_partial_score/mean": 0.5693359375,
      "rewards/symbolic_reward_partial_score/std": 0.34047579765319824,
      "rewards/tag_count_reward/mean": -0.197265625,
      "rewards/tag_count_reward/std": 0.3983237147331238,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9911985397338867,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 16.304428100585938,
      "step": 253
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6171875,
      "entropy": 0.21794138103723526,
      "epoch": 0.40705128205128205,
      "grad_norm": 297.2830505371094,
      "learning_rate": 1e-06,
      "loss": 0.3116,
      "step": 254
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.3515625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.2055223435163498,
      "epoch": 0.40865384615384615,
      "grad_norm": 0.2487526535987854,
      "learning_rate": 1e-06,
      "loss": 0.2665,
      "step": 255
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.21468888968229294,
      "epoch": 0.41025641025641024,
      "grad_norm": 0.02222493290901184,
      "learning_rate": 1e-06,
      "loss": 0.2176,
      "step": 256
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.244140625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3774.0,
      "completions/mean_length": 5068.212890625,
      "completions/mean_terminated_length": 1413.242919921875,
      "completions/min_length": 307.0,
      "completions/min_terminated_length": 307.0,
      "entropy": 0.21585044264793396,
      "epoch": 0.41185897435897434,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 821.6210327148438,
      "learning_rate": 1e-06,
      "loss": 0.2695,
      "num_tokens": 267516868.0,
      "reward": 0.2098761796951294,
      "reward_std": 0.16895067691802979,
      "rewards/progression_diversity/mean": -0.11101497709751129,
      "rewards/progression_diversity/std": 0.20189963281154633,
      "rewards/symbolic_reward_accuracy/mean": 0.11328125,
      "rewards/symbolic_reward_accuracy/std": 0.3172462284564972,
      "rewards/symbolic_reward_partial_score/mean": 0.5431314706802368,
      "rewards/symbolic_reward_partial_score/std": 0.334158331155777,
      "rewards/tag_count_reward/mean": -0.19921875,
      "rewards/tag_count_reward/std": 0.39980348944664,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9911006689071655,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 17.146984100341797,
      "step": 257
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.22612430155277252,
      "epoch": 0.41346153846153844,
      "grad_norm": 543.9043579101562,
      "learning_rate": 1e-06,
      "loss": 0.3986,
      "step": 258
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.244947150349617,
      "epoch": 0.4150641025641026,
      "grad_norm": 313.92132568359375,
      "learning_rate": 1e-06,
      "loss": 0.1894,
      "step": 259
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.328125,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6484375,
      "entropy": 0.2207338884472847,
      "epoch": 0.4166666666666667,
      "grad_norm": 0.4175088405609131,
      "learning_rate": 1e-06,
      "loss": 0.2577,
      "step": 260
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5952.0,
      "completions/mean_length": 5339.615234375,
      "completions/mean_terminated_length": 1503.144775390625,
      "completions/min_length": 364.0,
      "completions/min_terminated_length": 364.0,
      "entropy": 0.23514293134212494,
      "epoch": 0.4182692307692308,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 860.50244140625,
      "learning_rate": 1e-06,
      "loss": 0.2181,
      "num_tokens": 271111039.0,
      "reward": 0.2324339747428894,
      "reward_std": 0.19143269956111908,
      "rewards/progression_diversity/mean": -0.11207312345504761,
      "rewards/progression_diversity/std": 0.19719642400741577,
      "rewards/symbolic_reward_accuracy/mean": 0.16015625,
      "rewards/symbolic_reward_accuracy/std": 0.3671095669269562,
      "rewards/symbolic_reward_partial_score/mean": 0.524609386920929,
      "rewards/symbolic_reward_partial_score/std": 0.3499920070171356,
      "rewards/tag_count_reward/mean": -0.19921875,
      "rewards/tag_count_reward/std": 0.39980348944664,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9931377172470093,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 16.724380493164062,
      "step": 261
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.19652889668941498,
      "epoch": 0.4198717948717949,
      "grad_norm": 72.61575317382812,
      "learning_rate": 1e-06,
      "loss": 0.2661,
      "step": 262
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.2538982629776001,
      "epoch": 0.421474358974359,
      "grad_norm": 1.3014869689941406,
      "learning_rate": 1e-06,
      "loss": 0.1651,
      "step": 263
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.3515625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.21324076503515244,
      "epoch": 0.4230769230769231,
      "grad_norm": 7.824087142944336,
      "learning_rate": 1e-06,
      "loss": 0.3078,
      "step": 264
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.23046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4048.0,
      "completions/mean_length": 4864.2734375,
      "completions/mean_terminated_length": 1414.2030029296875,
      "completions/min_length": 305.0,
      "completions/min_terminated_length": 305.0,
      "entropy": 0.24157585948705673,
      "epoch": 0.42467948717948717,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 681.8773193359375,
      "learning_rate": 1e-06,
      "loss": 0.2296,
      "num_tokens": 274522011.0,
      "reward": 0.23941701650619507,
      "reward_std": 0.200226292014122,
      "rewards/progression_diversity/mean": -0.0978497862815857,
      "rewards/progression_diversity/std": 0.1862473040819168,
      "rewards/symbolic_reward_accuracy/mean": 0.1484375,
      "rewards/symbolic_reward_accuracy/std": 0.35588082671165466,
      "rewards/symbolic_reward_partial_score/mean": 0.5630371570587158,
      "rewards/symbolic_reward_partial_score/std": 0.34202542901039124,
      "rewards/tag_count_reward/mean": -0.17578125,
      "rewards/tag_count_reward/std": 0.3810062110424042,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9933131337165833,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 16.649776458740234,
      "step": 265
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.25425516068935394,
      "epoch": 0.42628205128205127,
      "grad_norm": 0.02002408355474472,
      "learning_rate": 1e-06,
      "loss": 0.2305,
      "step": 266
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.34375,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6484375,
      "entropy": 0.22516103833913803,
      "epoch": 0.42788461538461536,
      "grad_norm": 0.027018295601010323,
      "learning_rate": 1e-06,
      "loss": 0.2392,
      "step": 267
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.237248495221138,
      "epoch": 0.42948717948717946,
      "grad_norm": 2.79491925239563,
      "learning_rate": 1e-06,
      "loss": 0.2575,
      "step": 268
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4029.0,
      "completions/mean_length": 4960.13671875,
      "completions/mean_terminated_length": 1463.03564453125,
      "completions/min_length": 328.0,
      "completions/min_terminated_length": 328.0,
      "entropy": 0.24601341784000397,
      "epoch": 0.4310897435897436,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1408.3753662109375,
      "learning_rate": 1e-06,
      "loss": 0.2201,
      "num_tokens": 278050673.0,
      "reward": 0.20770949125289917,
      "reward_std": 0.19745418429374695,
      "rewards/progression_diversity/mean": -0.10600487887859344,
      "rewards/progression_diversity/std": 0.1964379996061325,
      "rewards/symbolic_reward_accuracy/mean": 0.1171875,
      "rewards/symbolic_reward_accuracy/std": 0.32195815443992615,
      "rewards/symbolic_reward_partial_score/mean": 0.5253255367279053,
      "rewards/symbolic_reward_partial_score/std": 0.3296595513820648,
      "rewards/tag_count_reward/mean": -0.19140625,
      "rewards/tag_count_reward/std": 0.3937928080558777,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9925892949104309,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 16.926132202148438,
      "step": 269
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.24742096662521362,
      "epoch": 0.4326923076923077,
      "grad_norm": 0.18998423218727112,
      "learning_rate": 1e-06,
      "loss": 0.1687,
      "step": 270
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.23993898928165436,
      "epoch": 0.4342948717948718,
      "grad_norm": 0.016234349459409714,
      "learning_rate": 1e-06,
      "loss": 0.3282,
      "step": 271
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.22421810775995255,
      "epoch": 0.4358974358974359,
      "grad_norm": 0.023402217775583267,
      "learning_rate": 1e-06,
      "loss": 0.2581,
      "step": 272
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5016.0,
      "completions/mean_length": 4715.15234375,
      "completions/mean_terminated_length": 1595.7574462890625,
      "completions/min_length": 318.0,
      "completions/min_terminated_length": 318.0,
      "entropy": 0.2802828401327133,
      "epoch": 0.4375,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 389.8497619628906,
      "learning_rate": 1e-06,
      "loss": 0.1786,
      "num_tokens": 281394383.0,
      "reward": 0.2193732112646103,
      "reward_std": 0.20153236389160156,
      "rewards/progression_diversity/mean": -0.09344275295734406,
      "rewards/progression_diversity/std": 0.18624533712863922,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.5436035394668579,
      "rewards/symbolic_reward_partial_score/std": 0.3351518511772156,
      "rewards/tag_count_reward/mean": -0.177734375,
      "rewards/tag_count_reward/std": 0.3826628625392914,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0049525499343872,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 14.498746871948242,
      "step": 273
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.25365348160266876,
      "epoch": 0.4391025641025641,
      "grad_norm": 250.75274658203125,
      "learning_rate": 1e-06,
      "loss": 0.2038,
      "step": 274
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.2545860484242439,
      "epoch": 0.4407051282051282,
      "grad_norm": 0.01937059499323368,
      "learning_rate": 1e-06,
      "loss": 0.2308,
      "step": 275
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.23184729367494583,
      "epoch": 0.4423076923076923,
      "grad_norm": 0.2053852677345276,
      "learning_rate": 1e-06,
      "loss": 0.2823,
      "step": 276
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.251953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4437.0,
      "completions/mean_length": 5307.53125,
      "completions/mean_terminated_length": 1576.814697265625,
      "completions/min_length": 357.0,
      "completions/min_terminated_length": 357.0,
      "entropy": 0.22293350100517273,
      "epoch": 0.4439102564102564,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 594.2986450195312,
      "learning_rate": 1e-06,
      "loss": 0.3091,
      "num_tokens": 285031903.0,
      "reward": 0.19161370396614075,
      "reward_std": 0.18198563158512115,
      "rewards/progression_diversity/mean": -0.11011378467082977,
      "rewards/progression_diversity/std": 0.1955021768808365,
      "rewards/symbolic_reward_accuracy/mean": 0.091796875,
      "rewards/symbolic_reward_accuracy/std": 0.289021372795105,
      "rewards/symbolic_reward_partial_score/mean": 0.5317057371139526,
      "rewards/symbolic_reward_partial_score/std": 0.3342364728450775,
      "rewards/tag_count_reward/mean": -0.21875,
      "rewards/tag_count_reward/std": 0.41380295157432556,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9914793372154236,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 17.115394592285156,
      "step": 277
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.2029179409146309,
      "epoch": 0.44551282051282054,
      "grad_norm": 56.31192398071289,
      "learning_rate": 1e-06,
      "loss": 0.3572,
      "step": 278
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3828125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6328125,
      "entropy": 0.25365860760211945,
      "epoch": 0.44711538461538464,
      "grad_norm": 0.015985539183020592,
      "learning_rate": 1e-06,
      "loss": 0.1795,
      "step": 279
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.25643934309482574,
      "epoch": 0.44871794871794873,
      "grad_norm": 8.044807434082031,
      "learning_rate": 1e-06,
      "loss": 0.2118,
      "step": 280
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.236328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5886.0,
      "completions/mean_length": 5134.87890625,
      "completions/mean_terminated_length": 1653.693115234375,
      "completions/min_length": 457.0,
      "completions/min_terminated_length": 457.0,
      "entropy": 0.27302658557891846,
      "epoch": 0.45032051282051283,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 5191.970703125,
      "learning_rate": 1e-06,
      "loss": 0.1912,
      "num_tokens": 288559489.0,
      "reward": 0.20244288444519043,
      "reward_std": 0.1676143854856491,
      "rewards/progression_diversity/mean": -0.10288003087043762,
      "rewards/progression_diversity/std": 0.19318129122257233,
      "rewards/symbolic_reward_accuracy/mean": 0.123046875,
      "rewards/symbolic_reward_accuracy/std": 0.32881227135658264,
      "rewards/symbolic_reward_partial_score/mean": 0.5037597417831421,
      "rewards/symbolic_reward_partial_score/std": 0.34105032682418823,
      "rewards/tag_count_reward/mean": -0.21484375,
      "rewards/tag_count_reward/std": 0.4111155867576599,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0008915662765503,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 15.52652359008789,
      "step": 281
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.2505979463458061,
      "epoch": 0.4519230769230769,
      "grad_norm": 1842.8763427734375,
      "learning_rate": 1e-06,
      "loss": 0.2985,
      "step": 282
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.3671875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5859375,
      "entropy": 0.23081808537244797,
      "epoch": 0.453525641025641,
      "grad_norm": 0.18885204195976257,
      "learning_rate": 1e-06,
      "loss": 0.3087,
      "step": 283
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6171875,
      "entropy": 0.2553938925266266,
      "epoch": 0.4551282051282051,
      "grad_norm": 0.015816714614629745,
      "learning_rate": 1e-06,
      "loss": 0.224,
      "step": 284
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.22265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4693.0,
      "completions/mean_length": 4956.9609375,
      "completions/mean_terminated_length": 1683.889404296875,
      "completions/min_length": 392.0,
      "completions/min_terminated_length": 392.0,
      "entropy": 0.2509460896253586,
      "epoch": 0.4567307692307692,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 501.21697998046875,
      "learning_rate": 1e-06,
      "loss": 0.2214,
      "num_tokens": 291920845.0,
      "reward": 0.19516658782958984,
      "reward_std": 0.18093836307525635,
      "rewards/progression_diversity/mean": -0.09760037809610367,
      "rewards/progression_diversity/std": 0.18880048394203186,
      "rewards/symbolic_reward_accuracy/mean": 0.09375,
      "rewards/symbolic_reward_accuracy/std": 0.29176566004753113,
      "rewards/symbolic_reward_partial_score/mean": 0.5301106572151184,
      "rewards/symbolic_reward_partial_score/std": 0.3172612190246582,
      "rewards/tag_count_reward/mean": -0.19140625,
      "rewards/tag_count_reward/std": 0.3937928080558777,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9981337189674377,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 15.91734790802002,
      "step": 285
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3515625,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.2802078127861023,
      "epoch": 0.4583333333333333,
      "grad_norm": 113.05368041992188,
      "learning_rate": 1e-06,
      "loss": 0.2016,
      "step": 286
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.359375,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.65625,
      "entropy": 0.24139665812253952,
      "epoch": 0.4599358974358974,
      "grad_norm": 0.025173721835017204,
      "learning_rate": 1e-06,
      "loss": 0.292,
      "step": 287
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3515625,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6171875,
      "entropy": 0.27211469411849976,
      "epoch": 0.46153846153846156,
      "grad_norm": 3.040055513381958,
      "learning_rate": 1e-06,
      "loss": 0.2398,
      "step": 288
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.220703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5427.0,
      "completions/mean_length": 4934.23046875,
      "completions/mean_terminated_length": 1691.5638427734375,
      "completions/min_length": 393.0,
      "completions/min_terminated_length": 393.0,
      "entropy": 0.2553727924823761,
      "epoch": 0.46314102564102566,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1051.9158935546875,
      "learning_rate": 1e-06,
      "loss": 0.2454,
      "num_tokens": 295288899.0,
      "reward": 0.2229723334312439,
      "reward_std": 0.21836064755916595,
      "rewards/progression_diversity/mean": -0.09778685122728348,
      "rewards/progression_diversity/std": 0.1902952492237091,
      "rewards/symbolic_reward_accuracy/mean": 0.13671875,
      "rewards/symbolic_reward_accuracy/std": 0.3438861668109894,
      "rewards/symbolic_reward_partial_score/mean": 0.540771484375,
      "rewards/symbolic_reward_partial_score/std": 0.33182287216186523,
      "rewards/tag_count_reward/mean": -0.203125,
      "rewards/tag_count_reward/std": 0.4027182459831238,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.004034161567688,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 14.605140686035156,
      "step": 289
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.24111786484718323,
      "epoch": 0.46474358974358976,
      "grad_norm": 3.2952563762664795,
      "learning_rate": 1e-06,
      "loss": 0.2509,
      "step": 290
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.2719377726316452,
      "epoch": 0.46634615384615385,
      "grad_norm": 0.019442997872829437,
      "learning_rate": 1e-06,
      "loss": 0.1819,
      "step": 291
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.65625,
      "entropy": 0.2737286537885666,
      "epoch": 0.46794871794871795,
      "grad_norm": 0.02424200251698494,
      "learning_rate": 1e-06,
      "loss": 0.2014,
      "step": 292
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.16015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4168.0,
      "completions/mean_length": 4012.345703125,
      "completions/mean_terminated_length": 1653.0999755859375,
      "completions/min_length": 321.0,
      "completions/min_terminated_length": 321.0,
      "entropy": 0.2804551422595978,
      "epoch": 0.46955128205128205,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 807.3659057617188,
      "learning_rate": 1e-06,
      "loss": 0.1918,
      "num_tokens": 298231508.0,
      "reward": 0.1925782412290573,
      "reward_std": 0.16427671909332275,
      "rewards/progression_diversity/mean": -0.07176616787910461,
      "rewards/progression_diversity/std": 0.16853131353855133,
      "rewards/symbolic_reward_accuracy/mean": 0.08203125,
      "rewards/symbolic_reward_accuracy/std": 0.2746807038784027,
      "rewards/symbolic_reward_partial_score/mean": 0.5264810919761658,
      "rewards/symbolic_reward_partial_score/std": 0.3188028335571289,
      "rewards/tag_count_reward/mean": -0.138671875,
      "rewards/tag_count_reward/std": 0.34594178199768066,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0244982242584229,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 504.0,
      "sampling/sampling_logp_difference/mean": 11.09280776977539,
      "step": 293
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.30911885201931,
      "epoch": 0.47115384615384615,
      "grad_norm": 343.7190246582031,
      "learning_rate": 1e-06,
      "loss": 0.1829,
      "step": 294
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.2846119552850723,
      "epoch": 0.47275641025641024,
      "grad_norm": 650.09619140625,
      "learning_rate": 1e-06,
      "loss": 0.2431,
      "step": 295
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.3162567764520645,
      "epoch": 0.47435897435897434,
      "grad_norm": 0.03714370355010033,
      "learning_rate": 1e-06,
      "loss": 0.1759,
      "step": 296
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.228515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4304.0,
      "completions/mean_length": 5002.0859375,
      "completions/mean_terminated_length": 1630.734130859375,
      "completions/min_length": 367.0,
      "completions/min_terminated_length": 367.0,
      "entropy": 0.2860996425151825,
      "epoch": 0.47596153846153844,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 530.2379150390625,
      "learning_rate": 1e-06,
      "loss": 0.2122,
      "num_tokens": 301642864.0,
      "reward": 0.253789484500885,
      "reward_std": 0.20528803765773773,
      "rewards/progression_diversity/mean": -0.10005459934473038,
      "rewards/progression_diversity/std": 0.1889607459306717,
      "rewards/symbolic_reward_accuracy/mean": 0.189453125,
      "rewards/symbolic_reward_accuracy/std": 0.3922513723373413,
      "rewards/symbolic_reward_partial_score/mean": 0.5315917730331421,
      "rewards/symbolic_reward_partial_score/std": 0.363020122051239,
      "rewards/tag_count_reward/mean": -0.18359375,
      "rewards/tag_count_reward/std": 0.3875311613082886,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0061748027801514,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 14.482137680053711,
      "step": 297
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.27232350409030914,
      "epoch": 0.4775641025641026,
      "grad_norm": 98.60669708251953,
      "learning_rate": 1e-06,
      "loss": 0.2171,
      "step": 298
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.40625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.2478073537349701,
      "epoch": 0.4791666666666667,
      "grad_norm": 10.104867935180664,
      "learning_rate": 1e-06,
      "loss": 0.2962,
      "step": 299
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.28513647615909576,
      "epoch": 0.4807692307692308,
      "grad_norm": 0.036038998514413834,
      "learning_rate": 1e-06,
      "loss": 0.2311,
      "step": 300
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.19140625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4377.0,
      "completions/mean_length": 4445.552734375,
      "completions/mean_terminated_length": 1619.54345703125,
      "completions/min_length": 378.0,
      "completions/min_terminated_length": 378.0,
      "entropy": 0.3068581074476242,
      "epoch": 0.4823717948717949,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 330.840087890625,
      "learning_rate": 1e-06,
      "loss": 0.1263,
      "num_tokens": 304794523.0,
      "reward": 0.2764816880226135,
      "reward_std": 0.20622991025447845,
      "rewards/progression_diversity/mean": -0.08181288093328476,
      "rewards/progression_diversity/std": 0.17579428851604462,
      "rewards/symbolic_reward_accuracy/mean": 0.205078125,
      "rewards/symbolic_reward_accuracy/std": 0.4041535556316376,
      "rewards/symbolic_reward_partial_score/mean": 0.5688639283180237,
      "rewards/symbolic_reward_partial_score/std": 0.3486393988132477,
      "rewards/tag_count_reward/mean": -0.1640625,
      "rewards/tag_count_reward/std": 0.37069445848464966,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0138086080551147,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 508.0,
      "sampling/sampling_logp_difference/mean": 13.007370948791504,
      "step": 301
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2967989444732666,
      "epoch": 0.483974358974359,
      "grad_norm": 484.8408203125,
      "learning_rate": 1e-06,
      "loss": 0.2192,
      "step": 302
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.26128391921520233,
      "epoch": 0.4855769230769231,
      "grad_norm": 281.1661376953125,
      "learning_rate": 1e-06,
      "loss": 0.2839,
      "step": 303
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.2628230154514313,
      "epoch": 0.48717948717948717,
      "grad_norm": 0.027397677302360535,
      "learning_rate": 1e-06,
      "loss": 0.2977,
      "step": 304
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.19140625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4609.0,
      "completions/mean_length": 4459.234375,
      "completions/mean_terminated_length": 1636.4637451171875,
      "completions/min_length": 363.0,
      "completions/min_terminated_length": 363.0,
      "entropy": 0.28316499292850494,
      "epoch": 0.48878205128205127,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 847.23779296875,
      "learning_rate": 1e-06,
      "loss": 0.2316,
      "num_tokens": 307925859.0,
      "reward": 0.25070130825042725,
      "reward_std": 0.16120730340480804,
      "rewards/progression_diversity/mean": -0.08758464455604553,
      "rewards/progression_diversity/std": 0.18611328303813934,
      "rewards/symbolic_reward_accuracy/mean": 0.1640625,
      "rewards/symbolic_reward_accuracy/std": 0.37069445848464966,
      "rewards/symbolic_reward_partial_score/mean": 0.5677571296691895,
      "rewards/symbolic_reward_partial_score/std": 0.333774209022522,
      "rewards/tag_count_reward/mean": -0.171875,
      "rewards/tag_count_reward/std": 0.3776407241821289,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0140712261199951,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 512.0,
      "sampling/sampling_logp_difference/mean": 13.223786354064941,
      "step": 305
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.375,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.2871243506669998,
      "epoch": 0.49038461538461536,
      "grad_norm": 393.9886474609375,
      "learning_rate": 1e-06,
      "loss": 0.2205,
      "step": 306
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.3013138175010681,
      "epoch": 0.49198717948717946,
      "grad_norm": 0.020085537806153297,
      "learning_rate": 1e-06,
      "loss": 0.1944,
      "step": 307
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.2749166488647461,
      "epoch": 0.4935897435897436,
      "grad_norm": 0.01898149587213993,
      "learning_rate": 1e-06,
      "loss": 0.2895,
      "step": 308
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.16796875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5109.0,
      "completions/mean_length": 4079.05859375,
      "completions/mean_terminated_length": 1594.9625244140625,
      "completions/min_length": 361.0,
      "completions/min_terminated_length": 361.0,
      "entropy": 0.2986243665218353,
      "epoch": 0.4951923076923077,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 579.7413330078125,
      "learning_rate": 1e-06,
      "loss": 0.2209,
      "num_tokens": 310795841.0,
      "reward": 0.24147483706474304,
      "reward_std": 0.18043410778045654,
      "rewards/progression_diversity/mean": -0.0688263550400734,
      "rewards/progression_diversity/std": 0.15821245312690735,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.589111328125,
      "rewards/symbolic_reward_partial_score/std": 0.3147651255130768,
      "rewards/tag_count_reward/mean": -0.142578125,
      "rewards/tag_count_reward/std": 0.3499840497970581,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0225772857666016,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 512.0,
      "sampling/sampling_logp_difference/mean": 11.870687484741211,
      "step": 309
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.3223755210638046,
      "epoch": 0.4967948717948718,
      "grad_norm": 256.777099609375,
      "learning_rate": 1e-06,
      "loss": 0.1514,
      "step": 310
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.3142601400613785,
      "epoch": 0.4983974358974359,
      "grad_norm": 0.41229262948036194,
      "learning_rate": 1e-06,
      "loss": 0.1924,
      "step": 311
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.29151175916194916,
      "epoch": 0.5,
      "grad_norm": 0.013734478503465652,
      "learning_rate": 1e-06,
      "loss": 0.2275,
      "step": 312
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.150390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4919.0,
      "completions/mean_length": 3832.859375,
      "completions/mean_terminated_length": 1611.1632080078125,
      "completions/min_length": 374.0,
      "completions/min_terminated_length": 374.0,
      "entropy": 0.29491689801216125,
      "epoch": 0.5016025641025641,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 841.0869140625,
      "learning_rate": 1e-06,
      "loss": 0.2169,
      "num_tokens": 313608521.0,
      "reward": 0.32326555252075195,
      "reward_std": 0.19986380636692047,
      "rewards/progression_diversity/mean": -0.06456150114536285,
      "rewards/progression_diversity/std": 0.15824860334396362,
      "rewards/symbolic_reward_accuracy/mean": 0.251953125,
      "rewards/symbolic_reward_accuracy/std": 0.43455907702445984,
      "rewards/symbolic_reward_partial_score/mean": 0.6194173097610474,
      "rewards/symbolic_reward_partial_score/std": 0.3335736393928528,
      "rewards/tag_count_reward/mean": -0.130859375,
      "rewards/tag_count_reward/std": 0.33757632970809937,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.027578353881836,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 512.0,
      "sampling/sampling_logp_difference/mean": 10.916871070861816,
      "step": 313
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.33330613374710083,
      "epoch": 0.5032051282051282,
      "grad_norm": 125.96926879882812,
      "learning_rate": 1e-06,
      "loss": 0.1417,
      "step": 314
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2990605980157852,
      "epoch": 0.5048076923076923,
      "grad_norm": 0.028910215944051743,
      "learning_rate": 1e-06,
      "loss": 0.1997,
      "step": 315
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.31929296255111694,
      "epoch": 0.5064102564102564,
      "grad_norm": 114.29574584960938,
      "learning_rate": 1e-06,
      "loss": 0.1296,
      "step": 316
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6009.0,
      "completions/mean_length": 4275.66015625,
      "completions/mean_terminated_length": 1762.6085205078125,
      "completions/min_length": 381.0,
      "completions/min_terminated_length": 381.0,
      "entropy": 0.32325243949890137,
      "epoch": 0.5080128205128205,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 370.2263488769531,
      "learning_rate": 1e-06,
      "loss": 0.1346,
      "num_tokens": 316649083.0,
      "reward": 0.24201995134353638,
      "reward_std": 0.18078970909118652,
      "rewards/progression_diversity/mean": -0.06900203227996826,
      "rewards/progression_diversity/std": 0.1568773090839386,
      "rewards/symbolic_reward_accuracy/mean": 0.142578125,
      "rewards/symbolic_reward_accuracy/std": 0.3499840497970581,
      "rewards/symbolic_reward_partial_score/mean": 0.5694499015808105,
      "rewards/symbolic_reward_partial_score/std": 0.31546229124069214,
      "rewards/tag_count_reward/mean": -0.13671875,
      "rewards/tag_count_reward/std": 0.3438861668109894,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0173547267913818,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 512.0,
      "sampling/sampling_logp_difference/mean": 12.444093704223633,
      "step": 317
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.2793115973472595,
      "epoch": 0.5096153846153846,
      "grad_norm": 7.101436614990234,
      "learning_rate": 1e-06,
      "loss": 0.2415,
      "step": 318
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.2891732454299927,
      "epoch": 0.5112179487179487,
      "grad_norm": 0.021407851949334145,
      "learning_rate": 1e-06,
      "loss": 0.2313,
      "step": 319
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.28606168925762177,
      "epoch": 0.5128205128205128,
      "grad_norm": 0.01678905449807644,
      "learning_rate": 1e-06,
      "loss": 0.2457,
      "step": 320
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.15234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4942.0,
      "completions/mean_length": 3954.216796875,
      "completions/mean_terminated_length": 1720.2926025390625,
      "completions/min_length": 366.0,
      "completions/min_terminated_length": 366.0,
      "entropy": 0.2966272532939911,
      "epoch": 0.5144230769230769,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 758.809814453125,
      "learning_rate": 1e-06,
      "loss": 0.2448,
      "num_tokens": 319542234.0,
      "reward": 0.23621943593025208,
      "reward_std": 0.17420727014541626,
      "rewards/progression_diversity/mean": -0.062140051275491714,
      "rewards/progression_diversity/std": 0.15392348170280457,
      "rewards/symbolic_reward_accuracy/mean": 0.126953125,
      "rewards/symbolic_reward_accuracy/std": 0.33324605226516724,
      "rewards/symbolic_reward_partial_score/mean": 0.5752767324447632,
      "rewards/symbolic_reward_partial_score/std": 0.3146626949310303,
      "rewards/tag_count_reward/mean": -0.119140625,
      "rewards/tag_count_reward/std": 0.32427072525024414,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0240211486816406,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 512.0,
      "sampling/sampling_logp_difference/mean": 11.527631759643555,
      "step": 321
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.33031274378299713,
      "epoch": 0.5160256410256411,
      "grad_norm": 0.021069565787911415,
      "learning_rate": 1e-06,
      "loss": 0.1706,
      "step": 322
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.3213012218475342,
      "epoch": 0.5176282051282052,
      "grad_norm": 0.34873709082603455,
      "learning_rate": 1e-06,
      "loss": 0.1742,
      "step": 323
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.30144381523132324,
      "epoch": 0.5192307692307693,
      "grad_norm": 0.02283627726137638,
      "learning_rate": 1e-06,
      "loss": 0.2115,
      "step": 324
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1796875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4747.0,
      "completions/mean_length": 4465.17578125,
      "completions/mean_terminated_length": 1854.3857421875,
      "completions/min_length": 530.0,
      "completions/min_terminated_length": 530.0,
      "entropy": 0.32395896315574646,
      "epoch": 0.5208333333333334,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 343.3765563964844,
      "learning_rate": 1e-06,
      "loss": 0.1178,
      "num_tokens": 322758580.0,
      "reward": 0.20879912376403809,
      "reward_std": 0.1630467027425766,
      "rewards/progression_diversity/mean": -0.07272521406412125,
      "rewards/progression_diversity/std": 0.16325299441814423,
      "rewards/symbolic_reward_accuracy/mean": 0.099609375,
      "rewards/symbolic_reward_accuracy/std": 0.29977133870124817,
      "rewards/symbolic_reward_partial_score/mean": 0.5441243648529053,
      "rewards/symbolic_reward_partial_score/std": 0.3322213888168335,
      "rewards/tag_count_reward/mean": -0.134765625,
      "rewards/tag_count_reward/std": 0.3418070077896118,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0155704021453857,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 512.0,
      "sampling/sampling_logp_difference/mean": 12.669414520263672,
      "step": 325
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.3014257103204727,
      "epoch": 0.5224358974358975,
      "grad_norm": 0.020518580451607704,
      "learning_rate": 1e-06,
      "loss": 0.2602,
      "step": 326
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.328125,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.29575052857398987,
      "epoch": 0.5240384615384616,
      "grad_norm": 530.0026245117188,
      "learning_rate": 1e-06,
      "loss": 0.2193,
      "step": 327
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.3828125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.2522171512246132,
      "epoch": 0.5256410256410257,
      "grad_norm": 0.015645645558834076,
      "learning_rate": 1e-06,
      "loss": 0.3602,
      "step": 328
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5689.0,
      "completions/mean_length": 4896.400390625,
      "completions/mean_terminated_length": 1968.1888427734375,
      "completions/min_length": 432.0,
      "completions/min_terminated_length": 432.0,
      "entropy": 0.28659459948539734,
      "epoch": 0.5272435897435898,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 472.4561462402344,
      "learning_rate": 1e-06,
      "loss": 0.1946,
      "num_tokens": 326107713.0,
      "reward": 0.27258506417274475,
      "reward_std": 0.22720691561698914,
      "rewards/progression_diversity/mean": -0.0842689722776413,
      "rewards/progression_diversity/std": 0.17504683136940002,
      "rewards/symbolic_reward_accuracy/mean": 0.20703125,
      "rewards/symbolic_reward_accuracy/std": 0.40557438135147095,
      "rewards/symbolic_reward_partial_score/mean": 0.554003894329071,
      "rewards/symbolic_reward_partial_score/std": 0.3567145764827728,
      "rewards/tag_count_reward/mean": -0.169921875,
      "rewards/tag_count_reward/std": 0.3759314715862274,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.014817714691162,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 512.0,
      "sampling/sampling_logp_difference/mean": 12.392074584960938,
      "step": 329
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.2568201795220375,
      "epoch": 0.5288461538461539,
      "grad_norm": 0.02229255437850952,
      "learning_rate": 1e-06,
      "loss": 0.2685,
      "step": 330
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.2770218998193741,
      "epoch": 0.530448717948718,
      "grad_norm": 0.020292697474360466,
      "learning_rate": 1e-06,
      "loss": 0.2106,
      "step": 331
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.27906210720539093,
      "epoch": 0.532051282051282,
      "grad_norm": 0.01652703620493412,
      "learning_rate": 1e-06,
      "loss": 0.2365,
      "step": 332
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.240234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6400.0,
      "completions/mean_length": 5536.8125,
      "completions/mean_terminated_length": 2106.98193359375,
      "completions/min_length": 382.0,
      "completions/min_terminated_length": 382.0,
      "entropy": 0.24760161340236664,
      "epoch": 0.5336538461538461,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 698.7391967773438,
      "learning_rate": 1e-06,
      "loss": 0.2491,
      "num_tokens": 329775969.0,
      "reward": 0.1934526264667511,
      "reward_std": 0.1800420880317688,
      "rewards/progression_diversity/mean": -0.09467939287424088,
      "rewards/progression_diversity/std": 0.17836976051330566,
      "rewards/symbolic_reward_accuracy/mean": 0.107421875,
      "rewards/symbolic_reward_accuracy/std": 0.30995169281959534,
      "rewards/symbolic_reward_partial_score/mean": 0.49760740995407104,
      "rewards/symbolic_reward_partial_score/std": 0.3303731083869934,
      "rewards/tag_count_reward/mean": -0.193359375,
      "rewards/tag_count_reward/std": 0.39531853795051575,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0133090019226074,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 12.529105186462402,
      "step": 333
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.27296438813209534,
      "epoch": 0.5352564102564102,
      "grad_norm": 238.9752960205078,
      "learning_rate": 1e-06,
      "loss": 0.2225,
      "step": 334
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.248204804956913,
      "epoch": 0.5368589743589743,
      "grad_norm": 106.70677185058594,
      "learning_rate": 1e-06,
      "loss": 0.2809,
      "step": 335
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.2727896124124527,
      "epoch": 0.5384615384615384,
      "grad_norm": 0.031822893768548965,
      "learning_rate": 1e-06,
      "loss": 0.2527,
      "step": 336
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5026.0,
      "completions/mean_length": 5709.119140625,
      "completions/mean_terminated_length": 2001.002685546875,
      "completions/min_length": 427.0,
      "completions/min_terminated_length": 427.0,
      "entropy": 0.25325363129377365,
      "epoch": 0.5400641025641025,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 363.06060791015625,
      "learning_rate": 1e-06,
      "loss": 0.2604,
      "num_tokens": 333527678.0,
      "reward": 0.23530638217926025,
      "reward_std": 0.22434313595294952,
      "rewards/progression_diversity/mean": -0.1065697893500328,
      "rewards/progression_diversity/std": 0.18764851987361908,
      "rewards/symbolic_reward_accuracy/mean": 0.171875,
      "rewards/symbolic_reward_accuracy/std": 0.3776407241821289,
      "rewards/symbolic_reward_partial_score/mean": 0.5112141966819763,
      "rewards/symbolic_reward_partial_score/std": 0.3657578229904175,
      "rewards/tag_count_reward/mean": -0.201171875,
      "rewards/tag_count_reward/std": 0.4012683033943176,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.006093144416809,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 14.130290031433105,
      "step": 337
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.24173004925251007,
      "epoch": 0.5416666666666666,
      "grad_norm": 319.6545715332031,
      "learning_rate": 1e-06,
      "loss": 0.2882,
      "step": 338
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.2622825503349304,
      "epoch": 0.5432692307692307,
      "grad_norm": 219.5463409423828,
      "learning_rate": 1e-06,
      "loss": 0.2093,
      "step": 339
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.2560427188873291,
      "epoch": 0.5448717948717948,
      "grad_norm": 0.026842230930924416,
      "learning_rate": 1e-06,
      "loss": 0.2614,
      "step": 340
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.279296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5854.0,
      "completions/mean_length": 6195.84765625,
      "completions/mean_terminated_length": 2247.593505859375,
      "completions/min_length": 562.0,
      "completions/min_terminated_length": 562.0,
      "entropy": 0.23094411194324493,
      "epoch": 0.5464743589743589,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 853.015625,
      "learning_rate": 1e-06,
      "loss": 0.3191,
      "num_tokens": 337591040.0,
      "reward": 0.1426115334033966,
      "reward_std": 0.17017614841461182,
      "rewards/progression_diversity/mean": -0.10945233702659607,
      "rewards/progression_diversity/std": 0.18531769514083862,
      "rewards/symbolic_reward_accuracy/mean": 0.064453125,
      "rewards/symbolic_reward_accuracy/std": 0.24579854309558868,
      "rewards/symbolic_reward_partial_score/mean": 0.4288899898529053,
      "rewards/symbolic_reward_partial_score/std": 0.330017626285553,
      "rewards/tag_count_reward/mean": -0.236328125,
      "rewards/tag_count_reward/std": 0.42524150013923645,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0073082447052002,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 13.80842399597168,
      "step": 341
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.2576480060815811,
      "epoch": 0.5480769230769231,
      "grad_norm": 14.71423053741455,
      "learning_rate": 1e-06,
      "loss": 0.2245,
      "step": 342
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.3515625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.2539537325501442,
      "epoch": 0.5496794871794872,
      "grad_norm": 0.01885385625064373,
      "learning_rate": 1e-06,
      "loss": 0.2631,
      "step": 343
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.2588704973459244,
      "epoch": 0.5512820512820513,
      "grad_norm": 0.022954288870096207,
      "learning_rate": 1e-06,
      "loss": 0.2359,
      "step": 344
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.25,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5730.0,
      "completions/mean_length": 5692.001953125,
      "completions/mean_terminated_length": 2128.002685546875,
      "completions/min_length": 496.0,
      "completions/min_terminated_length": 496.0,
      "entropy": 0.28620630502700806,
      "epoch": 0.5528846153846154,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 664.2322998046875,
      "learning_rate": 1e-06,
      "loss": 0.2053,
      "num_tokens": 341301089.0,
      "reward": 0.21951040625572205,
      "reward_std": 0.22114171087741852,
      "rewards/progression_diversity/mean": -0.09241549670696259,
      "rewards/progression_diversity/std": 0.17231649160385132,
      "rewards/symbolic_reward_accuracy/mean": 0.1484375,
      "rewards/symbolic_reward_accuracy/std": 0.35588082671165466,
      "rewards/symbolic_reward_partial_score/mean": 0.5049642324447632,
      "rewards/symbolic_reward_partial_score/std": 0.3561265468597412,
      "rewards/tag_count_reward/mean": -0.201171875,
      "rewards/tag_count_reward/std": 0.4012683033943176,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0092263221740723,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 13.769186019897461,
      "step": 345
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.2516386955976486,
      "epoch": 0.5544871794871795,
      "grad_norm": 547.4846801757812,
      "learning_rate": 1e-06,
      "loss": 0.3033,
      "step": 346
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.2582557052373886,
      "epoch": 0.5560897435897436,
      "grad_norm": 0.3687719404697418,
      "learning_rate": 1e-06,
      "loss": 0.2881,
      "step": 347
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.25732411444187164,
      "epoch": 0.5576923076923077,
      "grad_norm": 0.022600045427680016,
      "learning_rate": 1e-06,
      "loss": 0.1909,
      "step": 348
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.23046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5114.0,
      "completions/mean_length": 5463.091796875,
      "completions/mean_terminated_length": 2192.36279296875,
      "completions/min_length": 562.0,
      "completions/min_terminated_length": 562.0,
      "entropy": 0.2584301382303238,
      "epoch": 0.5592948717948718,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1072.4036865234375,
      "learning_rate": 1e-06,
      "loss": 0.2856,
      "num_tokens": 344947984.0,
      "reward": 0.17411813139915466,
      "reward_std": 0.17677854001522064,
      "rewards/progression_diversity/mean": -0.08086328208446503,
      "rewards/progression_diversity/std": 0.1584118902683258,
      "rewards/symbolic_reward_accuracy/mean": 0.083984375,
      "rewards/symbolic_reward_accuracy/std": 0.2776356339454651,
      "rewards/symbolic_reward_partial_score/mean": 0.4769693911075592,
      "rewards/symbolic_reward_partial_score/std": 0.3174276053905487,
      "rewards/tag_count_reward/mean": -0.185546875,
      "rewards/tag_count_reward/std": 0.38912075757980347,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.015385389328003,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 12.646538734436035,
      "step": 349
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.28206463158130646,
      "epoch": 0.5608974358974359,
      "grad_norm": 23.78731918334961,
      "learning_rate": 1e-06,
      "loss": 0.2093,
      "step": 350
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.2668316960334778,
      "epoch": 0.5625,
      "grad_norm": 0.019982578232884407,
      "learning_rate": 1e-06,
      "loss": 0.2188,
      "step": 351
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.27186111360788345,
      "epoch": 0.5641025641025641,
      "grad_norm": 0.04332799091935158,
      "learning_rate": 1e-06,
      "loss": 0.2327,
      "step": 352
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.255859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5015.0,
      "completions/mean_length": 5713.78515625,
      "completions/mean_terminated_length": 2045.0235595703125,
      "completions/min_length": 365.0,
      "completions/min_terminated_length": 365.0,
      "entropy": 0.24826274812221527,
      "epoch": 0.5657051282051282,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 2501.171875,
      "learning_rate": 1e-06,
      "loss": 0.2487,
      "num_tokens": 348796562.0,
      "reward": 0.2261546105146408,
      "reward_std": 0.22966080904006958,
      "rewards/progression_diversity/mean": -0.09401117265224457,
      "rewards/progression_diversity/std": 0.1731572151184082,
      "rewards/symbolic_reward_accuracy/mean": 0.166015625,
      "rewards/symbolic_reward_accuracy/std": 0.3724585771560669,
      "rewards/symbolic_reward_partial_score/mean": 0.4926595091819763,
      "rewards/symbolic_reward_partial_score/std": 0.3556332588195801,
      "rewards/tag_count_reward/mean": -0.203125,
      "rewards/tag_count_reward/std": 0.4027182459831238,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0071918964385986,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 14.334271430969238,
      "step": 353
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.23615773022174835,
      "epoch": 0.5673076923076923,
      "grad_norm": 180.40103149414062,
      "learning_rate": 1e-06,
      "loss": 0.3324,
      "step": 354
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.2617553621530533,
      "epoch": 0.5689102564102564,
      "grad_norm": 15.31356143951416,
      "learning_rate": 1e-06,
      "loss": 0.2533,
      "step": 355
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.28039678931236267,
      "epoch": 0.5705128205128205,
      "grad_norm": 0.03650696948170662,
      "learning_rate": 1e-06,
      "loss": 0.209,
      "step": 356
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.30859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6289.0,
      "completions/mean_length": 6547.625,
      "completions/mean_terminated_length": 2157.37841796875,
      "completions/min_length": 431.0,
      "completions/min_terminated_length": 431.0,
      "entropy": 0.2463904321193695,
      "epoch": 0.5721153846153846,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 776.6242065429688,
      "learning_rate": 1e-06,
      "loss": 0.2591,
      "num_tokens": 353049346.0,
      "reward": 0.15728139877319336,
      "reward_std": 0.19684484601020813,
      "rewards/progression_diversity/mean": -0.11365720629692078,
      "rewards/progression_diversity/std": 0.1853574961423874,
      "rewards/symbolic_reward_accuracy/mean": 0.083984375,
      "rewards/symbolic_reward_accuracy/std": 0.2776356339454651,
      "rewards/symbolic_reward_partial_score/mean": 0.4564453363418579,
      "rewards/symbolic_reward_partial_score/std": 0.3446153700351715,
      "rewards/tag_count_reward/mean": -0.2890625,
      "rewards/tag_count_reward/std": 0.45377036929130554,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0021424293518066,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 15.470850944519043,
      "step": 357
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.23972290009260178,
      "epoch": 0.5737179487179487,
      "grad_norm": 157.12100219726562,
      "learning_rate": 1e-06,
      "loss": 0.2861,
      "step": 358
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.2611324340105057,
      "epoch": 0.5753205128205128,
      "grad_norm": 0.026206420734524727,
      "learning_rate": 1e-06,
      "loss": 0.2928,
      "step": 359
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6328125,
      "entropy": 0.2634492665529251,
      "epoch": 0.5769230769230769,
      "grad_norm": 6.0313310623168945,
      "learning_rate": 1e-06,
      "loss": 0.346,
      "step": 360
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2890625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5884.0,
      "completions/mean_length": 6286.490234375,
      "completions/mean_terminated_length": 2180.909423828125,
      "completions/min_length": 398.0,
      "completions/min_terminated_length": 398.0,
      "entropy": 0.2577165514230728,
      "epoch": 0.5785256410256411,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 268.4336853027344,
      "learning_rate": 1e-06,
      "loss": 0.2247,
      "num_tokens": 357130125.0,
      "reward": 0.1658135950565338,
      "reward_std": 0.19578394293785095,
      "rewards/progression_diversity/mean": -0.116884745657444,
      "rewards/progression_diversity/std": 0.19544658064842224,
      "rewards/symbolic_reward_accuracy/mean": 0.09375,
      "rewards/symbolic_reward_accuracy/std": 0.29176566004753113,
      "rewards/symbolic_reward_partial_score/mean": 0.45634764432907104,
      "rewards/symbolic_reward_partial_score/std": 0.3385258913040161,
      "rewards/tag_count_reward/mean": -0.26171875,
      "rewards/tag_count_reward/std": 0.44000017642974854,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.001070261001587,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 520.0,
      "sampling/sampling_logp_difference/mean": 15.70700454711914,
      "step": 361
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.21536707878112793,
      "epoch": 0.5801282051282052,
      "grad_norm": 2007.2310791015625,
      "learning_rate": 1e-06,
      "loss": 0.3851,
      "step": 362
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.59375,
      "entropy": 0.23036614060401917,
      "epoch": 0.5817307692307693,
      "grad_norm": 188.44676208496094,
      "learning_rate": 1e-06,
      "loss": 0.3252,
      "step": 363
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5859375,
      "entropy": 0.24637820571660995,
      "epoch": 0.5833333333333334,
      "grad_norm": 0.025239424780011177,
      "learning_rate": 1e-06,
      "loss": 0.2805,
      "step": 364
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.3359375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5414.0,
      "completions/mean_length": 7036.28125,
      "completions/mean_terminated_length": 2307.435302734375,
      "completions/min_length": 542.0,
      "completions/min_terminated_length": 542.0,
      "entropy": 0.2505442202091217,
      "epoch": 0.5849358974358975,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 329.7117614746094,
      "learning_rate": 1e-06,
      "loss": 0.1805,
      "num_tokens": 361515437.0,
      "reward": 0.15802721679210663,
      "reward_std": 0.16592083871364594,
      "rewards/progression_diversity/mean": -0.14015081524848938,
      "rewards/progression_diversity/std": 0.20965632796287537,
      "rewards/symbolic_reward_accuracy/mean": 0.080078125,
      "rewards/symbolic_reward_accuracy/std": 0.271679550409317,
      "rewards/symbolic_reward_partial_score/mean": 0.47153323888778687,
      "rewards/symbolic_reward_partial_score/std": 0.3264097571372986,
      "rewards/tag_count_reward/mean": -0.30078125,
      "rewards/tag_count_reward/std": 0.45904624462127686,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9985228180885315,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 15.96454906463623,
      "step": 365
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.21816373616456985,
      "epoch": 0.5865384615384616,
      "grad_norm": 1167.26611328125,
      "learning_rate": 1e-06,
      "loss": 0.2898,
      "step": 366
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.3828125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6015625,
      "entropy": 0.2148098573088646,
      "epoch": 0.5881410256410257,
      "grad_norm": 8.457894325256348,
      "learning_rate": 1e-06,
      "loss": 0.3274,
      "step": 367
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.65625,
      "entropy": 0.2301519215106964,
      "epoch": 0.5897435897435898,
      "grad_norm": 40.98564147949219,
      "learning_rate": 1e-06,
      "loss": 0.2575,
      "step": 368
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.380859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5608.0,
      "completions/mean_length": 7724.8984375,
      "completions/mean_terminated_length": 2398.32177734375,
      "completions/min_length": 571.0,
      "completions/min_terminated_length": 571.0,
      "entropy": 0.20847928524017334,
      "epoch": 0.5913461538461539,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 825.861083984375,
      "learning_rate": 1e-06,
      "loss": 0.2632,
      "num_tokens": 366384745.0,
      "reward": 0.16988316178321838,
      "reward_std": 0.19900578260421753,
      "rewards/progression_diversity/mean": -0.1527988612651825,
      "rewards/progression_diversity/std": 0.21360373497009277,
      "rewards/symbolic_reward_accuracy/mean": 0.123046875,
      "rewards/symbolic_reward_accuracy/std": 0.32881227135658264,
      "rewards/symbolic_reward_partial_score/mean": 0.4359537959098816,
      "rewards/symbolic_reward_partial_score/std": 0.35540899634361267,
      "rewards/tag_count_reward/mean": -0.33203125,
      "rewards/tag_count_reward/std": 0.47140273451805115,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9913198947906494,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 16.873680114746094,
      "step": 369
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.1968770995736122,
      "epoch": 0.592948717948718,
      "grad_norm": 528.3468627929688,
      "learning_rate": 1e-06,
      "loss": 0.3157,
      "step": 370
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.3515625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.19093196094036102,
      "epoch": 0.594551282051282,
      "grad_norm": 962.9395751953125,
      "learning_rate": 1e-06,
      "loss": 0.3246,
      "step": 371
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.20564716309309006,
      "epoch": 0.5961538461538461,
      "grad_norm": 17.06043243408203,
      "learning_rate": 1e-06,
      "loss": 0.281,
      "step": 372
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2890625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6353.0,
      "completions/mean_length": 6398.890625,
      "completions/mean_terminated_length": 2339.010986328125,
      "completions/min_length": 467.0,
      "completions/min_terminated_length": 467.0,
      "entropy": 0.25001922249794006,
      "epoch": 0.5977564102564102,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 419.783447265625,
      "learning_rate": 1e-06,
      "loss": 0.2007,
      "num_tokens": 370463153.0,
      "reward": 0.15392863750457764,
      "reward_std": 0.17637325823307037,
      "rewards/progression_diversity/mean": -0.11885648965835571,
      "rewards/progression_diversity/std": 0.2015119045972824,
      "rewards/symbolic_reward_accuracy/mean": 0.052734375,
      "rewards/symbolic_reward_accuracy/std": 0.22372129559516907,
      "rewards/symbolic_reward_partial_score/mean": 0.5014322996139526,
      "rewards/symbolic_reward_partial_score/std": 0.3110915422439575,
      "rewards/tag_count_reward/mean": -0.26953125,
      "rewards/tag_count_reward/std": 0.44415023922920227,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9979949593544006,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 16.023746490478516,
      "step": 373
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.2377210631966591,
      "epoch": 0.5993589743589743,
      "grad_norm": 631.596435546875,
      "learning_rate": 1e-06,
      "loss": 0.3175,
      "step": 374
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.2239432856440544,
      "epoch": 0.6009615384615384,
      "grad_norm": 676.9954223632812,
      "learning_rate": 1e-06,
      "loss": 0.307,
      "step": 375
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.2474696934223175,
      "epoch": 0.6025641025641025,
      "grad_norm": 75.13766479492188,
      "learning_rate": 1e-06,
      "loss": 0.207,
      "step": 376
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.3046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6881.0,
      "completions/mean_length": 6655.6796875,
      "completions/mean_terminated_length": 2392.7080078125,
      "completions/min_length": 484.0,
      "completions/min_terminated_length": 484.0,
      "entropy": 0.2293773591518402,
      "epoch": 0.6041666666666666,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 938.2958984375,
      "learning_rate": 1e-06,
      "loss": 0.2922,
      "num_tokens": 374795517.0,
      "reward": 0.16000205278396606,
      "reward_std": 0.18099814653396606,
      "rewards/progression_diversity/mean": -0.11991134285926819,
      "rewards/progression_diversity/std": 0.19749855995178223,
      "rewards/symbolic_reward_accuracy/mean": 0.08203125,
      "rewards/symbolic_reward_accuracy/std": 0.2746807038784027,
      "rewards/symbolic_reward_partial_score/mean": 0.4598633050918579,
      "rewards/symbolic_reward_partial_score/std": 0.3230631649494171,
      "rewards/tag_count_reward/mean": -0.259765625,
      "rewards/tag_count_reward/std": 0.4389347732067108,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0005474090576172,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 520.0,
      "sampling/sampling_logp_difference/mean": 15.338663101196289,
      "step": 377
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.2615222707390785,
      "epoch": 0.6057692307692307,
      "grad_norm": 1569.6458740234375,
      "learning_rate": 1e-06,
      "loss": 0.1825,
      "step": 378
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.2232813537120819,
      "epoch": 0.6073717948717948,
      "grad_norm": 432.9651184082031,
      "learning_rate": 1e-06,
      "loss": 0.3715,
      "step": 379
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.24861778318881989,
      "epoch": 0.6089743589743589,
      "grad_norm": 11.820572853088379,
      "learning_rate": 1e-06,
      "loss": 0.2359,
      "step": 380
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.31640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5300.0,
      "completions/mean_length": 6716.076171875,
      "completions/mean_terminated_length": 2241.20849609375,
      "completions/min_length": 350.0,
      "completions/min_terminated_length": 350.0,
      "entropy": 0.2513297498226166,
      "epoch": 0.6105769230769231,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 598.6245727539062,
      "learning_rate": 1e-06,
      "loss": 0.1991,
      "num_tokens": 379059012.0,
      "reward": 0.2126418650150299,
      "reward_std": 0.20646947622299194,
      "rewards/progression_diversity/mean": -0.1269288957118988,
      "rewards/progression_diversity/std": 0.20075669884681702,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.4962402582168579,
      "rewards/symbolic_reward_partial_score/std": 0.35096409916877747,
      "rewards/tag_count_reward/mean": -0.287109375,
      "rewards/tag_count_reward/std": 0.45285552740097046,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9986449480056763,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 520.0,
      "sampling/sampling_logp_difference/mean": 15.811168670654297,
      "step": 381
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.2153405398130417,
      "epoch": 0.6121794871794872,
      "grad_norm": 698.4373779296875,
      "learning_rate": 1e-06,
      "loss": 0.4257,
      "step": 382
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.2359524369239807,
      "epoch": 0.6137820512820513,
      "grad_norm": 251.32823181152344,
      "learning_rate": 1e-06,
      "loss": 0.2656,
      "step": 383
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.2591070830821991,
      "epoch": 0.6153846153846154,
      "grad_norm": 1601.9156494140625,
      "learning_rate": 1e-06,
      "loss": 0.1929,
      "step": 384
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.25,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 13378.0,
      "completions/mean_length": 5822.728515625,
      "completions/mean_terminated_length": 2302.3046875,
      "completions/min_length": 601.0,
      "completions/min_terminated_length": 601.0,
      "entropy": 0.27983543276786804,
      "epoch": 0.6169871794871795,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 612.2117919921875,
      "learning_rate": 1e-06,
      "loss": 0.1586,
      "num_tokens": 382808633.0,
      "reward": 0.22261884808540344,
      "reward_std": 0.21902015805244446,
      "rewards/progression_diversity/mean": -0.10237417370080948,
      "rewards/progression_diversity/std": 0.1874011605978012,
      "rewards/symbolic_reward_accuracy/mean": 0.150390625,
      "rewards/symbolic_reward_accuracy/std": 0.35780346393585205,
      "rewards/symbolic_reward_partial_score/mean": 0.522167980670929,
      "rewards/symbolic_reward_partial_score/std": 0.3341177999973297,
      "rewards/tag_count_reward/mean": -0.232421875,
      "rewards/tag_count_reward/std": 0.42278963327407837,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0090166330337524,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 520.0,
      "sampling/sampling_logp_difference/mean": 13.79207706451416,
      "step": 385
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.24246982485055923,
      "epoch": 0.6185897435897436,
      "grad_norm": 324.906005859375,
      "learning_rate": 1e-06,
      "loss": 0.3371,
      "step": 386
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2653020918369293,
      "epoch": 0.6201923076923077,
      "grad_norm": 195.28736877441406,
      "learning_rate": 1e-06,
      "loss": 0.2096,
      "step": 387
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.2655164748430252,
      "epoch": 0.6217948717948718,
      "grad_norm": 68.55646514892578,
      "learning_rate": 1e-06,
      "loss": 0.2088,
      "step": 388
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.33984375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 12824.0,
      "completions/mean_length": 7204.025390625,
      "completions/mean_terminated_length": 2478.23974609375,
      "completions/min_length": 636.0,
      "completions/min_terminated_length": 636.0,
      "entropy": 0.21731575578451157,
      "epoch": 0.6233974358974359,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1553.82080078125,
      "learning_rate": 1e-06,
      "loss": 0.3047,
      "num_tokens": 387345190.0,
      "reward": 0.16609683632850647,
      "reward_std": 0.20702703297138214,
      "rewards/progression_diversity/mean": -0.1368994563817978,
      "rewards/progression_diversity/std": 0.20169976353645325,
      "rewards/symbolic_reward_accuracy/mean": 0.10546875,
      "rewards/symbolic_reward_accuracy/std": 0.3074568510055542,
      "rewards/symbolic_reward_partial_score/mean": 0.44949543476104736,
      "rewards/symbolic_reward_partial_score/std": 0.3339717984199524,
      "rewards/tag_count_reward/mean": -0.306640625,
      "rewards/tag_count_reward/std": 0.4615498185157776,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9934464693069458,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 520.0,
      "sampling/sampling_logp_difference/mean": 16.535369873046875,
      "step": 389
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.4921875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.65625,
      "entropy": 0.1892649158835411,
      "epoch": 0.625,
      "grad_norm": 104.23577880859375,
      "learning_rate": 1e-06,
      "loss": 0.348,
      "step": 390
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.24794911593198776,
      "epoch": 0.6266025641025641,
      "grad_norm": 0.032321542501449585,
      "learning_rate": 1e-06,
      "loss": 0.1668,
      "step": 391
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.578125,
      "entropy": 0.21500948816537857,
      "epoch": 0.6282051282051282,
      "grad_norm": 1.6792036294937134,
      "learning_rate": 1e-06,
      "loss": 0.2748,
      "step": 392
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.314453125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 13686.0,
      "completions/mean_length": 6848.802734375,
      "completions/mean_terminated_length": 2475.108154296875,
      "completions/min_length": 368.0,
      "completions/min_terminated_length": 368.0,
      "entropy": 0.23190298676490784,
      "epoch": 0.6298076923076923,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 609.8284301757812,
      "learning_rate": 1e-06,
      "loss": 0.2339,
      "num_tokens": 391783393.0,
      "reward": 0.14935705065727234,
      "reward_std": 0.17916199564933777,
      "rewards/progression_diversity/mean": -0.1165420189499855,
      "rewards/progression_diversity/std": 0.18923556804656982,
      "rewards/symbolic_reward_accuracy/mean": 0.078125,
      "rewards/symbolic_reward_accuracy/std": 0.26863065361976624,
      "rewards/symbolic_reward_partial_score/mean": 0.44054362177848816,
      "rewards/symbolic_reward_partial_score/std": 0.3393227458000183,
      "rewards/tag_count_reward/mean": -0.28515625,
      "rewards/tag_count_reward/std": 0.45193037390708923,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.002633810043335,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 520.0,
      "sampling/sampling_logp_difference/mean": 14.761566162109375,
      "step": 393
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.2784353643655777,
      "epoch": 0.6314102564102564,
      "grad_norm": 262.9246520996094,
      "learning_rate": 1e-06,
      "loss": 0.1924,
      "step": 394
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.3984375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.65625,
      "entropy": 0.21299445629119873,
      "epoch": 0.6330128205128205,
      "grad_norm": 244.2508087158203,
      "learning_rate": 1e-06,
      "loss": 0.3124,
      "step": 395
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.65625,
      "entropy": 0.22290880978107452,
      "epoch": 0.6346153846153846,
      "grad_norm": 0.022922182455658913,
      "learning_rate": 1e-06,
      "loss": 0.3096,
      "step": 396
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.322265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6123.0,
      "completions/mean_length": 6933.5625,
      "completions/mean_terminated_length": 2439.838623046875,
      "completions/min_length": 505.0,
      "completions/min_terminated_length": 505.0,
      "entropy": 0.25610800087451935,
      "epoch": 0.6362179487179487,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 492.3603210449219,
      "learning_rate": 1e-06,
      "loss": 0.2415,
      "num_tokens": 396120849.0,
      "reward": 0.14940249919891357,
      "reward_std": 0.18980038166046143,
      "rewards/progression_diversity/mean": -0.124204620718956,
      "rewards/progression_diversity/std": 0.19669218361377716,
      "rewards/symbolic_reward_accuracy/mean": 0.08203125,
      "rewards/symbolic_reward_accuracy/std": 0.2746807038784027,
      "rewards/symbolic_reward_partial_score/mean": 0.43704426288604736,
      "rewards/symbolic_reward_partial_score/std": 0.3477852940559387,
      "rewards/tag_count_reward/mean": -0.296875,
      "rewards/tag_count_reward/std": 0.45732781291007996,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0012847185134888,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 520.0,
      "sampling/sampling_logp_difference/mean": 15.557909965515137,
      "step": 397
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.23665452748537064,
      "epoch": 0.6378205128205128,
      "grad_norm": 1791.0264892578125,
      "learning_rate": 1e-06,
      "loss": 0.2836,
      "step": 398
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5625,
      "entropy": 0.2314099296927452,
      "epoch": 0.6394230769230769,
      "grad_norm": 193.3421630859375,
      "learning_rate": 1e-06,
      "loss": 0.3197,
      "step": 399
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.24476408958435059,
      "epoch": 0.6410256410256411,
      "grad_norm": 7.516962051391602,
      "learning_rate": 1e-06,
      "loss": 0.2566,
      "step": 400
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2890625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6095.0,
      "completions/mean_length": 6450.169921875,
      "completions/mean_terminated_length": 2411.14013671875,
      "completions/min_length": 392.0,
      "completions/min_terminated_length": 392.0,
      "entropy": 0.23504739999771118,
      "epoch": 0.6426282051282052,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 792.7737426757812,
      "learning_rate": 1e-06,
      "loss": 0.306,
      "num_tokens": 400280488.0,
      "reward": 0.18703122437000275,
      "reward_std": 0.192929208278656,
      "rewards/progression_diversity/mean": -0.10351836681365967,
      "rewards/progression_diversity/std": 0.17522378265857697,
      "rewards/symbolic_reward_accuracy/mean": 0.1171875,
      "rewards/symbolic_reward_accuracy/std": 0.32195815443992615,
      "rewards/symbolic_reward_partial_score/mean": 0.4745442867279053,
      "rewards/symbolic_reward_partial_score/std": 0.3318890631198883,
      "rewards/tag_count_reward/mean": -0.24609375,
      "rewards/tag_count_reward/std": 0.4311550557613373,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0084717273712158,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 520.0,
      "sampling/sampling_logp_difference/mean": 14.434258460998535,
      "step": 401
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.25011158734560013,
      "epoch": 0.6442307692307693,
      "grad_norm": 317.9718322753906,
      "learning_rate": 1e-06,
      "loss": 0.2485,
      "step": 402
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.261634424328804,
      "epoch": 0.6458333333333334,
      "grad_norm": 102.99897766113281,
      "learning_rate": 1e-06,
      "loss": 0.2611,
      "step": 403
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.2772502601146698,
      "epoch": 0.6474358974358975,
      "grad_norm": 31.346302032470703,
      "learning_rate": 1e-06,
      "loss": 0.1701,
      "step": 404
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.255859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7054.0,
      "completions/mean_length": 6076.56640625,
      "completions/mean_terminated_length": 2532.540771484375,
      "completions/min_length": 577.0,
      "completions/min_terminated_length": 577.0,
      "entropy": 0.2811954766511917,
      "epoch": 0.6490384615384616,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 863.798828125,
      "learning_rate": 1e-06,
      "loss": 0.2297,
      "num_tokens": 404250186.0,
      "reward": 0.1544865369796753,
      "reward_std": 0.17089833319187164,
      "rewards/progression_diversity/mean": -0.08845682442188263,
      "rewards/progression_diversity/std": 0.16534774005413055,
      "rewards/symbolic_reward_accuracy/mean": 0.05078125,
      "rewards/symbolic_reward_accuracy/std": 0.21976542472839355,
      "rewards/symbolic_reward_partial_score/mean": 0.4886067509651184,
      "rewards/symbolic_reward_partial_score/std": 0.3330814838409424,
      "rewards/tag_count_reward/mean": -0.216796875,
      "rewards/tag_count_reward/std": 0.4124660789966583,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0118536949157715,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 520.0,
      "sampling/sampling_logp_difference/mean": 13.839906692504883,
      "step": 405
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.269717812538147,
      "epoch": 0.6506410256410257,
      "grad_norm": 659.154052734375,
      "learning_rate": 1e-06,
      "loss": 0.2937,
      "step": 406
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.25652148574590683,
      "epoch": 0.6522435897435898,
      "grad_norm": 0.023287350311875343,
      "learning_rate": 1e-06,
      "loss": 0.2854,
      "step": 407
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.3671875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6015625,
      "entropy": 0.2793606072664261,
      "epoch": 0.6538461538461539,
      "grad_norm": 0.029362130910158157,
      "learning_rate": 1e-06,
      "loss": 0.2409,
      "step": 408
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7332.0,
      "completions/mean_length": 6744.724609375,
      "completions/mean_terminated_length": 2674.808349609375,
      "completions/min_length": 504.0,
      "completions/min_terminated_length": 504.0,
      "entropy": 0.23652763664722443,
      "epoch": 0.655448717948718,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 488.6410827636719,
      "learning_rate": 1e-06,
      "loss": 0.3058,
      "num_tokens": 408669469.0,
      "reward": 0.1471533626317978,
      "reward_std": 0.18105512857437134,
      "rewards/progression_diversity/mean": -0.10253537446260452,
      "rewards/progression_diversity/std": 0.17895787954330444,
      "rewards/symbolic_reward_accuracy/mean": 0.076171875,
      "rewards/symbolic_reward_accuracy/std": 0.26553234457969666,
      "rewards/symbolic_reward_partial_score/mean": 0.42687174677848816,
      "rewards/symbolic_reward_partial_score/std": 0.3458375334739685,
      "rewards/tag_count_reward/mean": -0.255859375,
      "rewards/tag_count_reward/std": 0.43676990270614624,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0076942443847656,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 520.0,
      "sampling/sampling_logp_difference/mean": 13.969697952270508,
      "step": 409
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.2513129413127899,
      "epoch": 0.657051282051282,
      "grad_norm": 781.8362426757812,
      "learning_rate": 1e-06,
      "loss": 0.3877,
      "step": 410
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.640625,
      "entropy": 0.2599927484989166,
      "epoch": 0.6586538461538461,
      "grad_norm": 8.609899520874023,
      "learning_rate": 1e-06,
      "loss": 0.2786,
      "step": 411
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.3515625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.23813901096582413,
      "epoch": 0.6602564102564102,
      "grad_norm": 0.03463734686374664,
      "learning_rate": 1e-06,
      "loss": 0.3067,
      "step": 412
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.248046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 13444.0,
      "completions/mean_length": 6043.830078125,
      "completions/mean_terminated_length": 2632.916748046875,
      "completions/min_length": 596.0,
      "completions/min_terminated_length": 596.0,
      "entropy": 0.2797359824180603,
      "epoch": 0.6618589743589743,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 446.9456787109375,
      "learning_rate": 1e-06,
      "loss": 0.2367,
      "num_tokens": 412617750.0,
      "reward": 0.1771666407585144,
      "reward_std": 0.18979674577713013,
      "rewards/progression_diversity/mean": -0.08070126920938492,
      "rewards/progression_diversity/std": 0.15634961426258087,
      "rewards/symbolic_reward_accuracy/mean": 0.09375,
      "rewards/symbolic_reward_accuracy/std": 0.29176566004753113,
      "rewards/symbolic_reward_partial_score/mean": 0.4754069149494171,
      "rewards/symbolic_reward_partial_score/std": 0.336232453584671,
      "rewards/tag_count_reward/mean": -0.208984375,
      "rewards/tag_count_reward/std": 0.40698084235191345,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0141136646270752,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 524.0,
      "sampling/sampling_logp_difference/mean": 12.855762481689453,
      "step": 413
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.2539880573749542,
      "epoch": 0.6634615384615384,
      "grad_norm": 368.5082092285156,
      "learning_rate": 1e-06,
      "loss": 0.3098,
      "step": 414
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2531498521566391,
      "epoch": 0.6650641025641025,
      "grad_norm": 162.2779541015625,
      "learning_rate": 1e-06,
      "loss": 0.2807,
      "step": 415
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.2950977236032486,
      "epoch": 0.6666666666666666,
      "grad_norm": 129.06211853027344,
      "learning_rate": 1e-06,
      "loss": 0.1478,
      "step": 416
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.19921875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 12448.0,
      "completions/mean_length": 5337.125,
      "completions/mean_terminated_length": 2588.8779296875,
      "completions/min_length": 577.0,
      "completions/min_terminated_length": 577.0,
      "entropy": 0.28342205286026,
      "epoch": 0.6682692307692307,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 795.7291259765625,
      "learning_rate": 1e-06,
      "loss": 0.2281,
      "num_tokens": 416218758.0,
      "reward": 0.18160498142242432,
      "reward_std": 0.17491479218006134,
      "rewards/progression_diversity/mean": -0.06606545299291611,
      "rewards/progression_diversity/std": 0.14889149367809296,
      "rewards/symbolic_reward_accuracy/mean": 0.087890625,
      "rewards/symbolic_reward_accuracy/std": 0.2834126651287079,
      "rewards/symbolic_reward_partial_score/mean": 0.48906251788139343,
      "rewards/symbolic_reward_partial_score/std": 0.3222104609012604,
      "rewards/tag_count_reward/mean": -0.171875,
      "rewards/tag_count_reward/std": 0.3776407241821289,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0209629535675049,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 524.0,
      "sampling/sampling_logp_difference/mean": 11.58393383026123,
      "step": 417
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.29566076397895813,
      "epoch": 0.6698717948717948,
      "grad_norm": 879.3897094726562,
      "learning_rate": 1e-06,
      "loss": 0.2653,
      "step": 418
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.2756947875022888,
      "epoch": 0.6714743589743589,
      "grad_norm": 92.3701400756836,
      "learning_rate": 1e-06,
      "loss": 0.2869,
      "step": 419
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.31204286217689514,
      "epoch": 0.6730769230769231,
      "grad_norm": 0.028281621634960175,
      "learning_rate": 1e-06,
      "loss": 0.1507,
      "step": 420
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6618.0,
      "completions/mean_length": 5795.564453125,
      "completions/mean_terminated_length": 2693.901611328125,
      "completions/min_length": 476.0,
      "completions/min_terminated_length": 476.0,
      "entropy": 0.28321386873722076,
      "epoch": 0.6746794871794872,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1091.48193359375,
      "learning_rate": 1e-06,
      "loss": 0.2515,
      "num_tokens": 420108631.0,
      "reward": 0.20123516023159027,
      "reward_std": 0.18499191105365753,
      "rewards/progression_diversity/mean": -0.06691451370716095,
      "rewards/progression_diversity/std": 0.14319713413715363,
      "rewards/symbolic_reward_accuracy/mean": 0.130859375,
      "rewards/symbolic_reward_accuracy/std": 0.33757632970809937,
      "rewards/symbolic_reward_partial_score/mean": 0.47119140625,
      "rewards/symbolic_reward_partial_score/std": 0.3467746376991272,
      "rewards/tag_count_reward/mean": -0.1796875,
      "rewards/tag_count_reward/std": 0.38430243730545044,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0261237621307373,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 524.0,
      "sampling/sampling_logp_difference/mean": 10.4873685836792,
      "step": 421
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.29514726996421814,
      "epoch": 0.6762820512820513,
      "grad_norm": 0.027155712246894836,
      "learning_rate": 1e-06,
      "loss": 0.1629,
      "step": 422
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.27398838102817535,
      "epoch": 0.6778846153846154,
      "grad_norm": 0.03398658335208893,
      "learning_rate": 1e-06,
      "loss": 0.2472,
      "step": 423
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.28561370074748993,
      "epoch": 0.6794871794871795,
      "grad_norm": 0.03232736140489578,
      "learning_rate": 1e-06,
      "loss": 0.1731,
      "step": 424
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.25390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8604.0,
      "completions/mean_length": 6224.42578125,
      "completions/mean_terminated_length": 2766.97900390625,
      "completions/min_length": 550.0,
      "completions/min_terminated_length": 550.0,
      "entropy": 0.28223684430122375,
      "epoch": 0.6810897435897436,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 1497.0281982421875,
      "learning_rate": 1e-06,
      "loss": 0.2344,
      "num_tokens": 424132385.0,
      "reward": 0.1739288568496704,
      "reward_std": 0.19503268599510193,
      "rewards/progression_diversity/mean": -0.07684145867824554,
      "rewards/progression_diversity/std": 0.15114691853523254,
      "rewards/symbolic_reward_accuracy/mean": 0.09375,
      "rewards/symbolic_reward_accuracy/std": 0.29176566004753113,
      "rewards/symbolic_reward_partial_score/mean": 0.46123045682907104,
      "rewards/symbolic_reward_partial_score/std": 0.3385530114173889,
      "rewards/tag_count_reward/mean": -0.19921875,
      "rewards/tag_count_reward/std": 0.39980348944664,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.015244483947754,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 524.0,
      "sampling/sampling_logp_difference/mean": 12.87832260131836,
      "step": 425
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.2601533979177475,
      "epoch": 0.6826923076923077,
      "grad_norm": 5.701667785644531,
      "learning_rate": 1e-06,
      "loss": 0.2746,
      "step": 426
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.28981634974479675,
      "epoch": 0.6842948717948718,
      "grad_norm": 0.028896579518914223,
      "learning_rate": 1e-06,
      "loss": 0.1887,
      "step": 427
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.3828125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.268518328666687,
      "epoch": 0.6858974358974359,
      "grad_norm": 0.03334295004606247,
      "learning_rate": 1e-06,
      "loss": 0.2642,
      "step": 428
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.267578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8233.0,
      "completions/mean_length": 6416.837890625,
      "completions/mean_terminated_length": 2775.501220703125,
      "completions/min_length": 445.0,
      "completions/min_terminated_length": 445.0,
      "entropy": 0.26624637842178345,
      "epoch": 0.6875,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 960.3082885742188,
      "learning_rate": 1e-06,
      "loss": 0.2586,
      "num_tokens": 428286878.0,
      "reward": 0.19910672307014465,
      "reward_std": 0.20500460267066956,
      "rewards/progression_diversity/mean": -0.07907611131668091,
      "rewards/progression_diversity/std": 0.1532817780971527,
      "rewards/symbolic_reward_accuracy/mean": 0.130859375,
      "rewards/symbolic_reward_accuracy/std": 0.33757632970809937,
      "rewards/symbolic_reward_partial_score/mean": 0.4729654788970947,
      "rewards/symbolic_reward_partial_score/std": 0.36421647667884827,
      "rewards/tag_count_reward/mean": -0.205078125,
      "rewards/tag_count_reward/std": 0.4041535556316376,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0138148069381714,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 524.0,
      "sampling/sampling_logp_difference/mean": 12.851956367492676,
      "step": 429
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.2550949156284332,
      "epoch": 0.6891025641025641,
      "grad_norm": 0.16806720197200775,
      "learning_rate": 1e-06,
      "loss": 0.2759,
      "step": 430
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.28539833426475525,
      "epoch": 0.6907051282051282,
      "grad_norm": 0.027446668595075607,
      "learning_rate": 1e-06,
      "loss": 0.2034,
      "step": 431
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5703125,
      "entropy": 0.2528308480978012,
      "epoch": 0.6923076923076923,
      "grad_norm": 0.4935199022293091,
      "learning_rate": 1e-06,
      "loss": 0.2769,
      "step": 432
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.22265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6021.0,
      "completions/mean_length": 5849.77734375,
      "completions/mean_terminated_length": 2832.437255859375,
      "completions/min_length": 514.0,
      "completions/min_terminated_length": 514.0,
      "entropy": 0.2618635445833206,
      "epoch": 0.6939102564102564,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 2757.423828125,
      "learning_rate": 1e-06,
      "loss": 0.2911,
      "num_tokens": 432106236.0,
      "reward": 0.14727163314819336,
      "reward_std": 0.16309772431850433,
      "rewards/progression_diversity/mean": -0.06434153020381927,
      "rewards/progression_diversity/std": 0.13670898973941803,
      "rewards/symbolic_reward_accuracy/mean": 0.0546875,
      "rewards/symbolic_reward_accuracy/std": 0.2275916188955307,
      "rewards/symbolic_reward_partial_score/mean": 0.44357097148895264,
      "rewards/symbolic_reward_partial_score/std": 0.31754741072654724,
      "rewards/tag_count_reward/mean": -0.1796875,
      "rewards/tag_count_reward/std": 0.38430243730545044,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.018375039100647,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 524.0,
      "sampling/sampling_logp_difference/mean": 12.233776092529297,
      "step": 433
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.2820604145526886,
      "epoch": 0.6955128205128205,
      "grad_norm": 319.56304931640625,
      "learning_rate": 1e-06,
      "loss": 0.2209,
      "step": 434
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.2742318958044052,
      "epoch": 0.6971153846153846,
      "grad_norm": 0.03161383047699928,
      "learning_rate": 1e-06,
      "loss": 0.232,
      "step": 435
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.28273947536945343,
      "epoch": 0.6987179487179487,
      "grad_norm": 0.10165081173181534,
      "learning_rate": 1e-06,
      "loss": 0.194,
      "step": 436
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 13592.0,
      "completions/mean_length": 5897.6796875,
      "completions/mean_terminated_length": 2825.92919921875,
      "completions/min_length": 496.0,
      "completions/min_terminated_length": 496.0,
      "entropy": 0.2722629904747009,
      "epoch": 0.7003205128205128,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 247.05191040039062,
      "learning_rate": 1e-06,
      "loss": 0.2686,
      "num_tokens": 435945464.0,
      "reward": 0.17971304059028625,
      "reward_std": 0.16216953098773956,
      "rewards/progression_diversity/mean": -0.043344639241695404,
      "rewards/progression_diversity/std": 0.10168621689081192,
      "rewards/symbolic_reward_accuracy/mean": 0.08984375,
      "rewards/symbolic_reward_accuracy/std": 0.2862374484539032,
      "rewards/symbolic_reward_partial_score/mean": 0.4813476502895355,
      "rewards/symbolic_reward_partial_score/std": 0.332892507314682,
      "rewards/tag_count_reward/mean": -0.181640625,
      "rewards/tag_count_reward/std": 0.38592514395713806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0249509811401367,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 524.0,
      "sampling/sampling_logp_difference/mean": 9.850525856018066,
      "step": 437
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.2895769029855728,
      "epoch": 0.7019230769230769,
      "grad_norm": 77.53079986572266,
      "learning_rate": 1e-06,
      "loss": 0.2535,
      "step": 438
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.3120630830526352,
      "epoch": 0.7035256410256411,
      "grad_norm": 0.035726454108953476,
      "learning_rate": 1e-06,
      "loss": 0.2106,
      "step": 439
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.3188182860612869,
      "epoch": 0.7051282051282052,
      "grad_norm": 26.437782287597656,
      "learning_rate": 1e-06,
      "loss": 0.2773,
      "step": 440
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.20703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 15335.0,
      "completions/mean_length": 5722.31640625,
      "completions/mean_terminated_length": 2938.72412109375,
      "completions/min_length": 602.0,
      "completions/min_terminated_length": 602.0,
      "entropy": 0.31573159992694855,
      "epoch": 0.7067307692307693,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 219.33189392089844,
      "learning_rate": 1e-06,
      "loss": 0.2088,
      "num_tokens": 439743898.0,
      "reward": 0.16550663113594055,
      "reward_std": 0.17046445608139038,
      "rewards/progression_diversity/mean": -0.00011906892905244604,
      "rewards/progression_diversity/std": 0.002519554691389203,
      "rewards/symbolic_reward_accuracy/mean": 0.068359375,
      "rewards/symbolic_reward_accuracy/std": 0.25260838866233826,
      "rewards/symbolic_reward_partial_score/mean": 0.4709635376930237,
      "rewards/symbolic_reward_partial_score/std": 0.31855061650276184,
      "rewards/tag_count_reward/mean": -0.16796875,
      "rewards/tag_count_reward/std": 0.374204158782959,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0505764484405518,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 1.717551350593567,
      "step": 441
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.3088010251522064,
      "epoch": 0.7083333333333334,
      "grad_norm": 0.03410027176141739,
      "learning_rate": 1e-06,
      "loss": 0.1864,
      "step": 442
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.3080977499485016,
      "epoch": 0.7099358974358975,
      "grad_norm": 0.03793443366885185,
      "learning_rate": 1e-06,
      "loss": 0.2849,
      "step": 443
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.3152337819337845,
      "epoch": 0.7115384615384616,
      "grad_norm": 0.02681701071560383,
      "learning_rate": 1e-06,
      "loss": 0.2083,
      "step": 444
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.240234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7462.0,
      "completions/mean_length": 6347.04296875,
      "completions/mean_terminated_length": 3173.403564453125,
      "completions/min_length": 535.0,
      "completions/min_terminated_length": 535.0,
      "entropy": 0.2724430710077286,
      "epoch": 0.7131410256410257,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 317.01702880859375,
      "learning_rate": 1e-06,
      "loss": 0.2599,
      "num_tokens": 443837808.0,
      "reward": 0.15419574081897736,
      "reward_std": 0.1597881019115448,
      "rewards/progression_diversity/mean": -0.0023014158941805363,
      "rewards/progression_diversity/std": 0.015193020924925804,
      "rewards/symbolic_reward_accuracy/mean": 0.060546875,
      "rewards/symbolic_reward_accuracy/std": 0.2387305200099945,
      "rewards/symbolic_reward_partial_score/mean": 0.4554687738418579,
      "rewards/symbolic_reward_partial_score/std": 0.3217701315879822,
      "rewards/tag_count_reward/mean": -0.1875,
      "rewards/tag_count_reward/std": 0.39069411158561707,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0374995470046997,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 482.0,
      "sampling/sampling_logp_difference/mean": 2.7787814140319824,
      "step": 445
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.3045268654823303,
      "epoch": 0.7147435897435898,
      "grad_norm": 0.030019085854291916,
      "learning_rate": 1e-06,
      "loss": 0.178,
      "step": 446
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.26429812610149384,
      "epoch": 0.7163461538461539,
      "grad_norm": 0.038714099675416946,
      "learning_rate": 1e-06,
      "loss": 0.3328,
      "step": 447
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.28812122344970703,
      "epoch": 0.717948717948718,
      "grad_norm": 0.03327601030468941,
      "learning_rate": 1e-06,
      "loss": 0.2087,
      "step": 448
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.25,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8403.0,
      "completions/mean_length": 6429.9140625,
      "completions/mean_terminated_length": 3111.885498046875,
      "completions/min_length": 725.0,
      "completions/min_terminated_length": 725.0,
      "entropy": 0.27920031547546387,
      "epoch": 0.719551282051282,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 255.1264190673828,
      "learning_rate": 1e-06,
      "loss": 0.2454,
      "num_tokens": 448087220.0,
      "reward": 0.14131547510623932,
      "reward_std": 0.1714935302734375,
      "rewards/progression_diversity/mean": -0.005660384893417358,
      "rewards/progression_diversity/std": 0.01929020695388317,
      "rewards/symbolic_reward_accuracy/mean": 0.052734375,
      "rewards/symbolic_reward_accuracy/std": 0.22372129559516907,
      "rewards/symbolic_reward_partial_score/mean": 0.43413084745407104,
      "rewards/symbolic_reward_partial_score/std": 0.3264063894748688,
      "rewards/tag_count_reward/mean": -0.205078125,
      "rewards/tag_count_reward/std": 0.4041535556316376,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0356756448745728,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 454.0,
      "sampling/sampling_logp_difference/mean": 3.235456943511963,
      "step": 449
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.27578917145729065,
      "epoch": 0.7211538461538461,
      "grad_norm": 1.0962835550308228,
      "learning_rate": 1e-06,
      "loss": 0.2381,
      "step": 450
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.2769124358892441,
      "epoch": 0.7227564102564102,
      "grad_norm": 0.03778436407446861,
      "learning_rate": 1e-06,
      "loss": 0.2373,
      "step": 451
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.25797323882579803,
      "epoch": 0.7243589743589743,
      "grad_norm": 0.07790114730596542,
      "learning_rate": 1e-06,
      "loss": 0.2823,
      "step": 452
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.23046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 13926.0,
      "completions/mean_length": 6445.306640625,
      "completions/mean_terminated_length": 3468.74365234375,
      "completions/min_length": 525.0,
      "completions/min_terminated_length": 525.0,
      "entropy": 0.2812911868095398,
      "epoch": 0.7259615384615384,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 242.18006896972656,
      "learning_rate": 1e-06,
      "loss": 0.2006,
      "num_tokens": 452275297.0,
      "reward": 0.14035208523273468,
      "reward_std": 0.1568242758512497,
      "rewards/progression_diversity/mean": -0.00873718224465847,
      "rewards/progression_diversity/std": 0.02793550305068493,
      "rewards/symbolic_reward_accuracy/mean": 0.0546875,
      "rewards/symbolic_reward_accuracy/std": 0.2275916188955307,
      "rewards/symbolic_reward_partial_score/mean": 0.42190754413604736,
      "rewards/symbolic_reward_partial_score/std": 0.31549105048179626,
      "rewards/tag_count_reward/mean": -0.189453125,
      "rewards/tag_count_reward/std": 0.3922513723373413,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0444124937057495,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 400.0,
      "sampling/sampling_logp_difference/mean": 3.6778717041015625,
      "step": 453
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2804671823978424,
      "epoch": 0.7275641025641025,
      "grad_norm": 0.040219251066446304,
      "learning_rate": 1e-06,
      "loss": 0.1904,
      "step": 454
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.26130877435207367,
      "epoch": 0.7291666666666666,
      "grad_norm": 0.05419059470295906,
      "learning_rate": 1e-06,
      "loss": 0.2466,
      "step": 455
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.6171875,
      "entropy": 0.2619156539440155,
      "epoch": 0.7307692307692307,
      "grad_norm": 6.435583114624023,
      "learning_rate": 1e-06,
      "loss": 0.3236,
      "step": 456
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.205078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7699.0,
      "completions/mean_length": 6036.04296875,
      "completions/mean_terminated_length": 3366.422607421875,
      "completions/min_length": 786.0,
      "completions/min_terminated_length": 786.0,
      "entropy": 0.30352291464805603,
      "epoch": 0.7323717948717948,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 363.8023681640625,
      "learning_rate": 1e-06,
      "loss": 0.1275,
      "num_tokens": 456286823.0,
      "reward": 0.13116447627544403,
      "reward_std": 0.14602042734622955,
      "rewards/progression_diversity/mean": -0.00855403020977974,
      "rewards/progression_diversity/std": 0.028096795082092285,
      "rewards/symbolic_reward_accuracy/mean": 0.03515625,
      "rewards/symbolic_reward_accuracy/std": 0.1843547374010086,
      "rewards/symbolic_reward_partial_score/mean": 0.4192708432674408,
      "rewards/symbolic_reward_partial_score/std": 0.30005842447280884,
      "rewards/tag_count_reward/mean": -0.15625,
      "rewards/tag_count_reward/std": 0.36344730854034424,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0444647073745728,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 430.0,
      "sampling/sampling_logp_difference/mean": 5.1553144454956055,
      "step": 457
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.2788728326559067,
      "epoch": 0.7339743589743589,
      "grad_norm": 0.03529650345444679,
      "learning_rate": 1e-06,
      "loss": 0.2712,
      "step": 458
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.2527826279401779,
      "epoch": 0.7355769230769231,
      "grad_norm": 0.02542269416153431,
      "learning_rate": 1e-06,
      "loss": 0.3004,
      "step": 459
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.27167366445064545,
      "epoch": 0.7371794871794872,
      "grad_norm": 0.02993706800043583,
      "learning_rate": 1e-06,
      "loss": 0.2269,
      "step": 460
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.185546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7170.0,
      "completions/mean_length": 5713.298828125,
      "completions/mean_terminated_length": 3282.32373046875,
      "completions/min_length": 848.0,
      "completions/min_terminated_length": 848.0,
      "entropy": 0.2820378243923187,
      "epoch": 0.7387820512820513,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 278.36627197265625,
      "learning_rate": 1e-06,
      "loss": 0.1824,
      "num_tokens": 460071312.0,
      "reward": 0.17337031662464142,
      "reward_std": 0.17275133728981018,
      "rewards/progression_diversity/mean": -0.004277626518160105,
      "rewards/progression_diversity/std": 0.01868487522006035,
      "rewards/symbolic_reward_accuracy/mean": 0.072265625,
      "rewards/symbolic_reward_accuracy/std": 0.2591804563999176,
      "rewards/symbolic_reward_partial_score/mean": 0.485595703125,
      "rewards/symbolic_reward_partial_score/std": 0.31720206141471863,
      "rewards/tag_count_reward/mean": -0.15625,
      "rewards/tag_count_reward/std": 0.36344730854034424,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0428156852722168,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 456.0,
      "sampling/sampling_logp_difference/mean": 5.391382217407227,
      "step": 461
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.2974891811609268,
      "epoch": 0.7403846153846154,
      "grad_norm": 0.03358296677470207,
      "learning_rate": 1e-06,
      "loss": 0.1633,
      "step": 462
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5546875,
      "entropy": 0.27032434195280075,
      "epoch": 0.7419871794871795,
      "grad_norm": 0.03503183275461197,
      "learning_rate": 1e-06,
      "loss": 0.2628,
      "step": 463
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.2741468995809555,
      "epoch": 0.7435897435897436,
      "grad_norm": 0.047111768275499344,
      "learning_rate": 1e-06,
      "loss": 0.2698,
      "step": 464
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.224609375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 15885.0,
      "completions/mean_length": 6353.427734375,
      "completions/mean_terminated_length": 3447.84619140625,
      "completions/min_length": 490.0,
      "completions/min_terminated_length": 490.0,
      "entropy": 0.259844645857811,
      "epoch": 0.7451923076923077,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 823.7109985351562,
      "learning_rate": 1e-06,
      "loss": 0.2066,
      "num_tokens": 464320411.0,
      "reward": 0.13118945062160492,
      "reward_std": 0.14172431826591492,
      "rewards/progression_diversity/mean": -0.009473828598856926,
      "rewards/progression_diversity/std": 0.026718810200691223,
      "rewards/symbolic_reward_accuracy/mean": 0.0390625,
      "rewards/symbolic_reward_accuracy/std": 0.1939331740140915,
      "rewards/symbolic_reward_partial_score/mean": 0.42329102754592896,
      "rewards/symbolic_reward_partial_score/std": 0.2852300703525543,
      "rewards/tag_count_reward/mean": -0.19140625,
      "rewards/tag_count_reward/std": 0.3937928080558777,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0352518558502197,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 482.0,
      "sampling/sampling_logp_difference/mean": 8.388681411743164,
      "step": 465
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.3671875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.609375,
      "entropy": 0.25056569278240204,
      "epoch": 0.7467948717948718,
      "grad_norm": 0.02025044709444046,
      "learning_rate": 1e-06,
      "loss": 0.2388,
      "step": 466
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.25776257365942,
      "epoch": 0.7483974358974359,
      "grad_norm": 0.026951203122735023,
      "learning_rate": 1e-06,
      "loss": 0.248,
      "step": 467
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.25720326602458954,
      "epoch": 0.75,
      "grad_norm": 0.03187187388539314,
      "learning_rate": 1e-06,
      "loss": 0.2192,
      "step": 468
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.154296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 12745.0,
      "completions/mean_length": 5212.98828125,
      "completions/mean_terminated_length": 3174.859130859375,
      "completions/min_length": 439.0,
      "completions/min_terminated_length": 439.0,
      "entropy": 0.28627626597881317,
      "epoch": 0.7516025641025641,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 330.2890930175781,
      "learning_rate": 1e-06,
      "loss": 0.1308,
      "num_tokens": 467844069.0,
      "reward": 0.1364753395318985,
      "reward_std": 0.13440468907356262,
      "rewards/progression_diversity/mean": -0.009204398840665817,
      "rewards/progression_diversity/std": 0.028541414067149162,
      "rewards/symbolic_reward_accuracy/mean": 0.021484375,
      "rewards/symbolic_reward_accuracy/std": 0.14513419568538666,
      "rewards/symbolic_reward_partial_score/mean": 0.4545735716819763,
      "rewards/symbolic_reward_partial_score/std": 0.28182342648506165,
      "rewards/tag_count_reward/mean": -0.126953125,
      "rewards/tag_count_reward/std": 0.33324605226516724,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0359776020050049,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 500.0,
      "sampling/sampling_logp_difference/mean": 9.518253326416016,
      "step": 469
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2785252183675766,
      "epoch": 0.7532051282051282,
      "grad_norm": 0.04603856801986694,
      "learning_rate": 1e-06,
      "loss": 0.226,
      "step": 470
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.28009602427482605,
      "epoch": 0.7548076923076923,
      "grad_norm": 0.03989138454198837,
      "learning_rate": 1e-06,
      "loss": 0.1848,
      "step": 471
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.28221395611763,
      "epoch": 0.7564102564102564,
      "grad_norm": 0.04469291865825653,
      "learning_rate": 1e-06,
      "loss": 0.1911,
      "step": 472
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.15234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7659.0,
      "completions/mean_length": 5128.966796875,
      "completions/mean_terminated_length": 3106.1728515625,
      "completions/min_length": 533.0,
      "completions/min_terminated_length": 533.0,
      "entropy": 0.26698917895555496,
      "epoch": 0.7580128205128205,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 593.4261474609375,
      "learning_rate": 1e-06,
      "loss": 0.2374,
      "num_tokens": 471326436.0,
      "reward": 0.17709538340568542,
      "reward_std": 0.15860500931739807,
      "rewards/progression_diversity/mean": -0.008724970743060112,
      "rewards/progression_diversity/std": 0.03066168539226055,
      "rewards/symbolic_reward_accuracy/mean": 0.07421875,
      "rewards/symbolic_reward_accuracy/std": 0.2623828947544098,
      "rewards/symbolic_reward_partial_score/mean": 0.48448890447616577,
      "rewards/symbolic_reward_partial_score/std": 0.2948387861251831,
      "rewards/tag_count_reward/mean": -0.126953125,
      "rewards/tag_count_reward/std": 0.33324605226516724,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0322108268737793,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 516.0,
      "sampling/sampling_logp_difference/mean": 10.949844360351562,
      "step": 473
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.28319013118743896,
      "epoch": 0.7596153846153846,
      "grad_norm": 0.02427232451736927,
      "learning_rate": 1e-06,
      "loss": 0.196,
      "step": 474
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.29725903272628784,
      "epoch": 0.7612179487179487,
      "grad_norm": 0.03828584402799606,
      "learning_rate": 1e-06,
      "loss": 0.1165,
      "step": 475
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.2925011217594147,
      "epoch": 0.7628205128205128,
      "grad_norm": 0.030051294714212418,
      "learning_rate": 1e-06,
      "loss": 0.1457,
      "step": 476
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.173828125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6983.0,
      "completions/mean_length": 5560.11328125,
      "completions/mean_terminated_length": 3282.7470703125,
      "completions/min_length": 565.0,
      "completions/min_terminated_length": 565.0,
      "entropy": 0.2824954092502594,
      "epoch": 0.7644230769230769,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1322.20654296875,
      "learning_rate": 1e-06,
      "loss": 0.126,
      "num_tokens": 475077582.0,
      "reward": 0.1587892770767212,
      "reward_std": 0.146462082862854,
      "rewards/progression_diversity/mean": -0.010721936821937561,
      "rewards/progression_diversity/std": 0.032617803663015366,
      "rewards/symbolic_reward_accuracy/mean": 0.0625,
      "rewards/symbolic_reward_accuracy/std": 0.2422981858253479,
      "rewards/symbolic_reward_partial_score/mean": 0.45152994990348816,
      "rewards/symbolic_reward_partial_score/std": 0.2799621820449829,
      "rewards/tag_count_reward/mean": -0.140625,
      "rewards/tag_count_reward/std": 0.3479743003845215,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0238516330718994,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 532.0,
      "sampling/sampling_logp_difference/mean": 12.856636047363281,
      "step": 477
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5390625,
      "entropy": 0.24741245806217194,
      "epoch": 0.7660256410256411,
      "grad_norm": 0.03778098151087761,
      "learning_rate": 1e-06,
      "loss": 0.2498,
      "step": 478
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.26974865794181824,
      "epoch": 0.7676282051282052,
      "grad_norm": 0.03347004950046539,
      "learning_rate": 1e-06,
      "loss": 0.193,
      "step": 479
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.2672978490591049,
      "epoch": 0.7692307692307693,
      "grad_norm": 7.450833320617676,
      "learning_rate": 1e-06,
      "loss": 0.1819,
      "step": 480
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.15625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 15833.0,
      "completions/mean_length": 5183.740234375,
      "completions/mean_terminated_length": 3109.6181640625,
      "completions/min_length": 548.0,
      "completions/min_terminated_length": 548.0,
      "entropy": 0.27999909222126007,
      "epoch": 0.7708333333333334,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 328.314208984375,
      "learning_rate": 1e-06,
      "loss": 0.1661,
      "num_tokens": 478652985.0,
      "reward": 0.16447407007217407,
      "reward_std": 0.16247986257076263,
      "rewards/progression_diversity/mean": -0.009626220911741257,
      "rewards/progression_diversity/std": 0.03150993958115578,
      "rewards/symbolic_reward_accuracy/mean": 0.0625,
      "rewards/symbolic_reward_accuracy/std": 0.2422981858253479,
      "rewards/symbolic_reward_partial_score/mean": 0.4671875238418579,
      "rewards/symbolic_reward_partial_score/std": 0.2948901653289795,
      "rewards/tag_count_reward/mean": -0.130859375,
      "rewards/tag_count_reward/std": 0.33757632970809937,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0216107368469238,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 544.0,
      "sampling/sampling_logp_difference/mean": 14.160472869873047,
      "step": 481
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.27840910851955414,
      "epoch": 0.7724358974358975,
      "grad_norm": 0.05339457094669342,
      "learning_rate": 1e-06,
      "loss": 0.1271,
      "step": 482
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.53125,
      "entropy": 0.2735966593027115,
      "epoch": 0.7740384615384616,
      "grad_norm": 0.03152226284146309,
      "learning_rate": 1e-06,
      "loss": 0.2084,
      "step": 483
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.27197548747062683,
      "epoch": 0.7756410256410257,
      "grad_norm": 0.027287054806947708,
      "learning_rate": 1e-06,
      "loss": 0.1888,
      "step": 484
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.15625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6238.0,
      "completions/mean_length": 5183.2421875,
      "completions/mean_terminated_length": 3109.02783203125,
      "completions/min_length": 741.0,
      "completions/min_terminated_length": 741.0,
      "entropy": 0.2760499268770218,
      "epoch": 0.7772435897435898,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 610.1857299804688,
      "learning_rate": 1e-06,
      "loss": 0.1777,
      "num_tokens": 482058533.0,
      "reward": 0.16969150304794312,
      "reward_std": 0.13450489938259125,
      "rewards/progression_diversity/mean": -0.0122962836176157,
      "rewards/progression_diversity/std": 0.034512363374233246,
      "rewards/symbolic_reward_accuracy/mean": 0.06640625,
      "rewards/symbolic_reward_accuracy/std": 0.2492343932390213,
      "rewards/symbolic_reward_partial_score/mean": 0.47880858182907104,
      "rewards/symbolic_reward_partial_score/std": 0.29609718918800354,
      "rewards/tag_count_reward/mean": -0.13671875,
      "rewards/tag_count_reward/std": 0.3438861668109894,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.017279863357544,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 560.0,
      "sampling/sampling_logp_difference/mean": 15.952376365661621,
      "step": 485
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2726978659629822,
      "epoch": 0.7788461538461539,
      "grad_norm": 0.029111366719007492,
      "learning_rate": 1e-06,
      "loss": 0.222,
      "step": 486
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.29166853427886963,
      "epoch": 0.780448717948718,
      "grad_norm": 0.03601072356104851,
      "learning_rate": 1e-06,
      "loss": 0.1032,
      "step": 487
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.546875,
      "entropy": 0.2872645705938339,
      "epoch": 0.782051282051282,
      "grad_norm": 0.02848963811993599,
      "learning_rate": 1e-06,
      "loss": 0.1847,
      "step": 488
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.12109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 13892.0,
      "completions/mean_length": 4681.236328125,
      "completions/mean_terminated_length": 3068.855712890625,
      "completions/min_length": 480.0,
      "completions/min_terminated_length": 480.0,
      "entropy": 0.2787110507488251,
      "epoch": 0.7836538461538461,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 408.6405944824219,
      "learning_rate": 1e-06,
      "loss": 0.2056,
      "num_tokens": 485320734.0,
      "reward": 0.17335151135921478,
      "reward_std": 0.15505962073802948,
      "rewards/progression_diversity/mean": -0.010063882917165756,
      "rewards/progression_diversity/std": 0.03297411650419235,
      "rewards/symbolic_reward_accuracy/mean": 0.05859375,
      "rewards/symbolic_reward_accuracy/std": 0.23509246110916138,
      "rewards/symbolic_reward_partial_score/mean": 0.49549156427383423,
      "rewards/symbolic_reward_partial_score/std": 0.2642830014228821,
      "rewards/tag_count_reward/mean": -0.103515625,
      "rewards/tag_count_reward/std": 0.30492907762527466,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0186336040496826,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 576.0,
      "sampling/sampling_logp_difference/mean": 15.60879135131836,
      "step": 489
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.28752225637435913,
      "epoch": 0.7852564102564102,
      "grad_norm": 0.027989018708467484,
      "learning_rate": 1e-06,
      "loss": 0.1339,
      "step": 490
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.28790904581546783,
      "epoch": 0.7868589743589743,
      "grad_norm": 0.028247352689504623,
      "learning_rate": 1e-06,
      "loss": 0.1596,
      "step": 491
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.2850726991891861,
      "epoch": 0.7884615384615384,
      "grad_norm": 0.030095672234892845,
      "learning_rate": 1e-06,
      "loss": 0.1193,
      "step": 492
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6460.0,
      "completions/mean_length": 4684.54296875,
      "completions/mean_terminated_length": 3247.767578125,
      "completions/min_length": 664.0,
      "completions/min_terminated_length": 664.0,
      "entropy": 0.2784768342971802,
      "epoch": 0.7900641025641025,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 769.5609741210938,
      "learning_rate": 1e-06,
      "loss": 0.1766,
      "num_tokens": 488602388.0,
      "reward": 0.16056574881076813,
      "reward_std": 0.1394985318183899,
      "rewards/progression_diversity/mean": -0.008366829715669155,
      "rewards/progression_diversity/std": 0.0340796634554863,
      "rewards/symbolic_reward_accuracy/mean": 0.0390625,
      "rewards/symbolic_reward_accuracy/std": 0.1939331740140915,
      "rewards/symbolic_reward_partial_score/mean": 0.4892740845680237,
      "rewards/symbolic_reward_partial_score/std": 0.26915544271469116,
      "rewards/tag_count_reward/mean": -0.095703125,
      "rewards/tag_count_reward/std": 0.2944713830947876,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0246632099151611,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 592.0,
      "sampling/sampling_logp_difference/mean": 13.432557106018066,
      "step": 493
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2812363803386688,
      "epoch": 0.7916666666666666,
      "grad_norm": 0.02783721312880516,
      "learning_rate": 1e-06,
      "loss": 0.1083,
      "step": 494
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2817336320877075,
      "epoch": 0.7932692307692307,
      "grad_norm": 0.027454564347863197,
      "learning_rate": 1e-06,
      "loss": 0.1656,
      "step": 495
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.28926005959510803,
      "epoch": 0.7948717948717948,
      "grad_norm": 0.052017685025930405,
      "learning_rate": 1e-06,
      "loss": 0.0861,
      "step": 496
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.080078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6732.0,
      "completions/mean_length": 4107.0703125,
      "completions/mean_terminated_length": 3038.378173828125,
      "completions/min_length": 598.0,
      "completions/min_terminated_length": 598.0,
      "entropy": 0.31334738433361053,
      "epoch": 0.7964743589743589,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 333.04046630859375,
      "learning_rate": 1e-06,
      "loss": 0.0064,
      "num_tokens": 491578936.0,
      "reward": 0.1752394437789917,
      "reward_std": 0.12819209694862366,
      "rewards/progression_diversity/mean": -0.006330978125333786,
      "rewards/progression_diversity/std": 0.025279799476265907,
      "rewards/symbolic_reward_accuracy/mean": 0.04296875,
      "rewards/symbolic_reward_accuracy/std": 0.2029850035905838,
      "rewards/symbolic_reward_partial_score/mean": 0.5198893547058105,
      "rewards/symbolic_reward_partial_score/std": 0.25531089305877686,
      "rewards/tag_count_reward/mean": -0.064453125,
      "rewards/tag_count_reward/std": 0.24579854309558868,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0289983749389648,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 608.0,
      "sampling/sampling_logp_difference/mean": 12.241405487060547,
      "step": 497
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.27990762889385223,
      "epoch": 0.7980769230769231,
      "grad_norm": 0.03255309909582138,
      "learning_rate": 1e-06,
      "loss": 0.1956,
      "step": 498
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.2794528156518936,
      "epoch": 0.7996794871794872,
      "grad_norm": 0.02955903671681881,
      "learning_rate": 1e-06,
      "loss": 0.1626,
      "step": 499
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.28499020636081696,
      "epoch": 0.8012820512820513,
      "grad_norm": 0.031167522072792053,
      "learning_rate": 1e-06,
      "loss": 0.1355,
      "step": 500
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7721.0,
      "completions/mean_length": 4657.025390625,
      "completions/mean_terminated_length": 3216.87060546875,
      "completions/min_length": 448.0,
      "completions/min_terminated_length": 448.0,
      "entropy": 0.2538497596979141,
      "epoch": 0.8028846153846154,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 414.44403076171875,
      "learning_rate": 1e-06,
      "loss": 0.1954,
      "num_tokens": 494832389.0,
      "reward": 0.20476049184799194,
      "reward_std": 0.15316171944141388,
      "rewards/progression_diversity/mean": -0.008816054090857506,
      "rewards/progression_diversity/std": 0.03022286854684353,
      "rewards/symbolic_reward_accuracy/mean": 0.107421875,
      "rewards/symbolic_reward_accuracy/std": 0.30995169281959534,
      "rewards/symbolic_reward_partial_score/mean": 0.49988603591918945,
      "rewards/symbolic_reward_partial_score/std": 0.2798597812652588,
      "rewards/tag_count_reward/mean": -0.095703125,
      "rewards/tag_count_reward/std": 0.2944713830947876,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.017579197883606,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 616.0,
      "sampling/sampling_logp_difference/mean": 14.574539184570312,
      "step": 501
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.26197493076324463,
      "epoch": 0.8044871794871795,
      "grad_norm": 209.1951141357422,
      "learning_rate": 1e-06,
      "loss": 0.1276,
      "step": 502
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.26250365376472473,
      "epoch": 0.8060897435897436,
      "grad_norm": 0.026032600551843643,
      "learning_rate": 1e-06,
      "loss": 0.1871,
      "step": 503
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.28040294349193573,
      "epoch": 0.8076923076923077,
      "grad_norm": 0.0345035195350647,
      "learning_rate": 1e-06,
      "loss": 0.0733,
      "step": 504
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6329.0,
      "completions/mean_length": 4636.939453125,
      "completions/mean_terminated_length": 3194.318115234375,
      "completions/min_length": 426.0,
      "completions/min_terminated_length": 426.0,
      "entropy": 0.2512836679816246,
      "epoch": 0.8092948717948718,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1516.8017578125,
      "learning_rate": 1e-06,
      "loss": 0.2065,
      "num_tokens": 498112726.0,
      "reward": 0.14471670985221863,
      "reward_std": 0.1163174957036972,
      "rewards/progression_diversity/mean": -0.007822653278708458,
      "rewards/progression_diversity/std": 0.027946053072810173,
      "rewards/symbolic_reward_accuracy/mean": 0.01953125,
      "rewards/symbolic_reward_accuracy/std": 0.1385180652141571,
      "rewards/symbolic_reward_partial_score/mean": 0.4754882752895355,
      "rewards/symbolic_reward_partial_score/std": 0.25455576181411743,
      "rewards/tag_count_reward/mean": -0.095703125,
      "rewards/tag_count_reward/std": 0.2944713830947876,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0157849788665771,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 628.0,
      "sampling/sampling_logp_difference/mean": 15.629146575927734,
      "step": 505
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2723834812641144,
      "epoch": 0.8108974358974359,
      "grad_norm": 0.023276617750525475,
      "learning_rate": 1e-06,
      "loss": 0.0898,
      "step": 506
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.24540285021066666,
      "epoch": 0.8125,
      "grad_norm": 2.523991584777832,
      "learning_rate": 1e-06,
      "loss": 0.1915,
      "step": 507
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.27531081438064575,
      "epoch": 0.8141025641025641,
      "grad_norm": 0.03945612907409668,
      "learning_rate": 1e-06,
      "loss": 0.0928,
      "step": 508
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7280.0,
      "completions/mean_length": 4566.736328125,
      "completions/mean_terminated_length": 3230.87158203125,
      "completions/min_length": 496.0,
      "completions/min_terminated_length": 496.0,
      "entropy": 0.25356655567884445,
      "epoch": 0.8157051282051282,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 2822.289306640625,
      "learning_rate": 1e-06,
      "loss": 0.2041,
      "num_tokens": 501219247.0,
      "reward": 0.21896670758724213,
      "reward_std": 0.16303659975528717,
      "rewards/progression_diversity/mean": -0.00762702152132988,
      "rewards/progression_diversity/std": 0.029197612777352333,
      "rewards/symbolic_reward_accuracy/mean": 0.1171875,
      "rewards/symbolic_reward_accuracy/std": 0.32195815443992615,
      "rewards/symbolic_reward_partial_score/mean": 0.5237630605697632,
      "rewards/symbolic_reward_partial_score/std": 0.2946349084377289,
      "rewards/tag_count_reward/mean": -0.083984375,
      "rewards/tag_count_reward/std": 0.2776356339454651,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.023378610610962,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 636.0,
      "sampling/sampling_logp_difference/mean": 13.583820343017578,
      "step": 509
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.28210754692554474,
      "epoch": 0.8173076923076923,
      "grad_norm": 0.023770952597260475,
      "learning_rate": 1e-06,
      "loss": 0.0548,
      "step": 510
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.26374469697475433,
      "epoch": 0.8189102564102564,
      "grad_norm": 0.039423782378435135,
      "learning_rate": 1e-06,
      "loss": 0.1377,
      "step": 511
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2674047648906708,
      "epoch": 0.8205128205128205,
      "grad_norm": 0.0316784493625164,
      "learning_rate": 1e-06,
      "loss": 0.1403,
      "step": 512
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.130859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6836.0,
      "completions/mean_length": 5036.466796875,
      "completions/mean_terminated_length": 3327.9619140625,
      "completions/min_length": 756.0,
      "completions/min_terminated_length": 756.0,
      "entropy": 0.24934273958206177,
      "epoch": 0.8221153846153846,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 2330.71728515625,
      "learning_rate": 1e-06,
      "loss": 0.1066,
      "num_tokens": 504675182.0,
      "reward": 0.14850318431854248,
      "reward_std": 0.10997183620929718,
      "rewards/progression_diversity/mean": -0.01491626538336277,
      "rewards/progression_diversity/std": 0.04920973256230354,
      "rewards/symbolic_reward_accuracy/mean": 0.02734375,
      "rewards/symbolic_reward_accuracy/std": 0.16324250400066376,
      "rewards/symbolic_reward_partial_score/mean": 0.4779297113418579,
      "rewards/symbolic_reward_partial_score/std": 0.262747585773468,
      "rewards/tag_count_reward/mean": -0.111328125,
      "rewards/tag_count_reward/std": 0.31484565138816833,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.004473090171814,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 640.0,
      "sampling/sampling_logp_difference/mean": 19.370677947998047,
      "step": 513
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.22383376955986023,
      "epoch": 0.8237179487179487,
      "grad_norm": 0.028715649619698524,
      "learning_rate": 1e-06,
      "loss": 0.2759,
      "step": 514
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.23630475997924805,
      "epoch": 0.8253205128205128,
      "grad_norm": 0.03143737465143204,
      "learning_rate": 1e-06,
      "loss": 0.2012,
      "step": 515
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.25393127650022507,
      "epoch": 0.8269230769230769,
      "grad_norm": 0.031186457723379135,
      "learning_rate": 1e-06,
      "loss": 0.0771,
      "step": 516
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6311.0,
      "completions/mean_length": 4890.7109375,
      "completions/mean_terminated_length": 3248.812744140625,
      "completions/min_length": 729.0,
      "completions/min_terminated_length": 729.0,
      "entropy": 0.25201961398124695,
      "epoch": 0.8285256410256411,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 291.3205261230469,
      "learning_rate": 1e-06,
      "loss": 0.1009,
      "num_tokens": 508036634.0,
      "reward": 0.1587275266647339,
      "reward_std": 0.1489463448524475,
      "rewards/progression_diversity/mean": -0.014943530783057213,
      "rewards/progression_diversity/std": 0.049044571816921234,
      "rewards/symbolic_reward_accuracy/mean": 0.04296875,
      "rewards/symbolic_reward_accuracy/std": 0.2029850035905838,
      "rewards/symbolic_reward_partial_score/mean": 0.47750651836395264,
      "rewards/symbolic_reward_partial_score/std": 0.25351831316947937,
      "rewards/tag_count_reward/mean": -0.1015625,
      "rewards/tag_count_reward/std": 0.30236753821372986,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0078704357147217,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 640.0,
      "sampling/sampling_logp_difference/mean": 17.498640060424805,
      "step": 517
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.249831885099411,
      "epoch": 0.8301282051282052,
      "grad_norm": 133.84860229492188,
      "learning_rate": 1e-06,
      "loss": 0.1318,
      "step": 518
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2366134375333786,
      "epoch": 0.8317307692307693,
      "grad_norm": 0.042024604976177216,
      "learning_rate": 1e-06,
      "loss": 0.1665,
      "step": 519
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.23418694734573364,
      "epoch": 0.8333333333333334,
      "grad_norm": 0.023398801684379578,
      "learning_rate": 1e-06,
      "loss": 0.1972,
      "step": 520
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.12109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8086.0,
      "completions/mean_length": 4970.57421875,
      "completions/mean_terminated_length": 3398.057861328125,
      "completions/min_length": 492.0,
      "completions/min_terminated_length": 492.0,
      "entropy": 0.24243058264255524,
      "epoch": 0.8349358974358975,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 2872.773193359375,
      "learning_rate": 1e-06,
      "loss": 0.1181,
      "num_tokens": 511476240.0,
      "reward": 0.14858070015907288,
      "reward_std": 0.11485429108142853,
      "rewards/progression_diversity/mean": -0.017418432980775833,
      "rewards/progression_diversity/std": 0.056952327489852905,
      "rewards/symbolic_reward_accuracy/mean": 0.033203125,
      "rewards/symbolic_reward_accuracy/std": 0.17934183776378632,
      "rewards/symbolic_reward_partial_score/mean": 0.4639485478401184,
      "rewards/symbolic_reward_partial_score/std": 0.2507692277431488,
      "rewards/tag_count_reward/mean": -0.103515625,
      "rewards/tag_count_reward/std": 0.30492907762527466,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.008687973022461,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 648.0,
      "sampling/sampling_logp_difference/mean": 16.52926254272461,
      "step": 521
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.23126275092363358,
      "epoch": 0.8365384615384616,
      "grad_norm": 10394.248046875,
      "learning_rate": 1e-06,
      "loss": 1.4171,
      "step": 522
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.2367812767624855,
      "epoch": 0.8381410256410257,
      "grad_norm": 2017.530517578125,
      "learning_rate": 1e-06,
      "loss": 0.5716,
      "step": 523
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.21209017932415009,
      "epoch": 0.8397435897435898,
      "grad_norm": 100393.40625,
      "learning_rate": 1e-06,
      "loss": 8.2323,
      "step": 524
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8267.0,
      "completions/mean_length": 4735.314453125,
      "completions/mean_terminated_length": 3418.50634765625,
      "completions/min_length": 669.0,
      "completions/min_terminated_length": 669.0,
      "entropy": 0.24131416529417038,
      "epoch": 0.8413461538461539,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1308.060302734375,
      "learning_rate": 1e-06,
      "loss": 0.0936,
      "num_tokens": 514695857.0,
      "reward": 0.18877416849136353,
      "reward_std": 0.15265323221683502,
      "rewards/progression_diversity/mean": -0.018579598516225815,
      "rewards/progression_diversity/std": 0.0655272826552391,
      "rewards/symbolic_reward_accuracy/mean": 0.078125,
      "rewards/symbolic_reward_accuracy/std": 0.26863065361976624,
      "rewards/symbolic_reward_partial_score/mean": 0.5009602904319763,
      "rewards/symbolic_reward_partial_score/std": 0.28437086939811707,
      "rewards/tag_count_reward/mean": -0.08203125,
      "rewards/tag_count_reward/std": 0.2746807038784027,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0153002738952637,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 656.0,
      "sampling/sampling_logp_difference/mean": 14.62424087524414,
      "step": 525
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2314378321170807,
      "epoch": 0.842948717948718,
      "grad_norm": 2592.095703125,
      "learning_rate": 1e-06,
      "loss": 0.7228,
      "step": 526
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.24521462619304657,
      "epoch": 0.844551282051282,
      "grad_norm": 108445.40625,
      "learning_rate": 1e-06,
      "loss": 21.7594,
      "step": 527
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.23320768773555756,
      "epoch": 0.8461538461538461,
      "grad_norm": 3017.028076171875,
      "learning_rate": 1e-06,
      "loss": 0.684,
      "step": 528
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.130859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6846.0,
      "completions/mean_length": 5188.595703125,
      "completions/mean_terminated_length": 3502.99560546875,
      "completions/min_length": 426.0,
      "completions/min_terminated_length": 426.0,
      "entropy": 0.23708263784646988,
      "epoch": 0.8477564102564102,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 436.8207092285156,
      "learning_rate": 1e-06,
      "loss": 0.03,
      "num_tokens": 518277074.0,
      "reward": 0.16587716341018677,
      "reward_std": 0.13942670822143555,
      "rewards/progression_diversity/mean": -0.018729275092482567,
      "rewards/progression_diversity/std": 0.06120576709508896,
      "rewards/symbolic_reward_accuracy/mean": 0.05078125,
      "rewards/symbolic_reward_accuracy/std": 0.21976542472839355,
      "rewards/symbolic_reward_partial_score/mean": 0.4884440302848816,
      "rewards/symbolic_reward_partial_score/std": 0.27574726939201355,
      "rewards/tag_count_reward/mean": -0.109375,
      "rewards/tag_count_reward/std": 0.31241437792778015,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0088396072387695,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 668.0,
      "sampling/sampling_logp_difference/mean": 15.507166862487793,
      "step": 529
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2199697643518448,
      "epoch": 0.8493589743589743,
      "grad_norm": 0.035212092101573944,
      "learning_rate": 1e-06,
      "loss": 0.1937,
      "step": 530
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.22101637721061707,
      "epoch": 0.8509615384615384,
      "grad_norm": 0.045379314571619034,
      "learning_rate": 1e-06,
      "loss": 0.1429,
      "step": 531
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.18882086127996445,
      "epoch": 0.8525641025641025,
      "grad_norm": 0.029790926724672318,
      "learning_rate": 1e-06,
      "loss": 0.2997,
      "step": 532
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.111328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7388.0,
      "completions/mean_length": 4781.244140625,
      "completions/mean_terminated_length": 3327.712158203125,
      "completions/min_length": 629.0,
      "completions/min_terminated_length": 629.0,
      "entropy": 0.24219012260437012,
      "epoch": 0.8541666666666666,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1252.3094482421875,
      "learning_rate": 1e-06,
      "loss": 0.0937,
      "num_tokens": 521490175.0,
      "reward": 0.19383737444877625,
      "reward_std": 0.18049459159374237,
      "rewards/progression_diversity/mean": -0.01616620644927025,
      "rewards/progression_diversity/std": 0.057862453162670135,
      "rewards/symbolic_reward_accuracy/mean": 0.0859375,
      "rewards/symbolic_reward_accuracy/std": 0.28054583072662354,
      "rewards/symbolic_reward_partial_score/mean": 0.5060384273529053,
      "rewards/symbolic_reward_partial_score/std": 0.28731000423431396,
      "rewards/tag_count_reward/mean": -0.09375,
      "rewards/tag_count_reward/std": 0.29176566004753113,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0143693685531616,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 680.0,
      "sampling/sampling_logp_difference/mean": 14.317965507507324,
      "step": 533
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.22050420194864273,
      "epoch": 0.8557692307692307,
      "grad_norm": 0.0265819001942873,
      "learning_rate": 1e-06,
      "loss": 0.17,
      "step": 534
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.23473266512155533,
      "epoch": 0.8573717948717948,
      "grad_norm": 0.02770194783806801,
      "learning_rate": 1e-06,
      "loss": 0.1197,
      "step": 535
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.23300082236528397,
      "epoch": 0.8589743589743589,
      "grad_norm": 0.03747548907995224,
      "learning_rate": 1e-06,
      "loss": 0.1186,
      "step": 536
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.130859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 14631.0,
      "completions/mean_length": 5389.201171875,
      "completions/mean_terminated_length": 3733.804443359375,
      "completions/min_length": 973.0,
      "completions/min_terminated_length": 973.0,
      "entropy": 0.20381084084510803,
      "epoch": 0.8605769230769231,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 805.9613037109375,
      "learning_rate": 1e-06,
      "loss": 0.1618,
      "num_tokens": 525175254.0,
      "reward": 0.14221106469631195,
      "reward_std": 0.12302105128765106,
      "rewards/progression_diversity/mean": -0.0181512963026762,
      "rewards/progression_diversity/std": 0.06077505275607109,
      "rewards/symbolic_reward_accuracy/mean": 0.029296875,
      "rewards/symbolic_reward_accuracy/std": 0.16880230605602264,
      "rewards/symbolic_reward_partial_score/mean": 0.44860023260116577,
      "rewards/symbolic_reward_partial_score/std": 0.24405843019485474,
      "rewards/tag_count_reward/mean": -0.09765625,
      "rewards/tag_count_reward/std": 0.29713961482048035,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.005043387413025,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 680.0,
      "sampling/sampling_logp_difference/mean": 17.193443298339844,
      "step": 537
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.20633187144994736,
      "epoch": 0.8621794871794872,
      "grad_norm": 0.02656022645533085,
      "learning_rate": 1e-06,
      "loss": 0.13,
      "step": 538
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2009679228067398,
      "epoch": 0.8637820512820513,
      "grad_norm": 0.0328981950879097,
      "learning_rate": 1e-06,
      "loss": 0.2296,
      "step": 539
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.22686323523521423,
      "epoch": 0.8653846153846154,
      "grad_norm": 0.034096311777830124,
      "learning_rate": 1e-06,
      "loss": 0.1219,
      "step": 540
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.15234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8314.0,
      "completions/mean_length": 5590.28515625,
      "completions/mean_terminated_length": 3650.40087890625,
      "completions/min_length": 556.0,
      "completions/min_terminated_length": 556.0,
      "entropy": 0.18773535639047623,
      "epoch": 0.8669871794871795,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1792.0411376953125,
      "learning_rate": 1e-06,
      "loss": 0.3456,
      "num_tokens": 528906344.0,
      "reward": 0.12807568907737732,
      "reward_std": 0.12272882461547852,
      "rewards/progression_diversity/mean": -0.020069371908903122,
      "rewards/progression_diversity/std": 0.0620153546333313,
      "rewards/symbolic_reward_accuracy/mean": 0.017578125,
      "rewards/symbolic_reward_accuracy/std": 0.13154059648513794,
      "rewards/symbolic_reward_partial_score/mean": 0.4406087398529053,
      "rewards/symbolic_reward_partial_score/std": 0.2600206136703491,
      "rewards/tag_count_reward/mean": -0.14453125,
      "rewards/tag_count_reward/std": 0.35197147727012634,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.999688982963562,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 684.0,
      "sampling/sampling_logp_difference/mean": 18.892946243286133,
      "step": 541
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2230483591556549,
      "epoch": 0.8685897435897436,
      "grad_norm": 0.030025260522961617,
      "learning_rate": 1e-06,
      "loss": 0.0836,
      "step": 542
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.21129625290632248,
      "epoch": 0.8701923076923077,
      "grad_norm": 0.029054047539830208,
      "learning_rate": 1e-06,
      "loss": 0.1741,
      "step": 543
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.21284447610378265,
      "epoch": 0.8717948717948718,
      "grad_norm": 0.03237487003207207,
      "learning_rate": 1e-06,
      "loss": 0.1292,
      "step": 544
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.146484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7899.0,
      "completions/mean_length": 5516.55078125,
      "completions/mean_terminated_length": 3651.427734375,
      "completions/min_length": 890.0,
      "completions/min_terminated_length": 890.0,
      "entropy": 0.22282154858112335,
      "epoch": 0.8733974358974359,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 214.0397186279297,
      "learning_rate": 1e-06,
      "loss": 0.0983,
      "num_tokens": 532566914.0,
      "reward": 0.13807618618011475,
      "reward_std": 0.11777358502149582,
      "rewards/progression_diversity/mean": -0.013183235190808773,
      "rewards/progression_diversity/std": 0.04880344495177269,
      "rewards/symbolic_reward_accuracy/mean": 0.01953125,
      "rewards/symbolic_reward_accuracy/std": 0.1385180652141571,
      "rewards/symbolic_reward_partial_score/mean": 0.4639485776424408,
      "rewards/symbolic_reward_partial_score/std": 0.25486207008361816,
      "rewards/tag_count_reward/mean": -0.126953125,
      "rewards/tag_count_reward/std": 0.33324605226516724,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0068590641021729,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 684.0,
      "sampling/sampling_logp_difference/mean": 16.14547348022461,
      "step": 545
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.20825005322694778,
      "epoch": 0.875,
      "grad_norm": 0.026716945692896843,
      "learning_rate": 1e-06,
      "loss": 0.1561,
      "step": 546
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.19253433495759964,
      "epoch": 0.8766025641025641,
      "grad_norm": 0.5219494104385376,
      "learning_rate": 1e-06,
      "loss": 0.2689,
      "step": 547
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.21415498107671738,
      "epoch": 0.8782051282051282,
      "grad_norm": 0.0328250527381897,
      "learning_rate": 1e-06,
      "loss": 0.1751,
      "step": 548
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.12109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7527.0,
      "completions/mean_length": 5125.4140625,
      "completions/mean_terminated_length": 3574.231201171875,
      "completions/min_length": 617.0,
      "completions/min_terminated_length": 617.0,
      "entropy": 0.20042593777179718,
      "epoch": 0.8798076923076923,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 668.1113891601562,
      "learning_rate": 1e-06,
      "loss": 0.172,
      "num_tokens": 536149046.0,
      "reward": 0.15214072167873383,
      "reward_std": 0.13333392143249512,
      "rewards/progression_diversity/mean": -0.01493225246667862,
      "rewards/progression_diversity/std": 0.061123333871364594,
      "rewards/symbolic_reward_accuracy/mean": 0.0390625,
      "rewards/symbolic_reward_accuracy/std": 0.1939331740140915,
      "rewards/symbolic_reward_partial_score/mean": 0.4685709476470947,
      "rewards/symbolic_reward_partial_score/std": 0.26284727454185486,
      "rewards/tag_count_reward/mean": -0.1171875,
      "rewards/tag_count_reward/std": 0.32195815443992615,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.012796401977539,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 684.0,
      "sampling/sampling_logp_difference/mean": 12.92884349822998,
      "step": 549
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.21232055872678757,
      "epoch": 0.8814102564102564,
      "grad_norm": 0.02926452085375786,
      "learning_rate": 1e-06,
      "loss": 0.1224,
      "step": 550
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.20979397743940353,
      "epoch": 0.8830128205128205,
      "grad_norm": 0.02844247967004776,
      "learning_rate": 1e-06,
      "loss": 0.1311,
      "step": 551
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.1974465250968933,
      "epoch": 0.8846153846153846,
      "grad_norm": 0.9733312129974365,
      "learning_rate": 1e-06,
      "loss": 0.2059,
      "step": 552
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.107421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 13505.0,
      "completions/mean_length": 5052.701171875,
      "completions/mean_terminated_length": 3688.978271484375,
      "completions/min_length": 442.0,
      "completions/min_terminated_length": 442.0,
      "entropy": 0.2104191780090332,
      "epoch": 0.8862179487179487,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 485.2064514160156,
      "learning_rate": 1e-06,
      "loss": 0.1146,
      "num_tokens": 539713629.0,
      "reward": 0.15813249349594116,
      "reward_std": 0.12568315863609314,
      "rewards/progression_diversity/mean": -0.006576837040483952,
      "rewards/progression_diversity/std": 0.032705530524253845,
      "rewards/symbolic_reward_accuracy/mean": 0.03515625,
      "rewards/symbolic_reward_accuracy/std": 0.1843547374010086,
      "rewards/symbolic_reward_partial_score/mean": 0.4889160096645355,
      "rewards/symbolic_reward_partial_score/std": 0.24816220998764038,
      "rewards/tag_count_reward/mean": -0.095703125,
      "rewards/tag_count_reward/std": 0.2944713830947876,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0161933898925781,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 684.0,
      "sampling/sampling_logp_difference/mean": 11.277822494506836,
      "step": 553
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.20011521875858307,
      "epoch": 0.8878205128205128,
      "grad_norm": 170.6609344482422,
      "learning_rate": 1e-06,
      "loss": 0.1556,
      "step": 554
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.19730040431022644,
      "epoch": 0.8894230769230769,
      "grad_norm": 0.03489186614751816,
      "learning_rate": 1e-06,
      "loss": 0.1825,
      "step": 555
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2150089591741562,
      "epoch": 0.8910256410256411,
      "grad_norm": 2.223613977432251,
      "learning_rate": 1e-06,
      "loss": 0.1191,
      "step": 556
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.146484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7381.0,
      "completions/mean_length": 5520.578125,
      "completions/mean_terminated_length": 3656.146240234375,
      "completions/min_length": 633.0,
      "completions/min_terminated_length": 633.0,
      "entropy": 0.21337107568979263,
      "epoch": 0.8926282051282052,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 2374.782470703125,
      "learning_rate": 1e-06,
      "loss": 0.0962,
      "num_tokens": 543413141.0,
      "reward": 0.1618734747171402,
      "reward_std": 0.1597273349761963,
      "rewards/progression_diversity/mean": -0.014801505953073502,
      "rewards/progression_diversity/std": 0.047020554542541504,
      "rewards/symbolic_reward_accuracy/mean": 0.06640625,
      "rewards/symbolic_reward_accuracy/std": 0.2492343932390213,
      "rewards/symbolic_reward_partial_score/mean": 0.44892579317092896,
      "rewards/symbolic_reward_partial_score/std": 0.27913251519203186,
      "rewards/tag_count_reward/mean": -0.125,
      "rewards/tag_count_reward/std": 0.3310423493385315,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9986110925674438,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 688.0,
      "sampling/sampling_logp_difference/mean": 18.847614288330078,
      "step": 557
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.18713711947202682,
      "epoch": 0.8942307692307693,
      "grad_norm": 461.1898193359375,
      "learning_rate": 1e-06,
      "loss": 0.229,
      "step": 558
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.19299649447202682,
      "epoch": 0.8958333333333334,
      "grad_norm": 0.025606518611311913,
      "learning_rate": 1e-06,
      "loss": 0.2068,
      "step": 559
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.1916685476899147,
      "epoch": 0.8974358974358975,
      "grad_norm": 0.03450322896242142,
      "learning_rate": 1e-06,
      "loss": 0.1615,
      "step": 560
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.130859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8600.0,
      "completions/mean_length": 5390.927734375,
      "completions/mean_terminated_length": 3735.791015625,
      "completions/min_length": 774.0,
      "completions/min_terminated_length": 774.0,
      "entropy": 0.20245271921157837,
      "epoch": 0.8990384615384616,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 967.3348999023438,
      "learning_rate": 1e-06,
      "loss": 0.1123,
      "num_tokens": 547056576.0,
      "reward": 0.13580447435379028,
      "reward_std": 0.12063577026128769,
      "rewards/progression_diversity/mean": -0.0128158088773489,
      "rewards/progression_diversity/std": 0.04528747498989105,
      "rewards/symbolic_reward_accuracy/mean": 0.015625,
      "rewards/symbolic_reward_accuracy/std": 0.12414088100194931,
      "rewards/symbolic_reward_partial_score/mean": 0.4648274779319763,
      "rewards/symbolic_reward_partial_score/std": 0.2545829117298126,
      "rewards/tag_count_reward/mean": -0.12890625,
      "rewards/tag_count_reward/std": 0.33542385697364807,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0074546337127686,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 696.0,
      "sampling/sampling_logp_difference/mean": 15.149176597595215,
      "step": 561
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.1998009830713272,
      "epoch": 0.9006410256410257,
      "grad_norm": 0.021788692101836205,
      "learning_rate": 1e-06,
      "loss": 0.1464,
      "step": 562
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.1879931315779686,
      "epoch": 0.9022435897435898,
      "grad_norm": 0.03801717236638069,
      "learning_rate": 1e-06,
      "loss": 0.2225,
      "step": 563
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.19708418101072311,
      "epoch": 0.9038461538461539,
      "grad_norm": 0.028179530054330826,
      "learning_rate": 1e-06,
      "loss": 0.199,
      "step": 564
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.123046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7721.0,
      "completions/mean_length": 5477.939453125,
      "completions/mean_terminated_length": 3947.6904296875,
      "completions/min_length": 612.0,
      "completions/min_terminated_length": 612.0,
      "entropy": 0.1928836703300476,
      "epoch": 0.905448717948718,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1236.47900390625,
      "learning_rate": 1e-06,
      "loss": 0.1333,
      "num_tokens": 550782689.0,
      "reward": 0.1491735279560089,
      "reward_std": 0.13501743972301483,
      "rewards/progression_diversity/mean": -0.01135750487446785,
      "rewards/progression_diversity/std": 0.042249299585819244,
      "rewards/symbolic_reward_accuracy/mean": 0.04296875,
      "rewards/symbolic_reward_accuracy/std": 0.2029850035905838,
      "rewards/symbolic_reward_partial_score/mean": 0.4500976502895355,
      "rewards/symbolic_reward_partial_score/std": 0.2508525252342224,
      "rewards/tag_count_reward/mean": -0.115234375,
      "rewards/tag_count_reward/std": 0.3196168541908264,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.008751392364502,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 696.0,
      "sampling/sampling_logp_difference/mean": 14.606096267700195,
      "step": 565
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.19845175743103027,
      "epoch": 0.907051282051282,
      "grad_norm": 32.34842300415039,
      "learning_rate": 1e-06,
      "loss": 0.1143,
      "step": 566
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5078125,
      "entropy": 0.185062974691391,
      "epoch": 0.9086538461538461,
      "grad_norm": 26.105823516845703,
      "learning_rate": 1e-06,
      "loss": 0.2236,
      "step": 567
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.190422885119915,
      "epoch": 0.9102564102564102,
      "grad_norm": 0.02777501754462719,
      "learning_rate": 1e-06,
      "loss": 0.1594,
      "step": 568
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.115234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8955.0,
      "completions/mean_length": 5115.3125,
      "completions/mean_terminated_length": 3647.646728515625,
      "completions/min_length": 683.0,
      "completions/min_terminated_length": 683.0,
      "entropy": 0.18562601506710052,
      "epoch": 0.9118589743589743,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 2351.0830078125,
      "learning_rate": 1e-06,
      "loss": 0.1649,
      "num_tokens": 554328513.0,
      "reward": 0.1948603093624115,
      "reward_std": 0.1654394268989563,
      "rewards/progression_diversity/mean": -0.011040431447327137,
      "rewards/progression_diversity/std": 0.044010140001773834,
      "rewards/symbolic_reward_accuracy/mean": 0.091796875,
      "rewards/symbolic_reward_accuracy/std": 0.289021372795105,
      "rewards/symbolic_reward_partial_score/mean": 0.49755859375,
      "rewards/symbolic_reward_partial_score/std": 0.2728171944618225,
      "rewards/tag_count_reward/mean": -0.09375,
      "rewards/tag_count_reward/std": 0.29176566004753113,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0104902982711792,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 700.0,
      "sampling/sampling_logp_difference/mean": 13.494155883789062,
      "step": 569
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.20058371126651764,
      "epoch": 0.9134615384615384,
      "grad_norm": 0.0499381348490715,
      "learning_rate": 1e-06,
      "loss": 0.1804,
      "step": 570
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.19234994798898697,
      "epoch": 0.9150641025641025,
      "grad_norm": 0.04124247282743454,
      "learning_rate": 1e-06,
      "loss": 0.1146,
      "step": 571
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.20797397196292877,
      "epoch": 0.9166666666666666,
      "grad_norm": 0.028247077018022537,
      "learning_rate": 1e-06,
      "loss": 0.0817,
      "step": 572
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.07421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 10550.0,
      "completions/mean_length": 4636.544921875,
      "completions/mean_terminated_length": 3694.765625,
      "completions/min_length": 733.0,
      "completions/min_terminated_length": 733.0,
      "entropy": 0.21711117029190063,
      "epoch": 0.9182692307692307,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 234.86900329589844,
      "learning_rate": 1e-06,
      "loss": 0.0689,
      "num_tokens": 557529192.0,
      "reward": 0.20868007838726044,
      "reward_std": 0.16052421927452087,
      "rewards/progression_diversity/mean": -0.0065048919059336185,
      "rewards/progression_diversity/std": 0.03668885678052902,
      "rewards/symbolic_reward_accuracy/mean": 0.1015625,
      "rewards/symbolic_reward_accuracy/std": 0.30236753821372986,
      "rewards/symbolic_reward_partial_score/mean": 0.5161295533180237,
      "rewards/symbolic_reward_partial_score/std": 0.28020334243774414,
      "rewards/tag_count_reward/mean": -0.0703125,
      "rewards/tag_count_reward/std": 0.25592297315597534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0254935026168823,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 700.0,
      "sampling/sampling_logp_difference/mean": 8.34628963470459,
      "step": 573
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.20679805427789688,
      "epoch": 0.9198717948717948,
      "grad_norm": 0.06306414306163788,
      "learning_rate": 1e-06,
      "loss": 0.1743,
      "step": 574
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.21248594671487808,
      "epoch": 0.9214743589743589,
      "grad_norm": 0.033059295266866684,
      "learning_rate": 1e-06,
      "loss": 0.1313,
      "step": 575
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.21614834666252136,
      "epoch": 0.9230769230769231,
      "grad_norm": 0.023236755281686783,
      "learning_rate": 1e-06,
      "loss": 0.0916,
      "step": 576
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.130859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7749.0,
      "completions/mean_length": 5531.244140625,
      "completions/mean_terminated_length": 3897.233642578125,
      "completions/min_length": 932.0,
      "completions/min_terminated_length": 932.0,
      "entropy": 0.1956375539302826,
      "epoch": 0.9246794871794872,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 377.7513122558594,
      "learning_rate": 1e-06,
      "loss": 0.0834,
      "num_tokens": 561189509.0,
      "reward": 0.15623126924037933,
      "reward_std": 0.13696351647377014,
      "rewards/progression_diversity/mean": -0.006757441908121109,
      "rewards/progression_diversity/std": 0.02823774889111519,
      "rewards/symbolic_reward_accuracy/mean": 0.0390625,
      "rewards/symbolic_reward_accuracy/std": 0.1939331740140915,
      "rewards/symbolic_reward_partial_score/mean": 0.4793294072151184,
      "rewards/symbolic_reward_partial_score/std": 0.26099953055381775,
      "rewards/tag_count_reward/mean": -0.109375,
      "rewards/tag_count_reward/std": 0.31241437792778015,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0143134593963623,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 704.0,
      "sampling/sampling_logp_difference/mean": 11.73034381866455,
      "step": 577
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.19732258468866348,
      "epoch": 0.9262820512820513,
      "grad_norm": 0.030800944194197655,
      "learning_rate": 1e-06,
      "loss": 0.1879,
      "step": 578
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.19447887688875198,
      "epoch": 0.9278846153846154,
      "grad_norm": 0.03231598064303398,
      "learning_rate": 1e-06,
      "loss": 0.1604,
      "step": 579
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.19631733745336533,
      "epoch": 0.9294871794871795,
      "grad_norm": 0.0303302314132452,
      "learning_rate": 1e-06,
      "loss": 0.176,
      "step": 580
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.119140625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 10004.0,
      "completions/mean_length": 5372.26953125,
      "completions/mean_terminated_length": 3882.8779296875,
      "completions/min_length": 494.0,
      "completions/min_terminated_length": 494.0,
      "entropy": 0.19707387685775757,
      "epoch": 0.9310897435897436,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 645.3209838867188,
      "learning_rate": 1e-06,
      "loss": 0.1367,
      "num_tokens": 564833183.0,
      "reward": 0.14636225998401642,
      "reward_std": 0.13317109644412994,
      "rewards/progression_diversity/mean": -0.0073294322937726974,
      "rewards/progression_diversity/std": 0.031211169436573982,
      "rewards/symbolic_reward_accuracy/mean": 0.03515625,
      "rewards/symbolic_reward_accuracy/std": 0.1843547374010086,
      "rewards/symbolic_reward_partial_score/mean": 0.4542643129825592,
      "rewards/symbolic_reward_partial_score/std": 0.2500608563423157,
      "rewards/tag_count_reward/mean": -0.109375,
      "rewards/tag_count_reward/std": 0.31241437792778015,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0127942562103271,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 704.0,
      "sampling/sampling_logp_difference/mean": 12.442346572875977,
      "step": 581
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.1957729086279869,
      "epoch": 0.9326923076923077,
      "grad_norm": 0.03353438526391983,
      "learning_rate": 1e-06,
      "loss": 0.121,
      "step": 582
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.19240333884954453,
      "epoch": 0.9342948717948718,
      "grad_norm": 6.355886936187744,
      "learning_rate": 1e-06,
      "loss": 0.1826,
      "step": 583
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.1818048357963562,
      "epoch": 0.9358974358974359,
      "grad_norm": 0.03699595853686333,
      "learning_rate": 1e-06,
      "loss": 0.1854,
      "step": 584
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.11328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 11314.0,
      "completions/mean_length": 5186.060546875,
      "completions/mean_terminated_length": 3755.486572265625,
      "completions/min_length": 563.0,
      "completions/min_terminated_length": 563.0,
      "entropy": 0.1932847574353218,
      "epoch": 0.9375,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 567.5565185546875,
      "learning_rate": 1e-06,
      "loss": 0.1316,
      "num_tokens": 568426046.0,
      "reward": 0.18020448088645935,
      "reward_std": 0.1428624838590622,
      "rewards/progression_diversity/mean": -0.00933791883289814,
      "rewards/progression_diversity/std": 0.04019331932067871,
      "rewards/symbolic_reward_accuracy/mean": 0.07421875,
      "rewards/symbolic_reward_accuracy/std": 0.2623828947544098,
      "rewards/symbolic_reward_partial_score/mean": 0.4857584536075592,
      "rewards/symbolic_reward_partial_score/std": 0.26446202397346497,
      "rewards/tag_count_reward/mean": -0.099609375,
      "rewards/tag_count_reward/std": 0.29977133870124817,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.010314702987671,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 704.0,
      "sampling/sampling_logp_difference/mean": 13.445650100708008,
      "step": 585
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.193020299077034,
      "epoch": 0.9391025641025641,
      "grad_norm": 0.04160737618803978,
      "learning_rate": 1e-06,
      "loss": 0.1617,
      "step": 586
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.19402997195720673,
      "epoch": 0.9407051282051282,
      "grad_norm": 0.02597195655107498,
      "learning_rate": 1e-06,
      "loss": 0.1175,
      "step": 587
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.1869238093495369,
      "epoch": 0.9423076923076923,
      "grad_norm": 0.028201278299093246,
      "learning_rate": 1e-06,
      "loss": 0.1579,
      "step": 588
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.103515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8968.0,
      "completions/mean_length": 5071.556640625,
      "completions/mean_terminated_length": 3765.32666015625,
      "completions/min_length": 465.0,
      "completions/min_terminated_length": 465.0,
      "entropy": 0.20189929753541946,
      "epoch": 0.9439102564102564,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 732.089111328125,
      "learning_rate": 1e-06,
      "loss": 0.1303,
      "num_tokens": 571972779.0,
      "reward": 0.1497948169708252,
      "reward_std": 0.12462593615055084,
      "rewards/progression_diversity/mean": -0.00635932432487607,
      "rewards/progression_diversity/std": 0.03013550490140915,
      "rewards/symbolic_reward_accuracy/mean": 0.03515625,
      "rewards/symbolic_reward_accuracy/std": 0.1843547374010086,
      "rewards/symbolic_reward_partial_score/mean": 0.4578613340854645,
      "rewards/symbolic_reward_partial_score/std": 0.25101757049560547,
      "rewards/tag_count_reward/mean": -0.0859375,
      "rewards/tag_count_reward/std": 0.28054583072662354,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.014004111289978,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 708.0,
      "sampling/sampling_logp_difference/mean": 11.998255729675293,
      "step": 589
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.19095805287361145,
      "epoch": 0.9455128205128205,
      "grad_norm": 124.67987060546875,
      "learning_rate": 1e-06,
      "loss": 0.1042,
      "step": 590
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.19328931719064713,
      "epoch": 0.9471153846153846,
      "grad_norm": 0.03191890940070152,
      "learning_rate": 1e-06,
      "loss": 0.1363,
      "step": 591
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.18839430809020996,
      "epoch": 0.9487179487179487,
      "grad_norm": 0.5878300666809082,
      "learning_rate": 1e-06,
      "loss": 0.1515,
      "step": 592
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.09765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8919.0,
      "completions/mean_length": 5028.666015625,
      "completions/mean_terminated_length": 3799.733642578125,
      "completions/min_length": 519.0,
      "completions/min_terminated_length": 519.0,
      "entropy": 0.20535209774971008,
      "epoch": 0.9503205128205128,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 697.3111572265625,
      "learning_rate": 1e-06,
      "loss": 0.1668,
      "num_tokens": 575335136.0,
      "reward": 0.1823686808347702,
      "reward_std": 0.14714612066745758,
      "rewards/progression_diversity/mean": -0.006297005340456963,
      "rewards/progression_diversity/std": 0.02788672409951687,
      "rewards/symbolic_reward_accuracy/mean": 0.064453125,
      "rewards/symbolic_reward_accuracy/std": 0.24579854309558868,
      "rewards/symbolic_reward_partial_score/mean": 0.5111002922058105,
      "rewards/symbolic_reward_partial_score/std": 0.2809692621231079,
      "rewards/tag_count_reward/mean": -0.095703125,
      "rewards/tag_count_reward/std": 0.2944713830947876,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.015028953552246,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 712.0,
      "sampling/sampling_logp_difference/mean": 12.842995643615723,
      "step": 593
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2099175900220871,
      "epoch": 0.9519230769230769,
      "grad_norm": 0.03140028938651085,
      "learning_rate": 1e-06,
      "loss": 0.0722,
      "step": 594
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.19498909264802933,
      "epoch": 0.9535256410256411,
      "grad_norm": 0.029133038595318794,
      "learning_rate": 1e-06,
      "loss": 0.1713,
      "step": 595
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.20219064503908157,
      "epoch": 0.9551282051282052,
      "grad_norm": 253.06321716308594,
      "learning_rate": 1e-06,
      "loss": 0.1111,
      "step": 596
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.087890625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7837.0,
      "completions/mean_length": 4973.22265625,
      "completions/mean_terminated_length": 3873.68310546875,
      "completions/min_length": 757.0,
      "completions/min_terminated_length": 757.0,
      "entropy": 0.21754413098096848,
      "epoch": 0.9567307692307693,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 279.1087646484375,
      "learning_rate": 1e-06,
      "loss": 0.0303,
      "num_tokens": 578751282.0,
      "reward": 0.14907152950763702,
      "reward_std": 0.11073170602321625,
      "rewards/progression_diversity/mean": -0.004956918768584728,
      "rewards/progression_diversity/std": 0.026602579280734062,
      "rewards/symbolic_reward_accuracy/mean": 0.029296875,
      "rewards/symbolic_reward_accuracy/std": 0.16880230605602264,
      "rewards/symbolic_reward_partial_score/mean": 0.4677734375,
      "rewards/symbolic_reward_partial_score/std": 0.25228434801101685,
      "rewards/tag_count_reward/mean": -0.087890625,
      "rewards/tag_count_reward/std": 0.2834126651287079,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.02189040184021,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 712.0,
      "sampling/sampling_logp_difference/mean": 9.754007339477539,
      "step": 597
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.20453065633773804,
      "epoch": 0.9583333333333334,
      "grad_norm": 189.28167724609375,
      "learning_rate": 1e-06,
      "loss": 0.2374,
      "step": 598
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.20873209089040756,
      "epoch": 0.9599358974358975,
      "grad_norm": 47.94603729248047,
      "learning_rate": 1e-06,
      "loss": 0.0853,
      "step": 599
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.19657935947179794,
      "epoch": 0.9615384615384616,
      "grad_norm": 0.04143374040722847,
      "learning_rate": 1e-06,
      "loss": 0.2157,
      "step": 600
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.111328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 13524.0,
      "completions/mean_length": 5178.005859375,
      "completions/mean_terminated_length": 3774.17822265625,
      "completions/min_length": 487.0,
      "completions/min_terminated_length": 487.0,
      "entropy": 0.20645452290773392,
      "epoch": 0.9631410256410257,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 354.8719482421875,
      "learning_rate": 1e-06,
      "loss": 0.1344,
      "num_tokens": 582303205.0,
      "reward": 0.1530710607767105,
      "reward_std": 0.11708801239728928,
      "rewards/progression_diversity/mean": -0.0058823819272220135,
      "rewards/progression_diversity/std": 0.034123148769140244,
      "rewards/symbolic_reward_accuracy/mean": 0.0390625,
      "rewards/symbolic_reward_accuracy/std": 0.1939331740140915,
      "rewards/symbolic_reward_partial_score/mean": 0.4681152403354645,
      "rewards/symbolic_reward_partial_score/std": 0.2567155063152313,
      "rewards/tag_count_reward/mean": -0.107421875,
      "rewards/tag_count_reward/std": 0.30995169281959534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.022865891456604,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 716.0,
      "sampling/sampling_logp_difference/mean": 8.755578994750977,
      "step": 601
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.2079230472445488,
      "epoch": 0.9647435897435898,
      "grad_norm": 0.13820700347423553,
      "learning_rate": 1e-06,
      "loss": 0.154,
      "step": 602
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2013714388012886,
      "epoch": 0.9663461538461539,
      "grad_norm": 0.05277189984917641,
      "learning_rate": 1e-06,
      "loss": 0.082,
      "step": 603
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.20772968977689743,
      "epoch": 0.967948717948718,
      "grad_norm": 164.80355834960938,
      "learning_rate": 1e-06,
      "loss": 0.1195,
      "step": 604
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 14972.0,
      "completions/mean_length": 5488.85546875,
      "completions/mean_terminated_length": 3932.406494140625,
      "completions/min_length": 697.0,
      "completions/min_terminated_length": 697.0,
      "entropy": 0.18739153444766998,
      "epoch": 0.969551282051282,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 190.9339141845703,
      "learning_rate": 1e-06,
      "loss": 0.1891,
      "num_tokens": 585997131.0,
      "reward": 0.13834749162197113,
      "reward_std": 0.10280528664588928,
      "rewards/progression_diversity/mean": -0.004118788987398148,
      "rewards/progression_diversity/std": 0.02404225990176201,
      "rewards/symbolic_reward_accuracy/mean": 0.013671875,
      "rewards/symbolic_reward_accuracy/std": 0.1162383034825325,
      "rewards/symbolic_reward_partial_score/mean": 0.47366535663604736,
      "rewards/symbolic_reward_partial_score/std": 0.26022228598594666,
      "rewards/tag_count_reward/mean": -0.119140625,
      "rewards/tag_count_reward/std": 0.32427072525024414,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.021909475326538,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 716.0,
      "sampling/sampling_logp_difference/mean": 8.64652156829834,
      "step": 605
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.19497393816709518,
      "epoch": 0.9711538461538461,
      "grad_norm": 1706.4573974609375,
      "learning_rate": 1e-06,
      "loss": 0.2708,
      "step": 606
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.20221728086471558,
      "epoch": 0.9727564102564102,
      "grad_norm": 0.10873478651046753,
      "learning_rate": 1e-06,
      "loss": 0.1281,
      "step": 607
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.20786841213703156,
      "epoch": 0.9743589743589743,
      "grad_norm": 0.033346377313137054,
      "learning_rate": 1e-06,
      "loss": 0.0572,
      "step": 608
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.091796875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6816.0,
      "completions/mean_length": 5060.9296875,
      "completions/mean_terminated_length": 3916.447509765625,
      "completions/min_length": 634.0,
      "completions/min_terminated_length": 634.0,
      "entropy": 0.21419016271829605,
      "epoch": 0.9759615384615384,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 586.444580078125,
      "learning_rate": 1e-06,
      "loss": 0.0856,
      "num_tokens": 589427335.0,
      "reward": 0.16882535815238953,
      "reward_std": 0.1255710870027542,
      "rewards/progression_diversity/mean": -0.005649726837873459,
      "rewards/progression_diversity/std": 0.029928982257843018,
      "rewards/symbolic_reward_accuracy/mean": 0.044921875,
      "rewards/symbolic_reward_accuracy/std": 0.20733514428138733,
      "rewards/symbolic_reward_partial_score/mean": 0.500439465045929,
      "rewards/symbolic_reward_partial_score/std": 0.2657742500305176,
      "rewards/tag_count_reward/mean": -0.08203125,
      "rewards/tag_count_reward/std": 0.2746807038784027,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.023430585861206,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 9.106830596923828,
      "step": 609
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.18879622966051102,
      "epoch": 0.9775641025641025,
      "grad_norm": 236.33624267578125,
      "learning_rate": 1e-06,
      "loss": 0.2667,
      "step": 610
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2075936570763588,
      "epoch": 0.9791666666666666,
      "grad_norm": 2521.72265625,
      "learning_rate": 1e-06,
      "loss": 0.1519,
      "step": 611
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2186291366815567,
      "epoch": 0.9807692307692307,
      "grad_norm": 0.11578943580389023,
      "learning_rate": 1e-06,
      "loss": 0.0334,
      "step": 612
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.09765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7141.0,
      "completions/mean_length": 5257.30859375,
      "completions/mean_terminated_length": 4053.12109375,
      "completions/min_length": 1060.0,
      "completions/min_terminated_length": 1060.0,
      "entropy": 0.1974191665649414,
      "epoch": 0.9823717948717948,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 265.5615539550781,
      "learning_rate": 1e-06,
      "loss": 0.22,
      "num_tokens": 592999621.0,
      "reward": 0.16088539361953735,
      "reward_std": 0.10721283406019211,
      "rewards/progression_diversity/mean": -0.0037470583338290453,
      "rewards/progression_diversity/std": 0.02392842248082161,
      "rewards/symbolic_reward_accuracy/mean": 0.046875,
      "rewards/symbolic_reward_accuracy/std": 0.21157780289649963,
      "rewards/symbolic_reward_partial_score/mean": 0.47390952706336975,
      "rewards/symbolic_reward_partial_score/std": 0.23947234451770782,
      "rewards/tag_count_reward/mean": -0.09375,
      "rewards/tag_count_reward/std": 0.29176566004753113,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0285604000091553,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 6.232402801513672,
      "step": 613
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.21706271171569824,
      "epoch": 0.9839743589743589,
      "grad_norm": 286162.34375,
      "learning_rate": 1e-06,
      "loss": 93.3376,
      "step": 614
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2068658247590065,
      "epoch": 0.9855769230769231,
      "grad_norm": 8.328322410583496,
      "learning_rate": 1e-06,
      "loss": 0.1488,
      "step": 615
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.21251578629016876,
      "epoch": 0.9871794871794872,
      "grad_norm": 284.380615234375,
      "learning_rate": 1e-06,
      "loss": 0.1166,
      "step": 616
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 9352.0,
      "completions/mean_length": 5079.92578125,
      "completions/mean_terminated_length": 4017.149658203125,
      "completions/min_length": 703.0,
      "completions/min_terminated_length": 703.0,
      "entropy": 0.20627272129058838,
      "epoch": 0.9887820512820513,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 218.53688049316406,
      "learning_rate": 1e-06,
      "loss": 0.0771,
      "num_tokens": 596455135.0,
      "reward": 0.16188350319862366,
      "reward_std": 0.10645517706871033,
      "rewards/progression_diversity/mean": -0.0030562267638742924,
      "rewards/progression_diversity/std": 0.020391173660755157,
      "rewards/symbolic_reward_accuracy/mean": 0.041015625,
      "rewards/symbolic_reward_accuracy/std": 0.19852031767368317,
      "rewards/symbolic_reward_partial_score/mean": 0.484375,
      "rewards/symbolic_reward_partial_score/std": 0.23614925146102905,
      "rewards/tag_count_reward/mean": -0.080078125,
      "rewards/tag_count_reward/std": 0.271679550409317,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.029923439025879,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 5.815011024475098,
      "step": 617
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.21643763035535812,
      "epoch": 0.9903846153846154,
      "grad_norm": 8340.802734375,
      "learning_rate": 1e-06,
      "loss": 0.641,
      "step": 618
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.21413151174783707,
      "epoch": 0.9919871794871795,
      "grad_norm": 0.045055363327264786,
      "learning_rate": 1e-06,
      "loss": 0.0645,
      "step": 619
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.20299791544675827,
      "epoch": 0.9935897435897436,
      "grad_norm": 2.9305126667022705,
      "learning_rate": 1e-06,
      "loss": 0.1387,
      "step": 620
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 9045.0,
      "completions/mean_length": 4689.990234375,
      "completions/mean_terminated_length": 3910.389892578125,
      "completions/min_length": 613.0,
      "completions/min_terminated_length": 613.0,
      "entropy": 0.2107323333621025,
      "epoch": 0.9951923076923077,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 822.1503295898438,
      "learning_rate": 1e-06,
      "loss": 0.1441,
      "num_tokens": 599732858.0,
      "reward": 0.1449926495552063,
      "reward_std": 0.08904429525136948,
      "rewards/progression_diversity/mean": -0.0031782728619873524,
      "rewards/progression_diversity/std": 0.03115173615515232,
      "rewards/symbolic_reward_accuracy/mean": 0.02734375,
      "rewards/symbolic_reward_accuracy/std": 0.16324250400066376,
      "rewards/symbolic_reward_partial_score/mean": 0.4482584595680237,
      "rewards/symbolic_reward_partial_score/std": 0.2407679557800293,
      "rewards/tag_count_reward/mean": -0.05859375,
      "rewards/tag_count_reward/std": 0.23509246110916138,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0356639623641968,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 3.8055062294006348,
      "step": 621
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.21765735745429993,
      "epoch": 0.9967948717948718,
      "grad_norm": 0.11155123263597488,
      "learning_rate": 1e-06,
      "loss": 0.074,
      "step": 622
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2163386046886444,
      "epoch": 0.9983974358974359,
      "grad_norm": 0.035584624856710434,
      "learning_rate": 1e-06,
      "loss": 0.0425,
      "step": 623
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2188253104686737,
      "epoch": 1.0,
      "grad_norm": 0.041072484105825424,
      "learning_rate": 1e-06,
      "loss": 0.127,
      "step": 624
    },
    {
      "epoch": 1.0,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.05615234375,
      "eval_completions/max_length": 16384.0,
      "eval_completions/max_terminated_length": 6932.09375,
      "eval_completions/mean_length": 4607.2421875,
      "eval_completions/mean_terminated_length": 3907.518020629883,
      "eval_completions/min_length": 1083.65625,
      "eval_completions/min_terminated_length": 1083.65625,
      "eval_entropy": 0.21419930551201105,
      "eval_frac_reward_zero_std": 0.0078125,
      "eval_loss": 0.040855128318071365,
      "eval_num_tokens": 599732858.0,
      "eval_reward": 0.15359255392104387,
      "eval_reward_std": 0.0873756860382855,
      "eval_rewards/progression_diversity/mean": -0.002012389669403092,
      "eval_rewards/progression_diversity/std": 0.016667645196321246,
      "eval_rewards/symbolic_reward_accuracy/mean": 0.02294921875,
      "eval_rewards/symbolic_reward_accuracy/std": 0.11482575349509716,
      "eval_rewards/symbolic_reward_partial_score/mean": 0.4836405459791422,
      "eval_rewards/symbolic_reward_partial_score/std": 0.2179896729066968,
      "eval_rewards/tag_count_reward/mean": -0.052490234375,
      "eval_rewards/tag_count_reward/std": 0.21463490999303758,
      "eval_runtime": 4410.5978,
      "eval_samples_per_second": 0.057,
      "eval_sampling/importance_sampling_ratio/max": 2.0,
      "eval_sampling/importance_sampling_ratio/mean": 1.0419876985251904,
      "eval_sampling/importance_sampling_ratio/min": 0.0,
      "eval_sampling/sampling_logp_difference/max": 687.4686508178711,
      "eval_sampling/sampling_logp_difference/mean": 2.2834634024184197,
      "eval_steps_per_second": 0.0,
      "step": 624
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.064453125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7068.0,
      "completions/mean_length": 4668.310546875,
      "completions/mean_terminated_length": 3861.175537109375,
      "completions/min_length": 449.0,
      "completions/min_terminated_length": 449.0,
      "entropy": 0.2172800451517105,
      "epoch": 1.001602564102564,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 199.8529510498047,
      "learning_rate": 1e-06,
      "loss": 0.0884,
      "num_tokens": 603060937.0,
      "reward": 0.1484626829624176,
      "reward_std": 0.08050423860549927,
      "rewards/progression_diversity/mean": -0.001877149916253984,
      "rewards/progression_diversity/std": 0.01999068818986416,
      "rewards/symbolic_reward_accuracy/mean": 0.01953125,
      "rewards/symbolic_reward_accuracy/std": 0.1385180652141571,
      "rewards/symbolic_reward_partial_score/mean": 0.4773600399494171,
      "rewards/symbolic_reward_partial_score/std": 0.22182992100715637,
      "rewards/tag_count_reward/mean": -0.064453125,
      "rewards/tag_count_reward/std": 0.24579854309558868,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0379219055175781,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 2.98622727394104,
      "step": 625
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.22022734582424164,
      "epoch": 1.0032051282051282,
      "grad_norm": 20.245939254760742,
      "learning_rate": 1e-06,
      "loss": 0.0854,
      "step": 626
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.22906509041786194,
      "epoch": 1.0048076923076923,
      "grad_norm": 0.04052429273724556,
      "learning_rate": 1e-06,
      "loss": 0.0291,
      "step": 627
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.3359375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.20621829479932785,
      "epoch": 1.0064102564102564,
      "grad_norm": 0.6211011409759521,
      "learning_rate": 1e-06,
      "loss": 0.1451,
      "step": 628
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6794.0,
      "completions/mean_length": 4257.8671875,
      "completions/mean_terminated_length": 3764.934814453125,
      "completions/min_length": 779.0,
      "completions/min_terminated_length": 779.0,
      "entropy": 0.2293422520160675,
      "epoch": 1.0080128205128205,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 39.52851104736328,
      "learning_rate": 1e-06,
      "loss": 0.0808,
      "num_tokens": 606104485.0,
      "reward": 0.172392338514328,
      "reward_std": 0.10835404694080353,
      "rewards/progression_diversity/mean": -0.001978690503165126,
      "rewards/progression_diversity/std": 0.029427627101540565,
      "rewards/symbolic_reward_accuracy/mean": 0.046875,
      "rewards/symbolic_reward_accuracy/std": 0.21157780289649963,
      "rewards/symbolic_reward_partial_score/mean": 0.4920247495174408,
      "rewards/symbolic_reward_partial_score/std": 0.22508575022220612,
      "rewards/tag_count_reward/mean": -0.033203125,
      "rewards/tag_count_reward/std": 0.17934183776378632,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.044852375984192,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 1.5544822216033936,
      "step": 629
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.23360887169837952,
      "epoch": 1.0096153846153846,
      "grad_norm": 0.04223166033625603,
      "learning_rate": 1e-06,
      "loss": 0.0465,
      "step": 630
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2338651716709137,
      "epoch": 1.0112179487179487,
      "grad_norm": 5.134727478027344,
      "learning_rate": 1e-06,
      "loss": 0.0475,
      "step": 631
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.22925083339214325,
      "epoch": 1.0128205128205128,
      "grad_norm": 0.20785175263881683,
      "learning_rate": 1e-06,
      "loss": 0.0922,
      "step": 632
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7371.0,
      "completions/mean_length": 4059.546875,
      "completions/mean_terminated_length": 3713.076171875,
      "completions/min_length": 838.0,
      "completions/min_terminated_length": 838.0,
      "entropy": 0.23891044408082962,
      "epoch": 1.0144230769230769,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 379.1642150878906,
      "learning_rate": 1e-06,
      "loss": -0.0021,
      "num_tokens": 609025469.0,
      "reward": 0.19592517614364624,
      "reward_std": 0.10350589454174042,
      "rewards/progression_diversity/mean": -0.0012319717789068818,
      "rewards/progression_diversity/std": 0.013662992045283318,
      "rewards/symbolic_reward_accuracy/mean": 0.078125,
      "rewards/symbolic_reward_accuracy/std": 0.26863065361976624,
      "rewards/symbolic_reward_partial_score/mean": 0.5053385496139526,
      "rewards/symbolic_reward_partial_score/std": 0.2347535789012909,
      "rewards/tag_count_reward/mean": -0.025390625,
      "rewards/tag_count_reward/std": 0.15746226906776428,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0454130172729492,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 1.7672368288040161,
      "step": 633
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.23257441073656082,
      "epoch": 1.016025641025641,
      "grad_norm": 0.036344923079013824,
      "learning_rate": 1e-06,
      "loss": 0.0181,
      "step": 634
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.22513434290885925,
      "epoch": 1.017628205128205,
      "grad_norm": 0.2888374924659729,
      "learning_rate": 1e-06,
      "loss": 0.1695,
      "step": 635
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.23959070444107056,
      "epoch": 1.0192307692307692,
      "grad_norm": 0.04031003266572952,
      "learning_rate": 1e-06,
      "loss": 0.0214,
      "step": 636
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6660.0,
      "completions/mean_length": 4129.09375,
      "completions/mean_terminated_length": 3784.578125,
      "completions/min_length": 864.0,
      "completions/min_terminated_length": 864.0,
      "entropy": 0.22954071313142776,
      "epoch": 1.0208333333333333,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 803.85888671875,
      "learning_rate": 1e-06,
      "loss": 0.0414,
      "num_tokens": 611954541.0,
      "reward": 0.17702984809875488,
      "reward_std": 0.10909046232700348,
      "rewards/progression_diversity/mean": -0.0016061984933912754,
      "rewards/progression_diversity/std": 0.013535745441913605,
      "rewards/symbolic_reward_accuracy/mean": 0.046875,
      "rewards/symbolic_reward_accuracy/std": 0.21157780289649963,
      "rewards/symbolic_reward_partial_score/mean": 0.5055176019668579,
      "rewards/symbolic_reward_partial_score/std": 0.2366805523633957,
      "rewards/tag_count_reward/mean": -0.02734375,
      "rewards/tag_count_reward/std": 0.16324250400066376,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.040963053703308,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 4.412889003753662,
      "step": 637
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.23653806746006012,
      "epoch": 1.0224358974358974,
      "grad_norm": 0.032999370247125626,
      "learning_rate": 1e-06,
      "loss": 0.0386,
      "step": 638
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.23699220269918442,
      "epoch": 1.0240384615384615,
      "grad_norm": 0.044161055237054825,
      "learning_rate": 1e-06,
      "loss": 0.0856,
      "step": 639
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2416703775525093,
      "epoch": 1.0256410256410255,
      "grad_norm": 0.04160209372639656,
      "learning_rate": 1e-06,
      "loss": -0.0036,
      "step": 640
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 14949.0,
      "completions/mean_length": 3966.068359375,
      "completions/mean_terminated_length": 3616.9697265625,
      "completions/min_length": 704.0,
      "completions/min_terminated_length": 704.0,
      "entropy": 0.2398141846060753,
      "epoch": 1.0272435897435896,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1.787557601928711,
      "learning_rate": 1e-06,
      "loss": 0.0664,
      "num_tokens": 614793424.0,
      "reward": 0.1760445237159729,
      "reward_std": 0.09620662778615952,
      "rewards/progression_diversity/mean": -0.001016853959299624,
      "rewards/progression_diversity/std": 0.013255119323730469,
      "rewards/symbolic_reward_accuracy/mean": 0.04296875,
      "rewards/symbolic_reward_accuracy/std": 0.2029850035905838,
      "rewards/symbolic_reward_partial_score/mean": 0.5093749761581421,
      "rewards/symbolic_reward_partial_score/std": 0.20591585338115692,
      "rewards/tag_count_reward/mean": -0.025390625,
      "rewards/tag_count_reward/std": 0.15746226906776428,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0430570840835571,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 3.0360569953918457,
      "step": 641
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.23307596892118454,
      "epoch": 1.0288461538461537,
      "grad_norm": 0.9455596208572388,
      "learning_rate": 1e-06,
      "loss": 0.0624,
      "step": 642
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.23648811876773834,
      "epoch": 1.0304487179487178,
      "grad_norm": 0.03208564594388008,
      "learning_rate": 1e-06,
      "loss": 0.0013,
      "step": 643
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.23585931956768036,
      "epoch": 1.032051282051282,
      "grad_norm": 0.05858050286769867,
      "learning_rate": 1e-06,
      "loss": 0.087,
      "step": 644
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6913.0,
      "completions/mean_length": 3554.791015625,
      "completions/mean_terminated_length": 3194.13037109375,
      "completions/min_length": 679.0,
      "completions/min_terminated_length": 679.0,
      "entropy": 0.23475763201713562,
      "epoch": 1.0336538461538463,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 2.5429258346557617,
      "learning_rate": 1e-06,
      "loss": 0.0621,
      "num_tokens": 617548181.0,
      "reward": 0.1708458811044693,
      "reward_std": 0.0782819613814354,
      "rewards/progression_diversity/mean": -0.0018388191238045692,
      "rewards/progression_diversity/std": 0.020668642595410347,
      "rewards/symbolic_reward_accuracy/mean": 0.02734375,
      "rewards/symbolic_reward_accuracy/std": 0.16324250400066376,
      "rewards/symbolic_reward_partial_score/mean": 0.5226725339889526,
      "rewards/symbolic_reward_partial_score/std": 0.20760734379291534,
      "rewards/tag_count_reward/mean": -0.0234375,
      "rewards/tag_count_reward/std": 0.15143637359142303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.043994665145874,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 2.3344595432281494,
      "step": 645
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.231798954308033,
      "epoch": 1.0352564102564104,
      "grad_norm": 0.04138017073273659,
      "learning_rate": 1e-06,
      "loss": 0.0803,
      "step": 646
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.24344021826982498,
      "epoch": 1.0368589743589745,
      "grad_norm": 0.03530874475836754,
      "learning_rate": 1e-06,
      "loss": 0.0385,
      "step": 647
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.2350650280714035,
      "epoch": 1.0384615384615385,
      "grad_norm": 0.035837192088365555,
      "learning_rate": 1e-06,
      "loss": 0.0492,
      "step": 648
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.041015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5463.0,
      "completions/mean_length": 3913.376953125,
      "completions/mean_terminated_length": 3380.01025390625,
      "completions/min_length": 648.0,
      "completions/min_terminated_length": 648.0,
      "entropy": 0.21108438819646835,
      "epoch": 1.0400641025641026,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 306.6093444824219,
      "learning_rate": 1e-06,
      "loss": 0.151,
      "num_tokens": 620433494.0,
      "reward": 0.18476390838623047,
      "reward_std": 0.11055901646614075,
      "rewards/progression_diversity/mean": -0.001637741457670927,
      "rewards/progression_diversity/std": 0.015057054348289967,
      "rewards/symbolic_reward_accuracy/mean": 0.064453125,
      "rewards/symbolic_reward_accuracy/std": 0.24579854309558868,
      "rewards/symbolic_reward_partial_score/mean": 0.5006998777389526,
      "rewards/symbolic_reward_partial_score/std": 0.23833897709846497,
      "rewards/tag_count_reward/mean": -0.041015625,
      "rewards/tag_count_reward/std": 0.19852031767368317,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.038619875907898,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 3.4865543842315674,
      "step": 649
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.23450932651758194,
      "epoch": 1.0416666666666667,
      "grad_norm": 0.03884749859571457,
      "learning_rate": 1e-06,
      "loss": 0.0365,
      "step": 650
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.22569099068641663,
      "epoch": 1.0432692307692308,
      "grad_norm": 0.039085451513528824,
      "learning_rate": 1e-06,
      "loss": 0.045,
      "step": 651
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.22521612793207169,
      "epoch": 1.044871794871795,
      "grad_norm": 0.03394408896565437,
      "learning_rate": 1e-06,
      "loss": 0.0285,
      "step": 652
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6388.0,
      "completions/mean_length": 4014.796875,
      "completions/mean_terminated_length": 3299.222900390625,
      "completions/min_length": 500.0,
      "completions/min_terminated_length": 500.0,
      "entropy": 0.21447355300188065,
      "epoch": 1.046474358974359,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 3.4653165340423584,
      "learning_rate": 1e-06,
      "loss": 0.0665,
      "num_tokens": 623435790.0,
      "reward": 0.18457384407520294,
      "reward_std": 0.12713655829429626,
      "rewards/progression_diversity/mean": -0.00013509648852050304,
      "rewards/progression_diversity/std": 0.003056884743273258,
      "rewards/symbolic_reward_accuracy/mean": 0.05859375,
      "rewards/symbolic_reward_accuracy/std": 0.23509246110916138,
      "rewards/symbolic_reward_partial_score/mean": 0.5162923336029053,
      "rewards/symbolic_reward_partial_score/std": 0.25753483176231384,
      "rewards/tag_count_reward/mean": -0.0546875,
      "rewards/tag_count_reward/std": 0.2275916188955307,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0391998291015625,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.8635755777359009,
      "step": 653
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.21595758199691772,
      "epoch": 1.0480769230769231,
      "grad_norm": 839.4369506835938,
      "learning_rate": 1e-06,
      "loss": 0.1091,
      "step": 654
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.20685593783855438,
      "epoch": 1.0496794871794872,
      "grad_norm": 0.036022186279296875,
      "learning_rate": 1e-06,
      "loss": 0.1617,
      "step": 655
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.21082329750061035,
      "epoch": 1.0512820512820513,
      "grad_norm": 0.031025860458612442,
      "learning_rate": 1e-06,
      "loss": 0.0654,
      "step": 656
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6231.0,
      "completions/mean_length": 4426.666015625,
      "completions/mean_terminated_length": 3302.472412109375,
      "completions/min_length": 605.0,
      "completions/min_terminated_length": 605.0,
      "entropy": 0.20376238226890564,
      "epoch": 1.0528846153846154,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 223.56224060058594,
      "learning_rate": 1e-06,
      "loss": 0.0537,
      "num_tokens": 626540099.0,
      "reward": 0.1801775097846985,
      "reward_std": 0.12979283928871155,
      "rewards/progression_diversity/mean": -0.00031599291833117604,
      "rewards/progression_diversity/std": 0.005377059802412987,
      "rewards/symbolic_reward_accuracy/mean": 0.064453125,
      "rewards/symbolic_reward_accuracy/std": 0.24579854309558868,
      "rewards/symbolic_reward_partial_score/mean": 0.5003417730331421,
      "rewards/symbolic_reward_partial_score/std": 0.23511230945587158,
      "rewards/tag_count_reward/mean": -0.0859375,
      "rewards/tag_count_reward/std": 0.28054583072662354,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0354280471801758,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 1.9148916006088257,
      "step": 657
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.20688576996326447,
      "epoch": 1.0544871794871795,
      "grad_norm": 141.19863891601562,
      "learning_rate": 1e-06,
      "loss": 0.145,
      "step": 658
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.20776833593845367,
      "epoch": 1.0560897435897436,
      "grad_norm": 4.635189056396484,
      "learning_rate": 1e-06,
      "loss": 0.1138,
      "step": 659
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.21761061251163483,
      "epoch": 1.0576923076923077,
      "grad_norm": 0.028185199946165085,
      "learning_rate": 1e-06,
      "loss": 0.1123,
      "step": 660
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.064453125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 16107.0,
      "completions/mean_length": 4155.037109375,
      "completions/mean_terminated_length": 3312.540771484375,
      "completions/min_length": 717.0,
      "completions/min_terminated_length": 717.0,
      "entropy": 0.2150147706270218,
      "epoch": 1.0592948717948718,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 94.42283630371094,
      "learning_rate": 1e-06,
      "loss": 0.0893,
      "num_tokens": 629565094.0,
      "reward": 0.16469581425189972,
      "reward_std": 0.10881762206554413,
      "rewards/progression_diversity/mean": -0.0006340488907881081,
      "rewards/progression_diversity/std": 0.008337623439729214,
      "rewards/symbolic_reward_accuracy/mean": 0.029296875,
      "rewards/symbolic_reward_accuracy/std": 0.16880230605602264,
      "rewards/symbolic_reward_partial_score/mean": 0.5112468004226685,
      "rewards/symbolic_reward_partial_score/std": 0.23792609572410583,
      "rewards/tag_count_reward/mean": -0.0625,
      "rewards/tag_count_reward/std": 0.2422981858253479,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0353161096572876,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 2.2100651264190674,
      "step": 661
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.21428611874580383,
      "epoch": 1.060897435897436,
      "grad_norm": 114.22917938232422,
      "learning_rate": 1e-06,
      "loss": 0.1123,
      "step": 662
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2211027517914772,
      "epoch": 1.0625,
      "grad_norm": 0.0436987467110157,
      "learning_rate": 1e-06,
      "loss": 0.0534,
      "step": 663
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.18693063408136368,
      "epoch": 1.064102564102564,
      "grad_norm": 2.552537679672241,
      "learning_rate": 1e-06,
      "loss": 0.2344,
      "step": 664
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6130.0,
      "completions/mean_length": 4010.59765625,
      "completions/mean_terminated_length": 3294.78076171875,
      "completions/min_length": 426.0,
      "completions/min_terminated_length": 426.0,
      "entropy": 0.2113461047410965,
      "epoch": 1.0657051282051282,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 246.3715362548828,
      "learning_rate": 1e-06,
      "loss": 0.0674,
      "num_tokens": 632547896.0,
      "reward": 0.1795777976512909,
      "reward_std": 0.10678594559431076,
      "rewards/progression_diversity/mean": -0.00022982530936133116,
      "rewards/progression_diversity/std": 0.0034447184298187494,
      "rewards/symbolic_reward_accuracy/mean": 0.052734375,
      "rewards/symbolic_reward_accuracy/std": 0.22372129559516907,
      "rewards/symbolic_reward_partial_score/mean": 0.5113606452941895,
      "rewards/symbolic_reward_partial_score/std": 0.2122807502746582,
      "rewards/tag_count_reward/mean": -0.0546875,
      "rewards/tag_count_reward/std": 0.2275916188955307,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0385127067565918,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 1.2567355632781982,
      "step": 665
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2206530198454857,
      "epoch": 1.0673076923076923,
      "grad_norm": 0.2656633257865906,
      "learning_rate": 1e-06,
      "loss": 0.0547,
      "step": 666
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2168770357966423,
      "epoch": 1.0689102564102564,
      "grad_norm": 3.0182042121887207,
      "learning_rate": 1e-06,
      "loss": 0.125,
      "step": 667
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.2043883576989174,
      "epoch": 1.0705128205128205,
      "grad_norm": 0.024378696456551552,
      "learning_rate": 1e-06,
      "loss": 0.1789,
      "step": 668
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.068359375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 15758.0,
      "completions/mean_length": 4331.119140625,
      "completions/mean_terminated_length": 3446.73583984375,
      "completions/min_length": 634.0,
      "completions/min_terminated_length": 634.0,
      "entropy": 0.20979925245046616,
      "epoch": 1.0721153846153846,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 180.8795928955078,
      "learning_rate": 1e-06,
      "loss": 0.1064,
      "num_tokens": 635626965.0,
      "reward": 0.15303614735603333,
      "reward_std": 0.08873963356018066,
      "rewards/progression_diversity/mean": -9.739572124090046e-05,
      "rewards/progression_diversity/std": 0.0017328658141195774,
      "rewards/symbolic_reward_accuracy/mean": 0.013671875,
      "rewards/symbolic_reward_accuracy/std": 0.1162383034825325,
      "rewards/symbolic_reward_partial_score/mean": 0.5055663585662842,
      "rewards/symbolic_reward_partial_score/std": 0.22250817716121674,
      "rewards/tag_count_reward/mean": -0.068359375,
      "rewards/tag_count_reward/std": 0.25260838866233826,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.034943699836731,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 2.280566692352295,
      "step": 669
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.19934061914682388,
      "epoch": 1.0737179487179487,
      "grad_norm": 0.06182999163866043,
      "learning_rate": 1e-06,
      "loss": 0.1761,
      "step": 670
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.21370511502027512,
      "epoch": 1.0753205128205128,
      "grad_norm": 0.03356883302330971,
      "learning_rate": 1e-06,
      "loss": 0.0624,
      "step": 671
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.20709837973117828,
      "epoch": 1.0769230769230769,
      "grad_norm": 7.616127967834473,
      "learning_rate": 1e-06,
      "loss": 0.1033,
      "step": 672
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 16086.0,
      "completions/mean_length": 3967.392578125,
      "completions/mean_terminated_length": 3488.862060546875,
      "completions/min_length": 503.0,
      "completions/min_terminated_length": 503.0,
      "entropy": 0.21507354080677032,
      "epoch": 1.078525641025641,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 239.9991912841797,
      "learning_rate": 1e-06,
      "loss": 0.0799,
      "num_tokens": 638507678.0,
      "reward": 0.22426804900169373,
      "reward_std": 0.14125576615333557,
      "rewards/progression_diversity/mean": -0.00044275925029069185,
      "rewards/progression_diversity/std": 0.008287720382213593,
      "rewards/symbolic_reward_accuracy/mean": 0.10546875,
      "rewards/symbolic_reward_accuracy/std": 0.3074568510055542,
      "rewards/symbolic_reward_partial_score/mean": 0.5490071773529053,
      "rewards/symbolic_reward_partial_score/std": 0.2383054494857788,
      "rewards/tag_count_reward/mean": -0.037109375,
      "rewards/tag_count_reward/std": 0.18921469151973724,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0389597415924072,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 1.678175687789917,
      "step": 673
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.21100656688213348,
      "epoch": 1.080128205128205,
      "grad_norm": 2.742002487182617,
      "learning_rate": 1e-06,
      "loss": 0.0992,
      "step": 674
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.21375902742147446,
      "epoch": 1.0817307692307692,
      "grad_norm": 0.04870569705963135,
      "learning_rate": 1e-06,
      "loss": 0.077,
      "step": 675
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.2123679220676422,
      "epoch": 1.0833333333333333,
      "grad_norm": 0.03225383535027504,
      "learning_rate": 1e-06,
      "loss": 0.0213,
      "step": 676
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.033203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7501.0,
      "completions/mean_length": 3955.189453125,
      "completions/mean_terminated_length": 3528.341552734375,
      "completions/min_length": 693.0,
      "completions/min_terminated_length": 693.0,
      "entropy": 0.19846639037132263,
      "epoch": 1.0849358974358974,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 93.89030456542969,
      "learning_rate": 1e-06,
      "loss": 0.1439,
      "num_tokens": 641467087.0,
      "reward": 0.18885713815689087,
      "reward_std": 0.09211981296539307,
      "rewards/progression_diversity/mean": -2.9897617423557676e-05,
      "rewards/progression_diversity/std": 0.0006765059079043567,
      "rewards/symbolic_reward_accuracy/mean": 0.05859375,
      "rewards/symbolic_reward_accuracy/std": 0.23509246110916138,
      "rewards/symbolic_reward_partial_score/mean": 0.5234049558639526,
      "rewards/symbolic_reward_partial_score/std": 0.22506952285766602,
      "rewards/tag_count_reward/mean": -0.033203125,
      "rewards/tag_count_reward/std": 0.17934183776378632,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0395513772964478,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 696.0,
      "sampling/sampling_logp_difference/mean": 0.8175865411758423,
      "step": 677
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.21475867927074432,
      "epoch": 1.0865384615384615,
      "grad_norm": 0.04221729189157486,
      "learning_rate": 1e-06,
      "loss": 0.0553,
      "step": 678
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2134266495704651,
      "epoch": 1.0881410256410255,
      "grad_norm": 0.03336584195494652,
      "learning_rate": 1e-06,
      "loss": 0.0467,
      "step": 679
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.21357693523168564,
      "epoch": 1.0897435897435896,
      "grad_norm": 0.08009527623653412,
      "learning_rate": 1e-06,
      "loss": 0.0293,
      "step": 680
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.033203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6588.0,
      "completions/mean_length": 3841.345703125,
      "completions/mean_terminated_length": 3410.588134765625,
      "completions/min_length": 429.0,
      "completions/min_terminated_length": 429.0,
      "entropy": 0.21339119970798492,
      "epoch": 1.0913461538461537,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 614.2329711914062,
      "learning_rate": 1e-06,
      "loss": 0.1081,
      "num_tokens": 644259776.0,
      "reward": 0.2062734067440033,
      "reward_std": 0.1276092380285263,
      "rewards/progression_diversity/mean": -0.0005905528087168932,
      "rewards/progression_diversity/std": 0.006680304650217295,
      "rewards/symbolic_reward_accuracy/mean": 0.076171875,
      "rewards/symbolic_reward_accuracy/std": 0.26553234457969666,
      "rewards/symbolic_reward_partial_score/mean": 0.5463216304779053,
      "rewards/symbolic_reward_partial_score/std": 0.24570463597774506,
      "rewards/tag_count_reward/mean": -0.033203125,
      "rewards/tag_count_reward/std": 0.17934183776378632,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0360360145568848,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 3.0028653144836426,
      "step": 681
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.21380823105573654,
      "epoch": 1.092948717948718,
      "grad_norm": 0.03602517396211624,
      "learning_rate": 1e-06,
      "loss": 0.0344,
      "step": 682
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.21220668405294418,
      "epoch": 1.094551282051282,
      "grad_norm": 0.03208322450518608,
      "learning_rate": 1e-06,
      "loss": 0.0598,
      "step": 683
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2145799621939659,
      "epoch": 1.0961538461538463,
      "grad_norm": 0.026013914495706558,
      "learning_rate": 1e-06,
      "loss": 0.0583,
      "step": 684
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.025390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6038.0,
      "completions/mean_length": 3467.587890625,
      "completions/mean_terminated_length": 3131.088134765625,
      "completions/min_length": 412.0,
      "completions/min_terminated_length": 412.0,
      "entropy": 0.21561385691165924,
      "epoch": 1.0977564102564104,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 72.71430206298828,
      "learning_rate": 1e-06,
      "loss": 0.0558,
      "num_tokens": 647003501.0,
      "reward": 0.19664551317691803,
      "reward_std": 0.09150449931621552,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.05859375,
      "rewards/symbolic_reward_accuracy/std": 0.23509246110916138,
      "rewards/symbolic_reward_partial_score/mean": 0.5467610359191895,
      "rewards/symbolic_reward_partial_score/std": 0.21742354333400726,
      "rewards/tag_count_reward/mean": -0.025390625,
      "rewards/tag_count_reward/std": 0.15746226906776428,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0402486324310303,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 696.0,
      "sampling/sampling_logp_difference/mean": 0.6293116211891174,
      "step": 685
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.21327327191829681,
      "epoch": 1.0993589743589745,
      "grad_norm": 0.03850381448864937,
      "learning_rate": 1e-06,
      "loss": 0.0258,
      "step": 686
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.21018948405981064,
      "epoch": 1.1009615384615385,
      "grad_norm": 696.9727783203125,
      "learning_rate": 1e-06,
      "loss": 0.2903,
      "step": 687
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2201865389943123,
      "epoch": 1.1025641025641026,
      "grad_norm": 0.04112207144498825,
      "learning_rate": 1e-06,
      "loss": 0.0406,
      "step": 688
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.025390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8348.0,
      "completions/mean_length": 3814.763671875,
      "completions/mean_terminated_length": 3487.30859375,
      "completions/min_length": 676.0,
      "completions/min_terminated_length": 676.0,
      "entropy": 0.21837550401687622,
      "epoch": 1.1041666666666667,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 371.62738037109375,
      "learning_rate": 1e-06,
      "loss": 0.0119,
      "num_tokens": 649739428.0,
      "reward": 0.1545269787311554,
      "reward_std": 0.053041648119688034,
      "rewards/progression_diversity/mean": -0.00042648223461583257,
      "rewards/progression_diversity/std": 0.006124202162027359,
      "rewards/symbolic_reward_accuracy/mean": 0.001953125,
      "rewards/symbolic_reward_accuracy/std": 0.04419417306780815,
      "rewards/symbolic_reward_partial_score/mean": 0.5196614265441895,
      "rewards/symbolic_reward_partial_score/std": 0.16844609379768372,
      "rewards/tag_count_reward/mean": -0.025390625,
      "rewards/tag_count_reward/std": 0.15746226906776428,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0396302938461304,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 2.3374648094177246,
      "step": 689
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.21820075809955597,
      "epoch": 1.1057692307692308,
      "grad_norm": 0.03433378040790558,
      "learning_rate": 1e-06,
      "loss": 0.1054,
      "step": 690
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.21881910413503647,
      "epoch": 1.107371794871795,
      "grad_norm": 0.03191978111863136,
      "learning_rate": 1e-06,
      "loss": 0.018,
      "step": 691
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.22215040028095245,
      "epoch": 1.108974358974359,
      "grad_norm": 0.02860073745250702,
      "learning_rate": 1e-06,
      "loss": 0.0554,
      "step": 692
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7847.0,
      "completions/mean_length": 3363.330078125,
      "completions/mean_terminated_length": 3103.954345703125,
      "completions/min_length": 541.0,
      "completions/min_terminated_length": 541.0,
      "entropy": 0.21577662229537964,
      "epoch": 1.1105769230769231,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 299.27655029296875,
      "learning_rate": 1e-06,
      "loss": 0.1036,
      "num_tokens": 652366973.0,
      "reward": 0.23104004561901093,
      "reward_std": 0.10753922909498215,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.099609375,
      "rewards/symbolic_reward_accuracy/std": 0.29977133870124817,
      "rewards/symbolic_reward_partial_score/mean": 0.5774251222610474,
      "rewards/symbolic_reward_partial_score/std": 0.2207508236169815,
      "rewards/tag_count_reward/mean": -0.01953125,
      "rewards/tag_count_reward/std": 0.1385180652141571,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0428193807601929,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 712.0,
      "sampling/sampling_logp_difference/mean": 0.8698887228965759,
      "step": 693
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.22989095747470856,
      "epoch": 1.1121794871794872,
      "grad_norm": 0.051207710057497025,
      "learning_rate": 1e-06,
      "loss": -0.0037,
      "step": 694
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.2215583696961403,
      "epoch": 1.1137820512820513,
      "grad_norm": 0.03025507554411888,
      "learning_rate": 1e-06,
      "loss": 0.0495,
      "step": 695
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.22304167598485947,
      "epoch": 1.1153846153846154,
      "grad_norm": 0.03082253411412239,
      "learning_rate": 1e-06,
      "loss": 0.017,
      "step": 696
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6001.0,
      "completions/mean_length": 3574.69140625,
      "completions/mean_terminated_length": 3371.369140625,
      "completions/min_length": 582.0,
      "completions/min_terminated_length": 582.0,
      "entropy": 0.22506634891033173,
      "epoch": 1.1169871794871795,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 0.03984767943620682,
      "learning_rate": 1e-06,
      "loss": 0.0065,
      "num_tokens": 655022047.0,
      "reward": 0.17143532633781433,
      "reward_std": 0.09424548596143723,
      "rewards/progression_diversity/mean": -2.249992212455254e-05,
      "rewards/progression_diversity/std": 0.000509115110617131,
      "rewards/symbolic_reward_accuracy/mean": 0.037109375,
      "rewards/symbolic_reward_accuracy/std": 0.18921469151973724,
      "rewards/symbolic_reward_partial_score/mean": 0.50244140625,
      "rewards/symbolic_reward_partial_score/std": 0.21935398876667023,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0433446168899536,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 1.2426462173461914,
      "step": 697
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.22180306166410446,
      "epoch": 1.1185897435897436,
      "grad_norm": 108.0488052368164,
      "learning_rate": 1e-06,
      "loss": 0.0712,
      "step": 698
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.227100670337677,
      "epoch": 1.1201923076923077,
      "grad_norm": 0.034184910356998444,
      "learning_rate": 1e-06,
      "loss": 0.039,
      "step": 699
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2247518002986908,
      "epoch": 1.1217948717948718,
      "grad_norm": 0.03340495377779007,
      "learning_rate": 1e-06,
      "loss": 0.0312,
      "step": 700
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.04296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6562.0,
      "completions/mean_length": 3566.66796875,
      "completions/mean_terminated_length": 2991.19580078125,
      "completions/min_length": 444.0,
      "completions/min_terminated_length": 444.0,
      "entropy": 0.20850242674350739,
      "epoch": 1.123397435897436,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 87.1917495727539,
      "learning_rate": 1e-06,
      "loss": 0.1712,
      "num_tokens": 657830693.0,
      "reward": 0.1684723198413849,
      "reward_std": 0.07193771004676819,
      "rewards/progression_diversity/mean": -0.0009145288495346904,
      "rewards/progression_diversity/std": 0.013037758879363537,
      "rewards/symbolic_reward_accuracy/mean": 0.01171875,
      "rewards/symbolic_reward_accuracy/std": 0.10772226005792618,
      "rewards/symbolic_reward_partial_score/mean": 0.5511881113052368,
      "rewards/symbolic_reward_partial_score/std": 0.19447360932826996,
      "rewards/tag_count_reward/mean": -0.0390625,
      "rewards/tag_count_reward/std": 0.1939331740140915,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0334614515304565,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 4.139370918273926,
      "step": 701
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.21462332457304,
      "epoch": 1.125,
      "grad_norm": 0.03925897926092148,
      "learning_rate": 1e-06,
      "loss": 0.1,
      "step": 702
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2194051519036293,
      "epoch": 1.126602564102564,
      "grad_norm": 0.032283466309309006,
      "learning_rate": 1e-06,
      "loss": 0.068,
      "step": 703
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.2256844863295555,
      "epoch": 1.1282051282051282,
      "grad_norm": 0.027121527120471,
      "learning_rate": 1e-06,
      "loss": 0.0355,
      "step": 704
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7975.0,
      "completions/mean_length": 3463.818359375,
      "completions/mean_terminated_length": 3284.726806640625,
      "completions/min_length": 473.0,
      "completions/min_terminated_length": 473.0,
      "entropy": 0.22059781849384308,
      "epoch": 1.1298076923076923,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.042044855654239655,
      "learning_rate": 1e-06,
      "loss": 0.0491,
      "num_tokens": 660390200.0,
      "reward": 0.19374506175518036,
      "reward_std": 0.11770647019147873,
      "rewards/progression_diversity/mean": -6.273954568314366e-06,
      "rewards/progression_diversity/std": 0.00014196339179761708,
      "rewards/symbolic_reward_accuracy/mean": 0.06640625,
      "rewards/symbolic_reward_accuracy/std": 0.2492343932390213,
      "rewards/symbolic_reward_partial_score/mean": 0.5175618529319763,
      "rewards/symbolic_reward_partial_score/std": 0.2401990294456482,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0451421737670898,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 712.0,
      "sampling/sampling_logp_difference/mean": 0.5323032736778259,
      "step": 705
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2260003685951233,
      "epoch": 1.1314102564102564,
      "grad_norm": 0.026536772027611732,
      "learning_rate": 1e-06,
      "loss": 0.0615,
      "step": 706
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.2393200546503067,
      "epoch": 1.1330128205128205,
      "grad_norm": 0.036755435168743134,
      "learning_rate": 1e-06,
      "loss": 0.0075,
      "step": 707
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.23273401707410812,
      "epoch": 1.1346153846153846,
      "grad_norm": 0.02972438745200634,
      "learning_rate": 1e-06,
      "loss": 0.0061,
      "step": 708
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5406.0,
      "completions/mean_length": 3774.552734375,
      "completions/mean_terminated_length": 3288.590087890625,
      "completions/min_length": 762.0,
      "completions/min_terminated_length": 762.0,
      "entropy": 0.218715637922287,
      "epoch": 1.1362179487179487,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1383.7510986328125,
      "learning_rate": 1e-06,
      "loss": 0.0569,
      "num_tokens": 663134643.0,
      "reward": 0.18339823186397552,
      "reward_std": 0.08036380261182785,
      "rewards/progression_diversity/mean": -0.0005094322841614485,
      "rewards/progression_diversity/std": 0.008040121756494045,
      "rewards/symbolic_reward_accuracy/mean": 0.041015625,
      "rewards/symbolic_reward_accuracy/std": 0.19852031767368317,
      "rewards/symbolic_reward_partial_score/mean": 0.5416829586029053,
      "rewards/symbolic_reward_partial_score/std": 0.21756330132484436,
      "rewards/tag_count_reward/mean": -0.037109375,
      "rewards/tag_count_reward/std": 0.18921469151973724,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.038155198097229,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 3.4172184467315674,
      "step": 709
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.23037738353013992,
      "epoch": 1.1378205128205128,
      "grad_norm": 0.028344059363007545,
      "learning_rate": 1e-06,
      "loss": 0.0167,
      "step": 710
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.22201494872570038,
      "epoch": 1.1394230769230769,
      "grad_norm": 0.04986124113202095,
      "learning_rate": 1e-06,
      "loss": 0.0701,
      "step": 711
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.22669429332017899,
      "epoch": 1.141025641025641,
      "grad_norm": 0.03080112673342228,
      "learning_rate": 1e-06,
      "loss": 0.0666,
      "step": 712
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5973.0,
      "completions/mean_length": 3245.638671875,
      "completions/mean_terminated_length": 3089.847900390625,
      "completions/min_length": 517.0,
      "completions/min_terminated_length": 517.0,
      "entropy": 0.23609895259141922,
      "epoch": 1.142628205128205,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 188.2184295654297,
      "learning_rate": 1e-06,
      "loss": 0.0473,
      "num_tokens": 665642874.0,
      "reward": 0.21525543928146362,
      "reward_std": 0.0937977284193039,
      "rewards/progression_diversity/mean": -0.000335338176228106,
      "rewards/progression_diversity/std": 0.005407850258052349,
      "rewards/symbolic_reward_accuracy/mean": 0.08203125,
      "rewards/symbolic_reward_accuracy/std": 0.2746807038784027,
      "rewards/symbolic_reward_partial_score/mean": 0.557373046875,
      "rewards/symbolic_reward_partial_score/std": 0.2241562008857727,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0440669059753418,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.9209955930709839,
      "step": 713
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.23579012602567673,
      "epoch": 1.1442307692307692,
      "grad_norm": 0.031340159475803375,
      "learning_rate": 1e-06,
      "loss": -0.0103,
      "step": 714
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.23112714290618896,
      "epoch": 1.1458333333333333,
      "grad_norm": 0.03565249219536781,
      "learning_rate": 1e-06,
      "loss": 0.021,
      "step": 715
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.2343195155262947,
      "epoch": 1.1474358974358974,
      "grad_norm": 0.03568951413035393,
      "learning_rate": 1e-06,
      "loss": 0.0219,
      "step": 716
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5608.0,
      "completions/mean_length": 3687.26171875,
      "completions/mean_terminated_length": 3224.627685546875,
      "completions/min_length": 512.0,
      "completions/min_terminated_length": 512.0,
      "entropy": 0.23093228042125702,
      "epoch": 1.1490384615384615,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 25.515024185180664,
      "learning_rate": 1e-06,
      "loss": 0.008,
      "num_tokens": 668387632.0,
      "reward": 0.1549118608236313,
      "reward_std": 0.06852319091558456,
      "rewards/progression_diversity/mean": -0.001002308912575245,
      "rewards/progression_diversity/std": 0.015932733193039894,
      "rewards/symbolic_reward_accuracy/mean": 0.015625,
      "rewards/symbolic_reward_accuracy/std": 0.12414088100194931,
      "rewards/symbolic_reward_partial_score/mean": 0.49687501788139343,
      "rewards/symbolic_reward_partial_score/std": 0.20396628975868225,
      "rewards/tag_count_reward/mean": -0.03515625,
      "rewards/tag_count_reward/std": 0.1843547374010086,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.041079044342041,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 1.6645785570144653,
      "step": 717
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.22284479439258575,
      "epoch": 1.1506410256410255,
      "grad_norm": 0.03507667034864426,
      "learning_rate": 1e-06,
      "loss": 0.0831,
      "step": 718
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2239360809326172,
      "epoch": 1.1522435897435896,
      "grad_norm": 0.03187626227736473,
      "learning_rate": 1e-06,
      "loss": 0.0759,
      "step": 719
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2218421921133995,
      "epoch": 1.1538461538461537,
      "grad_norm": 0.05089287832379341,
      "learning_rate": 1e-06,
      "loss": 0.0547,
      "step": 720
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.025390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5998.0,
      "completions/mean_length": 3646.466796875,
      "completions/mean_terminated_length": 3314.627197265625,
      "completions/min_length": 710.0,
      "completions/min_terminated_length": 710.0,
      "entropy": 0.2157435417175293,
      "epoch": 1.155448717948718,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 491.061767578125,
      "learning_rate": 1e-06,
      "loss": 0.0744,
      "num_tokens": 671153135.0,
      "reward": 0.16443172097206116,
      "reward_std": 0.08347620069980621,
      "rewards/progression_diversity/mean": -0.00018694471509661525,
      "rewards/progression_diversity/std": 0.00423007644712925,
      "rewards/symbolic_reward_accuracy/mean": 0.025390625,
      "rewards/symbolic_reward_accuracy/std": 0.15746226906776428,
      "rewards/symbolic_reward_partial_score/mean": 0.5057942867279053,
      "rewards/symbolic_reward_partial_score/std": 0.22736532986164093,
      "rewards/tag_count_reward/mean": -0.025390625,
      "rewards/tag_count_reward/std": 0.15746226906776428,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0419472455978394,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.7793669700622559,
      "step": 721
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.23054082691669464,
      "epoch": 1.157051282051282,
      "grad_norm": 0.038431961089372635,
      "learning_rate": 1e-06,
      "loss": 0.0223,
      "step": 722
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.22331822663545609,
      "epoch": 1.1586538461538463,
      "grad_norm": 0.030653027817606926,
      "learning_rate": 1e-06,
      "loss": 0.0586,
      "step": 723
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.22740525007247925,
      "epoch": 1.1602564102564104,
      "grad_norm": 0.03620804473757744,
      "learning_rate": 1e-06,
      "loss": 0.0345,
      "step": 724
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5344.0,
      "completions/mean_length": 3325.42578125,
      "completions/mean_terminated_length": 3222.602294921875,
      "completions/min_length": 689.0,
      "completions/min_terminated_length": 689.0,
      "entropy": 0.2329612523317337,
      "epoch": 1.1618589743589745,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.042414553463459015,
      "learning_rate": 1e-06,
      "loss": 0.0642,
      "num_tokens": 673598329.0,
      "reward": 0.18591231107711792,
      "reward_std": 0.08399780839681625,
      "rewards/progression_diversity/mean": -7.887653919169679e-05,
      "rewards/progression_diversity/std": 0.0017324578948318958,
      "rewards/symbolic_reward_accuracy/mean": 0.048828125,
      "rewards/symbolic_reward_accuracy/std": 0.2157193273305893,
      "rewards/symbolic_reward_partial_score/mean": 0.524658203125,
      "rewards/symbolic_reward_partial_score/std": 0.23230838775634766,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0474889278411865,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 724.0,
      "sampling/sampling_logp_difference/mean": 0.47188305854797363,
      "step": 725
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.23176733404397964,
      "epoch": 1.1634615384615385,
      "grad_norm": 61.60709762573242,
      "learning_rate": 1e-06,
      "loss": 0.0217,
      "step": 726
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23428219556808472,
      "epoch": 1.1650641025641026,
      "grad_norm": 0.038381967693567276,
      "learning_rate": 1e-06,
      "loss": 0.0206,
      "step": 727
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.23280168324708939,
      "epoch": 1.1666666666666667,
      "grad_norm": 0.02906009368598461,
      "learning_rate": 1e-06,
      "loss": -0.0278,
      "step": 728
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4910.0,
      "completions/mean_length": 3052.662109375,
      "completions/mean_terminated_length": 2867.871337890625,
      "completions/min_length": 534.0,
      "completions/min_terminated_length": 534.0,
      "entropy": 0.2368728667497635,
      "epoch": 1.1682692307692308,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.03455471992492676,
      "learning_rate": 1e-06,
      "loss": 0.0136,
      "num_tokens": 676083596.0,
      "reward": 0.19645507633686066,
      "reward_std": 0.09786352515220642,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.0625,
      "rewards/symbolic_reward_accuracy/std": 0.2422981858253479,
      "rewards/symbolic_reward_partial_score/mean": 0.5344075560569763,
      "rewards/symbolic_reward_partial_score/std": 0.22556960582733154,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0460604429244995,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 0.4928211271762848,
      "step": 729
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2318621650338173,
      "epoch": 1.169871794871795,
      "grad_norm": 0.04372532665729523,
      "learning_rate": 1e-06,
      "loss": -0.0078,
      "step": 730
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.23245908319950104,
      "epoch": 1.171474358974359,
      "grad_norm": 0.10342668741941452,
      "learning_rate": 1e-06,
      "loss": 0.0665,
      "step": 731
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2263612002134323,
      "epoch": 1.1730769230769231,
      "grad_norm": 36.62001419067383,
      "learning_rate": 1e-06,
      "loss": 0.1059,
      "step": 732
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5687.0,
      "completions/mean_length": 3134.9296875,
      "completions/mean_terminated_length": 2951.279296875,
      "completions/min_length": 668.0,
      "completions/min_terminated_length": 668.0,
      "entropy": 0.23260176926851273,
      "epoch": 1.1746794871794872,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.0437169186770916,
      "learning_rate": 1e-06,
      "loss": 0.0039,
      "num_tokens": 678543112.0,
      "reward": 0.1655300259590149,
      "reward_std": 0.08380260318517685,
      "rewards/progression_diversity/mean": -0.0007075904868543148,
      "rewards/progression_diversity/std": 0.012582487426698208,
      "rewards/symbolic_reward_accuracy/mean": 0.03515625,
      "rewards/symbolic_reward_accuracy/std": 0.1843547374010086,
      "rewards/symbolic_reward_partial_score/mean": 0.48538413643836975,
      "rewards/symbolic_reward_partial_score/std": 0.22206903994083405,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.042715311050415,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.9027316570281982,
      "step": 733
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.2210148423910141,
      "epoch": 1.1762820512820513,
      "grad_norm": 185.5412139892578,
      "learning_rate": 1e-06,
      "loss": 0.0454,
      "step": 734
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.22284194827079773,
      "epoch": 1.1778846153846154,
      "grad_norm": 0.0355803407728672,
      "learning_rate": 1e-06,
      "loss": 0.0143,
      "step": 735
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.22447482496500015,
      "epoch": 1.1794871794871795,
      "grad_norm": 0.03734675049781799,
      "learning_rate": 1e-06,
      "loss": 0.0278,
      "step": 736
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5229.0,
      "completions/mean_length": 2980.09765625,
      "completions/mean_terminated_length": 2847.909423828125,
      "completions/min_length": 523.0,
      "completions/min_terminated_length": 523.0,
      "entropy": 0.22639226913452148,
      "epoch": 1.1810897435897436,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 0.03936491161584854,
      "learning_rate": 1e-06,
      "loss": 0.0262,
      "num_tokens": 681020794.0,
      "reward": 0.19800294935703278,
      "reward_std": 0.10915393382310867,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.052734375,
      "rewards/symbolic_reward_accuracy/std": 0.22372129559516907,
      "rewards/symbolic_reward_partial_score/mean": 0.5577962398529053,
      "rewards/symbolic_reward_partial_score/std": 0.2130732387304306,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.043833613395691,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 0.7718756794929504,
      "step": 737
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.22041287273168564,
      "epoch": 1.1826923076923077,
      "grad_norm": 0.025883983820676804,
      "learning_rate": 1e-06,
      "loss": -0.0021,
      "step": 738
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.22687563300132751,
      "epoch": 1.1842948717948718,
      "grad_norm": 0.030974598601460457,
      "learning_rate": 1e-06,
      "loss": 0.0084,
      "step": 739
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.22128045558929443,
      "epoch": 1.185897435897436,
      "grad_norm": 0.05905633792281151,
      "learning_rate": 1e-06,
      "loss": 0.0396,
      "step": 740
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4864.0,
      "completions/mean_length": 3153.501953125,
      "completions/mean_terminated_length": 3023.023681640625,
      "completions/min_length": 746.0,
      "completions/min_terminated_length": 746.0,
      "entropy": 0.23090147227048874,
      "epoch": 1.1875,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 331.0104064941406,
      "learning_rate": 1e-06,
      "loss": 0.0263,
      "num_tokens": 683450971.0,
      "reward": 0.16851550340652466,
      "reward_std": 0.07939761132001877,
      "rewards/progression_diversity/mean": -0.000989372143521905,
      "rewards/progression_diversity/std": 0.013076537288725376,
      "rewards/symbolic_reward_accuracy/mean": 0.0390625,
      "rewards/symbolic_reward_accuracy/std": 0.1939331740140915,
      "rewards/symbolic_reward_partial_score/mean": 0.48753252625465393,
      "rewards/symbolic_reward_partial_score/std": 0.20784252882003784,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0429353713989258,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 2.1165027618408203,
      "step": 741
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.22930524498224258,
      "epoch": 1.189102564102564,
      "grad_norm": 0.029941659420728683,
      "learning_rate": 1e-06,
      "loss": 0.041,
      "step": 742
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.22974687814712524,
      "epoch": 1.1907051282051282,
      "grad_norm": 0.025151744484901428,
      "learning_rate": 1e-06,
      "loss": 0.0187,
      "step": 743
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.2339945062994957,
      "epoch": 1.1923076923076923,
      "grad_norm": 0.02276208996772766,
      "learning_rate": 1e-06,
      "loss": 0.0261,
      "step": 744
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5222.0,
      "completions/mean_length": 3041.041015625,
      "completions/mean_terminated_length": 2909.45361328125,
      "completions/min_length": 678.0,
      "completions/min_terminated_length": 678.0,
      "entropy": 0.23491770029067993,
      "epoch": 1.1939102564102564,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.05074154585599899,
      "learning_rate": 1e-06,
      "loss": 0.0085,
      "num_tokens": 685896720.0,
      "reward": 0.21035856008529663,
      "reward_std": 0.08227989077568054,
      "rewards/progression_diversity/mean": -0.0007646906888112426,
      "rewards/progression_diversity/std": 0.010030530393123627,
      "rewards/symbolic_reward_accuracy/mean": 0.0625,
      "rewards/symbolic_reward_accuracy/std": 0.2422981858253479,
      "rewards/symbolic_reward_partial_score/mean": 0.5794759392738342,
      "rewards/symbolic_reward_partial_score/std": 0.1977068930864334,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0442752838134766,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 2.015608072280884,
      "step": 745
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.2312600165605545,
      "epoch": 1.1955128205128205,
      "grad_norm": 0.03379121422767639,
      "learning_rate": 1e-06,
      "loss": 0.0464,
      "step": 746
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2323080375790596,
      "epoch": 1.1971153846153846,
      "grad_norm": 0.02059108205139637,
      "learning_rate": 1e-06,
      "loss": 0.0359,
      "step": 747
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2359512448310852,
      "epoch": 1.1987179487179487,
      "grad_norm": 0.026128700003027916,
      "learning_rate": 1e-06,
      "loss": -0.0019,
      "step": 748
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.033203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5081.0,
      "completions/mean_length": 3201.248046875,
      "completions/mean_terminated_length": 2748.507080078125,
      "completions/min_length": 579.0,
      "completions/min_terminated_length": 579.0,
      "entropy": 0.21788237243890762,
      "epoch": 1.2003205128205128,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 229.65496826171875,
      "learning_rate": 1e-06,
      "loss": 0.0528,
      "num_tokens": 688491791.0,
      "reward": 0.17160627245903015,
      "reward_std": 0.09746118634939194,
      "rewards/progression_diversity/mean": -0.0005057294620200992,
      "rewards/progression_diversity/std": 0.005976199172437191,
      "rewards/symbolic_reward_accuracy/mean": 0.0390625,
      "rewards/symbolic_reward_accuracy/std": 0.1939331740140915,
      "rewards/symbolic_reward_partial_score/mean": 0.5030273795127869,
      "rewards/symbolic_reward_partial_score/std": 0.2094988375902176,
      "rewards/tag_count_reward/mean": -0.02734375,
      "rewards/tag_count_reward/std": 0.16324250400066376,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0365022420883179,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 3.8983547687530518,
      "step": 749
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.23051826655864716,
      "epoch": 1.2019230769230769,
      "grad_norm": 0.03440074622631073,
      "learning_rate": 1e-06,
      "loss": 0.0177,
      "step": 750
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.22294911742210388,
      "epoch": 1.203525641025641,
      "grad_norm": 0.028672844171524048,
      "learning_rate": 1e-06,
      "loss": 0.0646,
      "step": 751
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2228538542985916,
      "epoch": 1.205128205128205,
      "grad_norm": 36.360984802246094,
      "learning_rate": 1e-06,
      "loss": 0.0626,
      "step": 752
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5348.0,
      "completions/mean_length": 2809.646484375,
      "completions/mean_terminated_length": 2729.640625,
      "completions/min_length": 551.0,
      "completions/min_terminated_length": 551.0,
      "entropy": 0.23874470591545105,
      "epoch": 1.2067307692307692,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 0.05420081317424774,
      "learning_rate": 1e-06,
      "loss": 0.0125,
      "num_tokens": 690740314.0,
      "reward": 0.1803320348262787,
      "reward_std": 0.08125782012939453,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.041015625,
      "rewards/symbolic_reward_accuracy/std": 0.19852031767368317,
      "rewards/symbolic_reward_partial_score/mean": 0.5210286378860474,
      "rewards/symbolic_reward_partial_score/std": 0.19359782338142395,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0487172603607178,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 0.39224666357040405,
      "step": 753
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.2426346391439438,
      "epoch": 1.2083333333333333,
      "grad_norm": 106.4404296875,
      "learning_rate": 1e-06,
      "loss": 0.0278,
      "step": 754
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.23718024045228958,
      "epoch": 1.2099358974358974,
      "grad_norm": 0.03360147029161453,
      "learning_rate": 1e-06,
      "loss": 0.0182,
      "step": 755
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.24055323004722595,
      "epoch": 1.2115384615384615,
      "grad_norm": 0.028352908790111542,
      "learning_rate": 1e-06,
      "loss": -0.0133,
      "step": 756
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.021484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4704.0,
      "completions/mean_length": 2882.4375,
      "completions/mean_terminated_length": 2585.99609375,
      "completions/min_length": 550.0,
      "completions/min_terminated_length": 550.0,
      "entropy": 0.2231486737728119,
      "epoch": 1.2131410256410255,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 601.402587890625,
      "learning_rate": 1e-06,
      "loss": 0.0973,
      "num_tokens": 693148362.0,
      "reward": 0.25671058893203735,
      "reward_std": 0.10707536339759827,
      "rewards/progression_diversity/mean": -0.000815525883808732,
      "rewards/progression_diversity/std": 0.010911921970546246,
      "rewards/symbolic_reward_accuracy/mean": 0.134765625,
      "rewards/symbolic_reward_accuracy/std": 0.3418070077896118,
      "rewards/symbolic_reward_partial_score/mean": 0.5927083492279053,
      "rewards/symbolic_reward_partial_score/std": 0.24597984552383423,
      "rewards/tag_count_reward/mean": -0.01953125,
      "rewards/tag_count_reward/std": 0.1385180652141571,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0398728847503662,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 3.051284074783325,
      "step": 757
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.23185677826404572,
      "epoch": 1.2147435897435896,
      "grad_norm": 0.03557441011071205,
      "learning_rate": 1e-06,
      "loss": 0.0471,
      "step": 758
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.23132619261741638,
      "epoch": 1.2163461538461537,
      "grad_norm": 0.018664143979549408,
      "learning_rate": 1e-06,
      "loss": 0.0251,
      "step": 759
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.2319474294781685,
      "epoch": 1.217948717948718,
      "grad_norm": 0.017118403688073158,
      "learning_rate": 1e-06,
      "loss": 0.0058,
      "step": 760
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5218.0,
      "completions/mean_length": 2768.322265625,
      "completions/mean_terminated_length": 2606.87158203125,
      "completions/min_length": 462.0,
      "completions/min_terminated_length": 462.0,
      "entropy": 0.2300676926970482,
      "epoch": 1.219551282051282,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 52.74298095703125,
      "learning_rate": 1e-06,
      "loss": 0.0637,
      "num_tokens": 695437359.0,
      "reward": 0.18998649716377258,
      "reward_std": 0.09169755131006241,
      "rewards/progression_diversity/mean": -0.00037375889951363206,
      "rewards/progression_diversity/std": 0.00622538710013032,
      "rewards/symbolic_reward_accuracy/mean": 0.0546875,
      "rewards/symbolic_reward_accuracy/std": 0.2275916188955307,
      "rewards/symbolic_reward_partial_score/mean": 0.52783203125,
      "rewards/symbolic_reward_partial_score/std": 0.2111833095550537,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045241117477417,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.6731740236282349,
      "step": 761
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.23460698127746582,
      "epoch": 1.2211538461538463,
      "grad_norm": 1694.8343505859375,
      "learning_rate": 1e-06,
      "loss": 0.1124,
      "step": 762
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.24004460126161575,
      "epoch": 1.2227564102564104,
      "grad_norm": 0.038948699831962585,
      "learning_rate": 1e-06,
      "loss": 0.0137,
      "step": 763
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.24492079764604568,
      "epoch": 1.2243589743589745,
      "grad_norm": 0.02253109961748123,
      "learning_rate": 1e-06,
      "loss": 0.0056,
      "step": 764
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4993.0,
      "completions/mean_length": 2854.958984375,
      "completions/mean_terminated_length": 2748.43115234375,
      "completions/min_length": 506.0,
      "completions/min_terminated_length": 506.0,
      "entropy": 0.2323237955570221,
      "epoch": 1.2259615384615385,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1602.8712158203125,
      "learning_rate": 1e-06,
      "loss": 0.0338,
      "num_tokens": 697774042.0,
      "reward": 0.187491774559021,
      "reward_std": 0.09734513610601425,
      "rewards/progression_diversity/mean": -0.00033558663562871516,
      "rewards/progression_diversity/std": 0.005632203537970781,
      "rewards/symbolic_reward_accuracy/mean": 0.048828125,
      "rewards/symbolic_reward_accuracy/std": 0.2157193273305893,
      "rewards/symbolic_reward_partial_score/mean": 0.5299316644668579,
      "rewards/symbolic_reward_partial_score/std": 0.21817044913768768,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0454214811325073,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.2177436351776123,
      "step": 765
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.22870472818613052,
      "epoch": 1.2275641025641026,
      "grad_norm": 0.02481686696410179,
      "learning_rate": 1e-06,
      "loss": 0.0111,
      "step": 766
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2385152205824852,
      "epoch": 1.2291666666666667,
      "grad_norm": 0.0412432886660099,
      "learning_rate": 1e-06,
      "loss": -0.0112,
      "step": 767
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.23105958104133606,
      "epoch": 1.2307692307692308,
      "grad_norm": 0.04013615846633911,
      "learning_rate": 1e-06,
      "loss": 0.0401,
      "step": 768
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5212.0,
      "completions/mean_length": 3133.908203125,
      "completions/mean_terminated_length": 2896.828857421875,
      "completions/min_length": 450.0,
      "completions/min_terminated_length": 450.0,
      "entropy": 0.2316986247897148,
      "epoch": 1.232371794871795,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 384.80572509765625,
      "learning_rate": 1e-06,
      "loss": 0.0112,
      "num_tokens": 700265723.0,
      "reward": 0.19464799761772156,
      "reward_std": 0.07056570053100586,
      "rewards/progression_diversity/mean": -0.0005334917223080993,
      "rewards/progression_diversity/std": 0.00818613264709711,
      "rewards/symbolic_reward_accuracy/mean": 0.044921875,
      "rewards/symbolic_reward_accuracy/std": 0.20733514428138733,
      "rewards/symbolic_reward_partial_score/mean": 0.5648600459098816,
      "rewards/symbolic_reward_partial_score/std": 0.20340082049369812,
      "rewards/tag_count_reward/mean": -0.017578125,
      "rewards/tag_count_reward/std": 0.13154059648513794,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.042616367340088,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 2.089385509490967,
      "step": 769
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.23712345212697983,
      "epoch": 1.233974358974359,
      "grad_norm": 0.03780589625239372,
      "learning_rate": 1e-06,
      "loss": 0.0076,
      "step": 770
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.2268049567937851,
      "epoch": 1.2355769230769231,
      "grad_norm": 0.028598081320524216,
      "learning_rate": 1e-06,
      "loss": 0.0633,
      "step": 771
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.22704008221626282,
      "epoch": 1.2371794871794872,
      "grad_norm": 0.036604560911655426,
      "learning_rate": 1e-06,
      "loss": 0.0556,
      "step": 772
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5111.0,
      "completions/mean_length": 2862.83203125,
      "completions/mean_terminated_length": 2729.4873046875,
      "completions/min_length": 484.0,
      "completions/min_terminated_length": 484.0,
      "entropy": 0.23467671126127243,
      "epoch": 1.2387820512820513,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 248.04339599609375,
      "learning_rate": 1e-06,
      "loss": 0.0378,
      "num_tokens": 702571989.0,
      "reward": 0.16636018455028534,
      "reward_std": 0.07103414833545685,
      "rewards/progression_diversity/mean": -0.00021245863172225654,
      "rewards/progression_diversity/std": 0.004727643448859453,
      "rewards/symbolic_reward_accuracy/mean": 0.017578125,
      "rewards/symbolic_reward_accuracy/std": 0.13154059648513794,
      "rewards/symbolic_reward_partial_score/mean": 0.5232909917831421,
      "rewards/symbolic_reward_partial_score/std": 0.18974992632865906,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0468013286590576,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.2262119054794312,
      "step": 773
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.2362857237458229,
      "epoch": 1.2403846153846154,
      "grad_norm": 0.048055924475193024,
      "learning_rate": 1e-06,
      "loss": 0.0191,
      "step": 774
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.23780542612075806,
      "epoch": 1.2419871794871795,
      "grad_norm": 0.02682042494416237,
      "learning_rate": 1e-06,
      "loss": -0.0151,
      "step": 775
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.2368232011795044,
      "epoch": 1.2435897435897436,
      "grad_norm": 0.09740027785301208,
      "learning_rate": 1e-06,
      "loss": 0.026,
      "step": 776
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5104.0,
      "completions/mean_length": 2685.009765625,
      "completions/mean_terminated_length": 2577.143798828125,
      "completions/min_length": 489.0,
      "completions/min_terminated_length": 489.0,
      "entropy": 0.2424686774611473,
      "epoch": 1.2451923076923077,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 399.4440612792969,
      "learning_rate": 1e-06,
      "loss": 0.0333,
      "num_tokens": 704759194.0,
      "reward": 0.1672852635383606,
      "reward_std": 0.0681874230504036,
      "rewards/progression_diversity/mean": -0.0009661708609201014,
      "rewards/progression_diversity/std": 0.01556091196835041,
      "rewards/symbolic_reward_accuracy/mean": 0.025390625,
      "rewards/symbolic_reward_accuracy/std": 0.15746226906776428,
      "rewards/symbolic_reward_partial_score/mean": 0.5088216066360474,
      "rewards/symbolic_reward_partial_score/std": 0.19267694652080536,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0457208156585693,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 2.5860910415649414,
      "step": 777
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.24405201524496078,
      "epoch": 1.2467948717948718,
      "grad_norm": 0.040985848754644394,
      "learning_rate": 1e-06,
      "loss": 0.0082,
      "step": 778
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.24193061143159866,
      "epoch": 1.248397435897436,
      "grad_norm": 0.029141481965780258,
      "learning_rate": 1e-06,
      "loss": 0.0323,
      "step": 779
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.24474172294139862,
      "epoch": 1.25,
      "grad_norm": 0.03860503062605858,
      "learning_rate": 1e-06,
      "loss": 0.005,
      "step": 780
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4847.0,
      "completions/mean_length": 2609.923828125,
      "completions/mean_terminated_length": 2555.907958984375,
      "completions/min_length": 534.0,
      "completions/min_terminated_length": 534.0,
      "entropy": 0.23170283436775208,
      "epoch": 1.251602564102564,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 0.0442928783595562,
      "learning_rate": 1e-06,
      "loss": 0.0059,
      "num_tokens": 707031571.0,
      "reward": 0.18274357914924622,
      "reward_std": 0.06679178774356842,
      "rewards/progression_diversity/mean": -0.0005459238309413195,
      "rewards/progression_diversity/std": 0.008116460405290127,
      "rewards/symbolic_reward_accuracy/mean": 0.0234375,
      "rewards/symbolic_reward_accuracy/std": 0.15143637359142303,
      "rewards/symbolic_reward_partial_score/mean": 0.5635905265808105,
      "rewards/symbolic_reward_partial_score/std": 0.20824402570724487,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045533299446106,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.4784348011016846,
      "step": 781
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.23342838883399963,
      "epoch": 1.2532051282051282,
      "grad_norm": 0.03521363064646721,
      "learning_rate": 1e-06,
      "loss": 0.0218,
      "step": 782
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2293540984392166,
      "epoch": 1.2548076923076923,
      "grad_norm": 0.028306540101766586,
      "learning_rate": 1e-06,
      "loss": -0.0201,
      "step": 783
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.22597584128379822,
      "epoch": 1.2564102564102564,
      "grad_norm": 0.024497980251908302,
      "learning_rate": 1e-06,
      "loss": 0.0313,
      "step": 784
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4343.0,
      "completions/mean_length": 2555.701171875,
      "completions/mean_terminated_length": 2474.198486328125,
      "completions/min_length": 456.0,
      "completions/min_terminated_length": 456.0,
      "entropy": 0.22865238785743713,
      "epoch": 1.2580128205128205,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 0.03943171352148056,
      "learning_rate": 1e-06,
      "loss": -0.0009,
      "num_tokens": 709195418.0,
      "reward": 0.193578839302063,
      "reward_std": 0.087828129529953,
      "rewards/progression_diversity/mean": -2.721707096497994e-05,
      "rewards/progression_diversity/std": 0.0006158520118333399,
      "rewards/symbolic_reward_accuracy/mean": 0.048828125,
      "rewards/symbolic_reward_accuracy/std": 0.2157193273305893,
      "rewards/symbolic_reward_partial_score/mean": 0.549560546875,
      "rewards/symbolic_reward_partial_score/std": 0.19566665589809418,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045490026473999,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.5368279218673706,
      "step": 785
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.22690270096063614,
      "epoch": 1.2596153846153846,
      "grad_norm": 0.0247288029640913,
      "learning_rate": 1e-06,
      "loss": 0.0111,
      "step": 786
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2224835455417633,
      "epoch": 1.2612179487179487,
      "grad_norm": 0.03503105789422989,
      "learning_rate": 1e-06,
      "loss": 0.0801,
      "step": 787
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.22508803755044937,
      "epoch": 1.2628205128205128,
      "grad_norm": 0.033731210976839066,
      "learning_rate": 1e-06,
      "loss": -0.0029,
      "step": 788
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4441.0,
      "completions/mean_length": 2418.279296875,
      "completions/mean_terminated_length": 2280.55029296875,
      "completions/min_length": 385.0,
      "completions/min_terminated_length": 385.0,
      "entropy": 0.2210671380162239,
      "epoch": 1.2644230769230769,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 72.21678924560547,
      "learning_rate": 1e-06,
      "loss": 0.0407,
      "num_tokens": 711362041.0,
      "reward": 0.17944911122322083,
      "reward_std": 0.08195337653160095,
      "rewards/progression_diversity/mean": -0.00040262757102027535,
      "rewards/progression_diversity/std": 0.006931956857442856,
      "rewards/symbolic_reward_accuracy/mean": 0.03125,
      "rewards/symbolic_reward_accuracy/std": 0.17416280508041382,
      "rewards/symbolic_reward_partial_score/mean": 0.5389323234558105,
      "rewards/symbolic_reward_partial_score/std": 0.18118998408317566,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0432648658752441,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.1448458433151245,
      "step": 789
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.22485744953155518,
      "epoch": 1.266025641025641,
      "grad_norm": 0.02760397456586361,
      "learning_rate": 1e-06,
      "loss": 0.021,
      "step": 790
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.22400467097759247,
      "epoch": 1.267628205128205,
      "grad_norm": 0.031739212572574615,
      "learning_rate": 1e-06,
      "loss": 0.0194,
      "step": 791
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2254156917333603,
      "epoch": 1.2692307692307692,
      "grad_norm": 0.03261100500822067,
      "learning_rate": 1e-06,
      "loss": 0.0151,
      "step": 792
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4488.0,
      "completions/mean_length": 2580.109375,
      "completions/mean_terminated_length": 2388.768310546875,
      "completions/min_length": 580.0,
      "completions/min_terminated_length": 580.0,
      "entropy": 0.2238694727420807,
      "epoch": 1.2708333333333333,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 277.4549255371094,
      "learning_rate": 1e-06,
      "loss": 0.053,
      "num_tokens": 713579937.0,
      "reward": 0.18201328814029694,
      "reward_std": 0.0664268359541893,
      "rewards/progression_diversity/mean": -0.0013090292923152447,
      "rewards/progression_diversity/std": 0.013768412172794342,
      "rewards/symbolic_reward_accuracy/mean": 0.021484375,
      "rewards/symbolic_reward_accuracy/std": 0.14513419568538666,
      "rewards/symbolic_reward_partial_score/mean": 0.5670410394668579,
      "rewards/symbolic_reward_partial_score/std": 0.1917123794555664,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0389268398284912,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 2.950878143310547,
      "step": 793
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.2173541858792305,
      "epoch": 1.2724358974358974,
      "grad_norm": 745.087890625,
      "learning_rate": 1e-06,
      "loss": 0.0668,
      "step": 794
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.22230473905801773,
      "epoch": 1.2740384615384617,
      "grad_norm": 0.03551755100488663,
      "learning_rate": 1e-06,
      "loss": 0.0133,
      "step": 795
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.22595015913248062,
      "epoch": 1.2756410256410255,
      "grad_norm": 0.025852041319012642,
      "learning_rate": 1e-06,
      "loss": -0.0077,
      "step": 796
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4958.0,
      "completions/mean_length": 2487.400390625,
      "completions/mean_terminated_length": 2405.4951171875,
      "completions/min_length": 417.0,
      "completions/min_terminated_length": 417.0,
      "entropy": 0.22661074995994568,
      "epoch": 1.2772435897435899,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 0.04282950237393379,
      "learning_rate": 1e-06,
      "loss": -0.0107,
      "num_tokens": 715731470.0,
      "reward": 0.24059391021728516,
      "reward_std": 0.08692439645528793,
      "rewards/progression_diversity/mean": -0.00017893135373014957,
      "rewards/progression_diversity/std": 0.0020902305841445923,
      "rewards/symbolic_reward_accuracy/mean": 0.10546875,
      "rewards/symbolic_reward_accuracy/std": 0.3074568510055542,
      "rewards/symbolic_reward_partial_score/mean": 0.5930013060569763,
      "rewards/symbolic_reward_partial_score/std": 0.2335280328989029,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.043853759765625,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.3285684883594513,
      "step": 797
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2201601192355156,
      "epoch": 1.2788461538461537,
      "grad_norm": 0.03756594657897949,
      "learning_rate": 1e-06,
      "loss": 0.0067,
      "step": 798
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2165718972682953,
      "epoch": 1.280448717948718,
      "grad_norm": 0.02026909776031971,
      "learning_rate": 1e-06,
      "loss": 0.0363,
      "step": 799
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2138141244649887,
      "epoch": 1.282051282051282,
      "grad_norm": 0.021396497264504433,
      "learning_rate": 1e-06,
      "loss": 0.0612,
      "step": 800
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5352.0,
      "completions/mean_length": 2308.837890625,
      "completions/mean_terminated_length": 2198.009765625,
      "completions/min_length": 462.0,
      "completions/min_terminated_length": 462.0,
      "entropy": 0.2189565673470497,
      "epoch": 1.2836538461538463,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.04113735631108284,
      "learning_rate": 1e-06,
      "loss": -0.0137,
      "num_tokens": 717814763.0,
      "reward": 0.18618622422218323,
      "reward_std": 0.06758680939674377,
      "rewards/progression_diversity/mean": -0.0005197992431931198,
      "rewards/progression_diversity/std": 0.008483768440783024,
      "rewards/symbolic_reward_accuracy/mean": 0.0390625,
      "rewards/symbolic_reward_accuracy/std": 0.1939331740140915,
      "rewards/symbolic_reward_partial_score/mean": 0.545117199420929,
      "rewards/symbolic_reward_partial_score/std": 0.2002282440662384,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0427607297897339,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.610470175743103,
      "step": 801
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.21421056240797043,
      "epoch": 1.2852564102564101,
      "grad_norm": 0.02349807135760784,
      "learning_rate": 1e-06,
      "loss": 0.0212,
      "step": 802
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.21766190230846405,
      "epoch": 1.2868589743589745,
      "grad_norm": 0.024225391447544098,
      "learning_rate": 1e-06,
      "loss": 0.0283,
      "step": 803
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.22310616821050644,
      "epoch": 1.2884615384615383,
      "grad_norm": 0.05020375922322273,
      "learning_rate": 1e-06,
      "loss": 0.0196,
      "step": 804
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4344.0,
      "completions/max_terminated_length": 4344.0,
      "completions/mean_length": 2237.517578125,
      "completions/mean_terminated_length": 2237.517578125,
      "completions/min_length": 442.0,
      "completions/min_terminated_length": 442.0,
      "entropy": 0.22092020511627197,
      "epoch": 1.2900641025641026,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.034722164273262024,
      "learning_rate": 1e-06,
      "loss": 0.0008,
      "num_tokens": 719703236.0,
      "reward": 0.24485130608081818,
      "reward_std": 0.12719884514808655,
      "rewards/progression_diversity/mean": -0.00022179064399097115,
      "rewards/progression_diversity/std": 0.003463014727458358,
      "rewards/symbolic_reward_accuracy/mean": 0.11328125,
      "rewards/symbolic_reward_accuracy/std": 0.3172462284564972,
      "rewards/symbolic_reward_partial_score/mean": 0.5896158814430237,
      "rewards/symbolic_reward_partial_score/std": 0.21620358526706696,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0438671112060547,
      "sampling/importance_sampling_ratio/min": 1.0395459602315473e-15,
      "sampling/sampling_logp_difference/max": 34.49999237060547,
      "sampling/sampling_logp_difference/mean": 0.09283532202243805,
      "step": 805
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.22089938074350357,
      "epoch": 1.2916666666666667,
      "grad_norm": 0.027210041880607605,
      "learning_rate": 1e-06,
      "loss": -0.0033,
      "step": 806
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.21874121576547623,
      "epoch": 1.2932692307692308,
      "grad_norm": 0.024275539442896843,
      "learning_rate": 1e-06,
      "loss": 0.01,
      "step": 807
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.03125,
      "clip_ratio/low_mean": 0.34375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.21499209105968475,
      "epoch": 1.294871794871795,
      "grad_norm": 0.03609732910990715,
      "learning_rate": 1e-06,
      "loss": 0.0045,
      "step": 808
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4430.0,
      "completions/mean_length": 2284.5625,
      "completions/mean_terminated_length": 2173.543212890625,
      "completions/min_length": 444.0,
      "completions/min_terminated_length": 444.0,
      "entropy": 0.21107368171215057,
      "epoch": 1.296474358974359,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.03534679487347603,
      "learning_rate": 1e-06,
      "loss": -0.0077,
      "num_tokens": 721772084.0,
      "reward": 0.1930440068244934,
      "reward_std": 0.0763799324631691,
      "rewards/progression_diversity/mean": -0.0007751868688501418,
      "rewards/progression_diversity/std": 0.01335845235735178,
      "rewards/symbolic_reward_accuracy/mean": 0.046875,
      "rewards/symbolic_reward_accuracy/std": 0.21157780289649963,
      "rewards/symbolic_reward_partial_score/mean": 0.5523600578308105,
      "rewards/symbolic_reward_partial_score/std": 0.20780852437019348,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.040164828300476,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 0.9508891105651855,
      "step": 809
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.21084369719028473,
      "epoch": 1.2980769230769231,
      "grad_norm": 0.031098656356334686,
      "learning_rate": 1e-06,
      "loss": 0.0288,
      "step": 810
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.2036626636981964,
      "epoch": 1.2996794871794872,
      "grad_norm": 0.04239289462566376,
      "learning_rate": 1e-06,
      "loss": 0.0488,
      "step": 811
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.20902617275714874,
      "epoch": 1.3012820512820513,
      "grad_norm": 0.03809356689453125,
      "learning_rate": 1e-06,
      "loss": 0.0013,
      "step": 812
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5018.0,
      "completions/mean_length": 2127.1875,
      "completions/mean_terminated_length": 1986.5877685546875,
      "completions/min_length": 377.0,
      "completions/min_terminated_length": 377.0,
      "entropy": 0.20260044187307358,
      "epoch": 1.3028846153846154,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 216.08346557617188,
      "learning_rate": 1e-06,
      "loss": 0.0261,
      "num_tokens": 723870292.0,
      "reward": 0.17801672220230103,
      "reward_std": 0.07631928473711014,
      "rewards/progression_diversity/mean": -8.632877870695665e-05,
      "rewards/progression_diversity/std": 0.0017233239486813545,
      "rewards/symbolic_reward_accuracy/mean": 0.0234375,
      "rewards/symbolic_reward_accuracy/std": 0.15143637359142303,
      "rewards/symbolic_reward_partial_score/mean": 0.5497721433639526,
      "rewards/symbolic_reward_partial_score/std": 0.19805040955543518,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.038082242012024,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.1360535621643066,
      "step": 813
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.20947977900505066,
      "epoch": 1.3044871794871795,
      "grad_norm": 0.03403943032026291,
      "learning_rate": 1e-06,
      "loss": 0.0095,
      "step": 814
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.19642490148544312,
      "epoch": 1.3060897435897436,
      "grad_norm": 0.020816028118133545,
      "learning_rate": 1e-06,
      "loss": 0.0554,
      "step": 815
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.20714019238948822,
      "epoch": 1.3076923076923077,
      "grad_norm": 0.02245701290667057,
      "learning_rate": 1e-06,
      "loss": -0.0076,
      "step": 816
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4542.0,
      "completions/mean_length": 2476.328125,
      "completions/mean_terminated_length": 2339.171630859375,
      "completions/min_length": 549.0,
      "completions/min_terminated_length": 549.0,
      "entropy": 0.20043767243623734,
      "epoch": 1.3092948717948718,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 203.6943817138672,
      "learning_rate": 1e-06,
      "loss": 0.0633,
      "num_tokens": 725929244.0,
      "reward": 0.2166806012392044,
      "reward_std": 0.10392580181360245,
      "rewards/progression_diversity/mean": -0.0003975919389631599,
      "rewards/progression_diversity/std": 0.004248625598847866,
      "rewards/symbolic_reward_accuracy/mean": 0.087890625,
      "rewards/symbolic_reward_accuracy/std": 0.2834126651287079,
      "rewards/symbolic_reward_partial_score/mean": 0.549755871295929,
      "rewards/symbolic_reward_partial_score/std": 0.21964266896247864,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.03891921043396,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.3425063192844391,
      "step": 817
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.2018059492111206,
      "epoch": 1.310897435897436,
      "grad_norm": 0.02881396934390068,
      "learning_rate": 1e-06,
      "loss": 0.0388,
      "step": 818
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.20520929992198944,
      "epoch": 1.3125,
      "grad_norm": 0.029828235507011414,
      "learning_rate": 1e-06,
      "loss": 0.0124,
      "step": 819
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.20675267279148102,
      "epoch": 1.314102564102564,
      "grad_norm": 0.02294883131980896,
      "learning_rate": 1e-06,
      "loss": -0.0073,
      "step": 820
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4985.0,
      "completions/mean_length": 2255.87109375,
      "completions/mean_terminated_length": 2228.22314453125,
      "completions/min_length": 510.0,
      "completions/min_terminated_length": 510.0,
      "entropy": 0.20751511305570602,
      "epoch": 1.3157051282051282,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 0.03021460585296154,
      "learning_rate": 1e-06,
      "loss": 0.0083,
      "num_tokens": 727907322.0,
      "reward": 0.21369871497154236,
      "reward_std": 0.09200935065746307,
      "rewards/progression_diversity/mean": -0.00024685371317900717,
      "rewards/progression_diversity/std": 0.004451336804777384,
      "rewards/symbolic_reward_accuracy/mean": 0.078125,
      "rewards/symbolic_reward_accuracy/std": 0.26863065361976624,
      "rewards/symbolic_reward_partial_score/mean": 0.5567382574081421,
      "rewards/symbolic_reward_partial_score/std": 0.222740039229393,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.040090799331665,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 0.3165525197982788,
      "step": 821
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.20631252229213715,
      "epoch": 1.3173076923076923,
      "grad_norm": 0.021766817197203636,
      "learning_rate": 1e-06,
      "loss": -0.0018,
      "step": 822
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.20735712349414825,
      "epoch": 1.3189102564102564,
      "grad_norm": 0.024145064875483513,
      "learning_rate": 1e-06,
      "loss": 0.0377,
      "step": 823
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.20985466986894608,
      "epoch": 1.3205128205128205,
      "grad_norm": 0.030014174059033394,
      "learning_rate": 1e-06,
      "loss": -0.0084,
      "step": 824
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4514.0,
      "completions/mean_length": 2213.6015625,
      "completions/mean_terminated_length": 2130.08251953125,
      "completions/min_length": 412.0,
      "completions/min_terminated_length": 412.0,
      "entropy": 0.21330777555704117,
      "epoch": 1.3221153846153846,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.032049596309661865,
      "learning_rate": 1e-06,
      "loss": 0.0345,
      "num_tokens": 729936686.0,
      "reward": 0.18624354898929596,
      "reward_std": 0.09241214394569397,
      "rewards/progression_diversity/mean": -0.000645598629489541,
      "rewards/progression_diversity/std": 0.011169749312102795,
      "rewards/symbolic_reward_accuracy/mean": 0.0546875,
      "rewards/symbolic_reward_accuracy/std": 0.2275916188955307,
      "rewards/symbolic_reward_partial_score/mean": 0.5134114027023315,
      "rewards/symbolic_reward_partial_score/std": 0.2185731679201126,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0392979383468628,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 1.1306374073028564,
      "step": 825
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.20698269456624985,
      "epoch": 1.3237179487179487,
      "grad_norm": 382.0038757324219,
      "learning_rate": 1e-06,
      "loss": 0.04,
      "step": 826
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2093319445848465,
      "epoch": 1.3253205128205128,
      "grad_norm": 0.02666737698018551,
      "learning_rate": 1e-06,
      "loss": 0.0096,
      "step": 827
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2111140564084053,
      "epoch": 1.3269230769230769,
      "grad_norm": 172.5316925048828,
      "learning_rate": 1e-06,
      "loss": 0.0007,
      "step": 828
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4570.0,
      "completions/mean_length": 2048.4921875,
      "completions/mean_terminated_length": 2020.4383544921875,
      "completions/min_length": 474.0,
      "completions/min_terminated_length": 474.0,
      "entropy": 0.22205299139022827,
      "epoch": 1.328525641025641,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.03405119851231575,
      "learning_rate": 1e-06,
      "loss": 0.0065,
      "num_tokens": 731851850.0,
      "reward": 0.20125804841518402,
      "reward_std": 0.0859595537185669,
      "rewards/progression_diversity/mean": -0.00017142272554337978,
      "rewards/progression_diversity/std": 0.0027693838346749544,
      "rewards/symbolic_reward_accuracy/mean": 0.056640625,
      "rewards/symbolic_reward_accuracy/std": 0.23138070106506348,
      "rewards/symbolic_reward_partial_score/mean": 0.5582357048988342,
      "rewards/symbolic_reward_partial_score/std": 0.21341325342655182,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0437791347503662,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.45466187596321106,
      "step": 829
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.22545113414525986,
      "epoch": 1.330128205128205,
      "grad_norm": 0.03629428148269653,
      "learning_rate": 1e-06,
      "loss": 0.0079,
      "step": 830
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.22470547258853912,
      "epoch": 1.3317307692307692,
      "grad_norm": 0.02873636782169342,
      "learning_rate": 1e-06,
      "loss": 0.0038,
      "step": 831
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2245195508003235,
      "epoch": 1.3333333333333333,
      "grad_norm": 0.027126438915729523,
      "learning_rate": 1e-06,
      "loss": 0.0426,
      "step": 832
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8032.0,
      "completions/mean_length": 1912.681640625,
      "completions/mean_terminated_length": 1884.362060546875,
      "completions/min_length": 492.0,
      "completions/min_terminated_length": 492.0,
      "entropy": 0.2240256890654564,
      "epoch": 1.3349358974358974,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 0.03682028129696846,
      "learning_rate": 1e-06,
      "loss": 0.0219,
      "num_tokens": 733749047.0,
      "reward": 0.20436623692512512,
      "reward_std": 0.07025286555290222,
      "rewards/progression_diversity/mean": -0.0003872342931572348,
      "rewards/progression_diversity/std": 0.005586910527199507,
      "rewards/symbolic_reward_accuracy/mean": 0.048828125,
      "rewards/symbolic_reward_accuracy/std": 0.2157193273305893,
      "rewards/symbolic_reward_partial_score/mean": 0.584228515625,
      "rewards/symbolic_reward_partial_score/std": 0.201664999127388,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0452286005020142,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.28562289476394653,
      "step": 833
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.22573984414339066,
      "epoch": 1.3365384615384617,
      "grad_norm": 0.023644646629691124,
      "learning_rate": 1e-06,
      "loss": -0.0002,
      "step": 834
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2348889708518982,
      "epoch": 1.3381410256410255,
      "grad_norm": 0.0235019288957119,
      "learning_rate": 1e-06,
      "loss": -0.0061,
      "step": 835
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.2224627062678337,
      "epoch": 1.3397435897435899,
      "grad_norm": 0.029762422665953636,
      "learning_rate": 1e-06,
      "loss": 0.0313,
      "step": 836
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4299.0,
      "completions/max_terminated_length": 4299.0,
      "completions/mean_length": 1805.880859375,
      "completions/mean_terminated_length": 1805.880859375,
      "completions/min_length": 452.0,
      "completions/min_terminated_length": 452.0,
      "entropy": 0.23368220031261444,
      "epoch": 1.3413461538461537,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 0.034925542771816254,
      "learning_rate": 1e-06,
      "loss": 0.0211,
      "num_tokens": 735455482.0,
      "reward": 0.2690373361110687,
      "reward_std": 0.11118809878826141,
      "rewards/progression_diversity/mean": -7.453490979969501e-05,
      "rewards/progression_diversity/std": 0.0010274524101987481,
      "rewards/symbolic_reward_accuracy/mean": 0.130859375,
      "rewards/symbolic_reward_accuracy/std": 0.33757632970809937,
      "rewards/symbolic_reward_partial_score/mean": 0.6350748538970947,
      "rewards/symbolic_reward_partial_score/std": 0.2342989295721054,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.046823501586914,
      "sampling/importance_sampling_ratio/min": 0.0010576416971161962,
      "sampling/sampling_logp_difference/max": 6.85171365737915,
      "sampling/sampling_logp_difference/mean": 0.0981544554233551,
      "step": 837
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.23005392402410507,
      "epoch": 1.342948717948718,
      "grad_norm": 0.02568940818309784,
      "learning_rate": 1e-06,
      "loss": 0.0137,
      "step": 838
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.22952641546726227,
      "epoch": 1.344551282051282,
      "grad_norm": 0.030425578355789185,
      "learning_rate": 1e-06,
      "loss": -0.0307,
      "step": 839
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.22608061879873276,
      "epoch": 1.3461538461538463,
      "grad_norm": 0.03146536648273468,
      "learning_rate": 1e-06,
      "loss": 0.042,
      "step": 840
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4922.0,
      "completions/mean_length": 1662.537109375,
      "completions/mean_terminated_length": 1604.8060302734375,
      "completions/min_length": 464.0,
      "completions/min_terminated_length": 464.0,
      "entropy": 0.23019438982009888,
      "epoch": 1.3477564102564101,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 86.43042755126953,
      "learning_rate": 1e-06,
      "loss": 0.0126,
      "num_tokens": 737145805.0,
      "reward": 0.24045243859291077,
      "reward_std": 0.12287962436676025,
      "rewards/progression_diversity/mean": -0.0006555759464390576,
      "rewards/progression_diversity/std": 0.009774475358426571,
      "rewards/symbolic_reward_accuracy/mean": 0.099609375,
      "rewards/symbolic_reward_accuracy/std": 0.29977133870124817,
      "rewards/symbolic_reward_partial_score/mean": 0.6036132574081421,
      "rewards/symbolic_reward_partial_score/std": 0.22525496780872345,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.043199896812439,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 2.1184022426605225,
      "step": 841
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2274625524878502,
      "epoch": 1.3493589743589745,
      "grad_norm": 0.033875465393066406,
      "learning_rate": 1e-06,
      "loss": 0.0174,
      "step": 842
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.2375042364001274,
      "epoch": 1.3509615384615383,
      "grad_norm": 0.02767942100763321,
      "learning_rate": 1e-06,
      "loss": 0.001,
      "step": 843
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2300783023238182,
      "epoch": 1.3525641025641026,
      "grad_norm": 0.01807348243892193,
      "learning_rate": 1e-06,
      "loss": 0.0017,
      "step": 844
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4712.0,
      "completions/mean_length": 1617.0,
      "completions/mean_terminated_length": 1559.09033203125,
      "completions/min_length": 433.0,
      "completions/min_terminated_length": 433.0,
      "entropy": 0.22667521983385086,
      "epoch": 1.3541666666666667,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.036902934312820435,
      "learning_rate": 1e-06,
      "loss": 0.0133,
      "num_tokens": 738841725.0,
      "reward": 0.2651147246360779,
      "reward_std": 0.10891143232584,
      "rewards/progression_diversity/mean": -0.00024866661988198757,
      "rewards/progression_diversity/std": 0.002786832395941019,
      "rewards/symbolic_reward_accuracy/mean": 0.126953125,
      "rewards/symbolic_reward_accuracy/std": 0.33324605226516724,
      "rewards/symbolic_reward_partial_score/mean": 0.6311197280883789,
      "rewards/symbolic_reward_partial_score/std": 0.21445629000663757,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0440601110458374,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.6725925803184509,
      "step": 845
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.22537444531917572,
      "epoch": 1.3557692307692308,
      "grad_norm": 0.02038760855793953,
      "learning_rate": 1e-06,
      "loss": 0.0638,
      "step": 846
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2280193567276001,
      "epoch": 1.357371794871795,
      "grad_norm": 0.024952072650194168,
      "learning_rate": 1e-06,
      "loss": -0.002,
      "step": 847
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.23109513521194458,
      "epoch": 1.358974358974359,
      "grad_norm": 0.020898301154375076,
      "learning_rate": 1e-06,
      "loss": 0.0146,
      "step": 848
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4054.0,
      "completions/max_terminated_length": 4054.0,
      "completions/mean_length": 1489.037109375,
      "completions/mean_terminated_length": 1489.037109375,
      "completions/min_length": 358.0,
      "completions/min_terminated_length": 358.0,
      "entropy": 0.2277429699897766,
      "epoch": 1.3605769230769231,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 0.028738927096128464,
      "learning_rate": 1e-06,
      "loss": 0.004,
      "num_tokens": 740499376.0,
      "reward": 0.2145245224237442,
      "reward_std": 0.07948393374681473,
      "rewards/progression_diversity/mean": -0.0001856325543485582,
      "rewards/progression_diversity/std": 0.002533870516344905,
      "rewards/symbolic_reward_accuracy/mean": 0.060546875,
      "rewards/symbolic_reward_accuracy/std": 0.2387305200099945,
      "rewards/symbolic_reward_partial_score/mean": 0.593994140625,
      "rewards/symbolic_reward_partial_score/std": 0.20968253910541534,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045418381690979,
      "sampling/importance_sampling_ratio/min": 6.543979334111549e-12,
      "sampling/sampling_logp_difference/max": 25.75247573852539,
      "sampling/sampling_logp_difference/mean": 0.09526436775922775,
      "step": 849
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.22307077050209045,
      "epoch": 1.3621794871794872,
      "grad_norm": 0.025526469573378563,
      "learning_rate": 1e-06,
      "loss": 0.0063,
      "step": 850
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.22553226351737976,
      "epoch": 1.3637820512820513,
      "grad_norm": 0.02305307425558567,
      "learning_rate": 1e-06,
      "loss": 0.0134,
      "step": 851
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.21545638144016266,
      "epoch": 1.3653846153846154,
      "grad_norm": 0.02928796596825123,
      "learning_rate": 1e-06,
      "loss": 0.0036,
      "step": 852
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3814.0,
      "completions/mean_length": 1567.021484375,
      "completions/mean_terminated_length": 1538.025390625,
      "completions/min_length": 455.0,
      "completions/min_terminated_length": 455.0,
      "entropy": 0.21113787591457367,
      "epoch": 1.3669871794871795,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.04044374078512192,
      "learning_rate": 1e-06,
      "loss": 0.0172,
      "num_tokens": 742260011.0,
      "reward": 0.17563673853874207,
      "reward_std": 0.04823887348175049,
      "rewards/progression_diversity/mean": -0.00029063018155284226,
      "rewards/progression_diversity/std": 0.00633732695132494,
      "rewards/symbolic_reward_accuracy/mean": 0.009765625,
      "rewards/symbolic_reward_accuracy/std": 0.09843364357948303,
      "rewards/symbolic_reward_partial_score/mean": 0.5659342408180237,
      "rewards/symbolic_reward_partial_score/std": 0.18250161409378052,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0408705472946167,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 0.9602516889572144,
      "step": 853
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.218083955347538,
      "epoch": 1.3685897435897436,
      "grad_norm": 0.021193062886595726,
      "learning_rate": 1e-06,
      "loss": -0.0058,
      "step": 854
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.20679837465286255,
      "epoch": 1.3701923076923077,
      "grad_norm": 1588.3087158203125,
      "learning_rate": 1e-06,
      "loss": 0.0694,
      "step": 855
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.20584507286548615,
      "epoch": 1.3717948717948718,
      "grad_norm": 0.02309674583375454,
      "learning_rate": 1e-06,
      "loss": -0.0003,
      "step": 856
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4411.0,
      "completions/max_terminated_length": 4411.0,
      "completions/mean_length": 1414.36328125,
      "completions/mean_terminated_length": 1414.36328125,
      "completions/min_length": 445.0,
      "completions/min_terminated_length": 445.0,
      "entropy": 0.21835298836231232,
      "epoch": 1.373397435897436,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 0.026670735329389572,
      "learning_rate": 1e-06,
      "loss": 0.0033,
      "num_tokens": 743893157.0,
      "reward": 0.23589275777339935,
      "reward_std": 0.11220350861549377,
      "rewards/progression_diversity/mean": -8.082183921942487e-05,
      "rewards/progression_diversity/std": 0.0013526254333555698,
      "rewards/symbolic_reward_accuracy/mean": 0.1015625,
      "rewards/symbolic_reward_accuracy/std": 0.30236753821372986,
      "rewards/symbolic_reward_partial_score/mean": 0.5831868648529053,
      "rewards/symbolic_reward_partial_score/std": 0.22691193222999573,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0431017875671387,
      "sampling/importance_sampling_ratio/min": 2.427903382340446e-05,
      "sampling/sampling_logp_difference/max": 10.625897407531738,
      "sampling/sampling_logp_difference/mean": 0.0924476757645607,
      "step": 857
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.22078397125005722,
      "epoch": 1.375,
      "grad_norm": 0.017304185777902603,
      "learning_rate": 1e-06,
      "loss": -0.0035,
      "step": 858
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.21548201143741608,
      "epoch": 1.376602564102564,
      "grad_norm": 0.017954610288143158,
      "learning_rate": 1e-06,
      "loss": 0.0101,
      "step": 859
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.21115773171186447,
      "epoch": 1.3782051282051282,
      "grad_norm": 0.02831229567527771,
      "learning_rate": 1e-06,
      "loss": -0.0007,
      "step": 860
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3687.0,
      "completions/mean_length": 1555.4765625,
      "completions/mean_terminated_length": 1526.4578857421875,
      "completions/min_length": 425.0,
      "completions/min_terminated_length": 425.0,
      "entropy": 0.2089003250002861,
      "epoch": 1.3798076923076923,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 104.46712493896484,
      "learning_rate": 1e-06,
      "loss": 0.0193,
      "num_tokens": 745612857.0,
      "reward": 0.18778178095817566,
      "reward_std": 0.06140269339084625,
      "rewards/progression_diversity/mean": -0.0006311247125267982,
      "rewards/progression_diversity/std": 0.005138562526553869,
      "rewards/symbolic_reward_accuracy/mean": 0.015625,
      "rewards/symbolic_reward_accuracy/std": 0.12414088100194931,
      "rewards/symbolic_reward_partial_score/mean": 0.5953613519668579,
      "rewards/symbolic_reward_partial_score/std": 0.1708049327135086,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0414655208587646,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 724.0,
      "sampling/sampling_logp_difference/mean": 0.20548060536384583,
      "step": 861
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2088572382926941,
      "epoch": 1.3814102564102564,
      "grad_norm": 0.025014329701662064,
      "learning_rate": 1e-06,
      "loss": 0.0155,
      "step": 862
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.21202120184898376,
      "epoch": 1.3830128205128205,
      "grad_norm": 0.028442082926630974,
      "learning_rate": 1e-06,
      "loss": 0.0071,
      "step": 863
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.21458961814641953,
      "epoch": 1.3846153846153846,
      "grad_norm": 0.02302766777575016,
      "learning_rate": 1e-06,
      "loss": 0.0094,
      "step": 864
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4167.0,
      "completions/mean_length": 1564.603515625,
      "completions/mean_terminated_length": 1477.2593994140625,
      "completions/min_length": 416.0,
      "completions/min_terminated_length": 416.0,
      "entropy": 0.2116372287273407,
      "epoch": 1.3862179487179487,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.032430339604616165,
      "learning_rate": 1e-06,
      "loss": 0.053,
      "num_tokens": 747279262.0,
      "reward": 0.2634183168411255,
      "reward_std": 0.09794985502958298,
      "rewards/progression_diversity/mean": -0.00045274931471794844,
      "rewards/progression_diversity/std": 0.008904572576284409,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.6300293207168579,
      "rewards/symbolic_reward_partial_score/std": 0.21683233976364136,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0390100479125977,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 1.8159689903259277,
      "step": 865
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.22469650208950043,
      "epoch": 1.3878205128205128,
      "grad_norm": 0.018467538058757782,
      "learning_rate": 1e-06,
      "loss": 0.0018,
      "step": 866
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.22149597853422165,
      "epoch": 1.3894230769230769,
      "grad_norm": 0.023728886619210243,
      "learning_rate": 1e-06,
      "loss": 0.0179,
      "step": 867
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.22125183045864105,
      "epoch": 1.391025641025641,
      "grad_norm": 0.019297420978546143,
      "learning_rate": 1e-06,
      "loss": 0.0045,
      "step": 868
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3547.0,
      "completions/mean_length": 1396.44921875,
      "completions/mean_terminated_length": 1367.119384765625,
      "completions/min_length": 462.0,
      "completions/min_terminated_length": 462.0,
      "entropy": 0.23219949007034302,
      "epoch": 1.392628205128205,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.035795051604509354,
      "learning_rate": 1e-06,
      "loss": 0.0184,
      "num_tokens": 748746020.0,
      "reward": 0.22064433991909027,
      "reward_std": 0.07329108566045761,
      "rewards/progression_diversity/mean": -0.0009962395997717977,
      "rewards/progression_diversity/std": 0.018096894025802612,
      "rewards/symbolic_reward_accuracy/mean": 0.052734375,
      "rewards/symbolic_reward_accuracy/std": 0.22372129559516907,
      "rewards/symbolic_reward_partial_score/mean": 0.6306965351104736,
      "rewards/symbolic_reward_partial_score/std": 0.2060777246952057,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0450772047042847,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 1.2045083045959473,
      "step": 869
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.23559828102588654,
      "epoch": 1.3942307692307692,
      "grad_norm": 0.01611843891441822,
      "learning_rate": 1e-06,
      "loss": 0.0068,
      "step": 870
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.23359952867031097,
      "epoch": 1.3958333333333333,
      "grad_norm": 0.02076306752860546,
      "learning_rate": 1e-06,
      "loss": -0.0021,
      "step": 871
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2402142956852913,
      "epoch": 1.3974358974358974,
      "grad_norm": 0.017290132120251656,
      "learning_rate": 1e-06,
      "loss": -0.0002,
      "step": 872
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3703.0,
      "completions/mean_length": 1506.962890625,
      "completions/mean_terminated_length": 1477.8492431640625,
      "completions/min_length": 442.0,
      "completions/min_terminated_length": 442.0,
      "entropy": 0.22817867249250412,
      "epoch": 1.3990384615384617,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 61.729652404785156,
      "learning_rate": 1e-06,
      "loss": 0.028,
      "num_tokens": 750412865.0,
      "reward": 0.23350869119167328,
      "reward_std": 0.08862544596195221,
      "rewards/progression_diversity/mean": -0.00020658165158238262,
      "rewards/progression_diversity/std": 0.0025547959376126528,
      "rewards/symbolic_reward_accuracy/mean": 0.08984375,
      "rewards/symbolic_reward_accuracy/std": 0.2862374484539032,
      "rewards/symbolic_reward_partial_score/mean": 0.5993326902389526,
      "rewards/symbolic_reward_partial_score/std": 0.23785848915576935,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045426368713379,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 712.0,
      "sampling/sampling_logp_difference/mean": 0.20375153422355652,
      "step": 873
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.22932185232639313,
      "epoch": 1.4006410256410255,
      "grad_norm": 0.02397080697119236,
      "learning_rate": 1e-06,
      "loss": 0.0007,
      "step": 874
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.229917511343956,
      "epoch": 1.4022435897435899,
      "grad_norm": 0.028196433559060097,
      "learning_rate": 1e-06,
      "loss": 0.0031,
      "step": 875
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2285906821489334,
      "epoch": 1.4038461538461537,
      "grad_norm": 0.021019669249653816,
      "learning_rate": 1e-06,
      "loss": -0.0085,
      "step": 876
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4288.0,
      "completions/max_terminated_length": 4288.0,
      "completions/mean_length": 1584.255859375,
      "completions/mean_terminated_length": 1584.255859375,
      "completions/min_length": 449.0,
      "completions/min_terminated_length": 449.0,
      "entropy": 0.22845402359962463,
      "epoch": 1.405448717948718,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.1292523443698883,
      "learning_rate": 1e-06,
      "loss": 0.0021,
      "num_tokens": 752032676.0,
      "reward": 0.21584956347942352,
      "reward_std": 0.0788746029138565,
      "rewards/progression_diversity/mean": -0.0004929413553327322,
      "rewards/progression_diversity/std": 0.0036183472257107496,
      "rewards/symbolic_reward_accuracy/mean": 0.0625,
      "rewards/symbolic_reward_accuracy/std": 0.2422981858253479,
      "rewards/symbolic_reward_partial_score/mean": 0.5945149660110474,
      "rewards/symbolic_reward_partial_score/std": 0.20550201833248138,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0462896823883057,
      "sampling/importance_sampling_ratio/min": 0.0012942898320034146,
      "sampling/sampling_logp_difference/max": 6.6497931480407715,
      "sampling/sampling_logp_difference/mean": 0.09629081189632416,
      "step": 877
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2279265597462654,
      "epoch": 1.407051282051282,
      "grad_norm": 0.019477983936667442,
      "learning_rate": 1e-06,
      "loss": -0.0207,
      "step": 878
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.22508171200752258,
      "epoch": 1.4086538461538463,
      "grad_norm": 0.029449041932821274,
      "learning_rate": 1e-06,
      "loss": 0.0071,
      "step": 879
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.2288508340716362,
      "epoch": 1.4102564102564101,
      "grad_norm": 0.022348174825310707,
      "learning_rate": 1e-06,
      "loss": 0.0093,
      "step": 880
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4524.0,
      "completions/mean_length": 1608.572265625,
      "completions/mean_terminated_length": 1579.657470703125,
      "completions/min_length": 518.0,
      "completions/min_terminated_length": 518.0,
      "entropy": 0.22652538120746613,
      "epoch": 1.4118589743589745,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 0.0312635712325573,
      "learning_rate": 1e-06,
      "loss": 0.018,
      "num_tokens": 753635801.0,
      "reward": 0.23253975808620453,
      "reward_std": 0.08958201110363007,
      "rewards/progression_diversity/mean": -0.0004195784858893603,
      "rewards/progression_diversity/std": 0.008441566489636898,
      "rewards/symbolic_reward_accuracy/mean": 0.08984375,
      "rewards/symbolic_reward_accuracy/std": 0.2862374484539032,
      "rewards/symbolic_reward_partial_score/mean": 0.5961099863052368,
      "rewards/symbolic_reward_partial_score/std": 0.23455365002155304,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0431697368621826,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 1.380308985710144,
      "step": 881
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.22486486285924911,
      "epoch": 1.4134615384615383,
      "grad_norm": 0.023399055004119873,
      "learning_rate": 1e-06,
      "loss": 0.0023,
      "step": 882
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.22499487549066544,
      "epoch": 1.4150641025641026,
      "grad_norm": 386.1463317871094,
      "learning_rate": 1e-06,
      "loss": 0.0152,
      "step": 883
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.23033517599105835,
      "epoch": 1.4166666666666667,
      "grad_norm": 0.02022228017449379,
      "learning_rate": 1e-06,
      "loss": 0.0148,
      "step": 884
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4005.0,
      "completions/mean_length": 1440.396484375,
      "completions/mean_terminated_length": 1411.152587890625,
      "completions/min_length": 492.0,
      "completions/min_terminated_length": 492.0,
      "entropy": 0.2310120165348053,
      "epoch": 1.4182692307692308,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 0.029161928221583366,
      "learning_rate": 1e-06,
      "loss": 0.0218,
      "num_tokens": 755244068.0,
      "reward": 0.27608346939086914,
      "reward_std": 0.10152895748615265,
      "rewards/progression_diversity/mean": -0.0005383668467402458,
      "rewards/progression_diversity/std": 0.006143567617982626,
      "rewards/symbolic_reward_accuracy/mean": 0.1484375,
      "rewards/symbolic_reward_accuracy/std": 0.35588082671165466,
      "rewards/symbolic_reward_partial_score/mean": 0.624072253704071,
      "rewards/symbolic_reward_partial_score/std": 0.22695200145244598,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0465407371520996,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 700.0,
      "sampling/sampling_logp_difference/mean": 0.22680458426475525,
      "step": 885
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.23101823031902313,
      "epoch": 1.419871794871795,
      "grad_norm": 0.01696612685918808,
      "learning_rate": 1e-06,
      "loss": 0.008,
      "step": 886
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.22741740196943283,
      "epoch": 1.421474358974359,
      "grad_norm": 0.03519902750849724,
      "learning_rate": 1e-06,
      "loss": -0.0015,
      "step": 887
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.22920458763837814,
      "epoch": 1.4230769230769231,
      "grad_norm": 0.01825595274567604,
      "learning_rate": 1e-06,
      "loss": -0.0034,
      "step": 888
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3919.0,
      "completions/max_terminated_length": 3919.0,
      "completions/mean_length": 1301.556640625,
      "completions/mean_terminated_length": 1301.556640625,
      "completions/min_length": 459.0,
      "completions/min_terminated_length": 459.0,
      "entropy": 0.22784828394651413,
      "epoch": 1.4246794871794872,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 0.03417736664414406,
      "learning_rate": 1e-06,
      "loss": 0.0058,
      "num_tokens": 756822881.0,
      "reward": 0.27989476919174194,
      "reward_std": 0.10528245568275452,
      "rewards/progression_diversity/mean": -0.0002708068350329995,
      "rewards/progression_diversity/std": 0.0028580420184880495,
      "rewards/symbolic_reward_accuracy/mean": 0.13671875,
      "rewards/symbolic_reward_accuracy/std": 0.3438861668109894,
      "rewards/symbolic_reward_partial_score/mean": 0.6595540046691895,
      "rewards/symbolic_reward_partial_score/std": 0.209646537899971,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0472159385681152,
      "sampling/importance_sampling_ratio/min": 9.738879924725552e-09,
      "sampling/sampling_logp_difference/max": 18.447139739990234,
      "sampling/sampling_logp_difference/mean": 0.09813030064105988,
      "step": 889
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.23206569254398346,
      "epoch": 1.4262820512820513,
      "grad_norm": 0.02267414890229702,
      "learning_rate": 1e-06,
      "loss": 0.0077,
      "step": 890
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.23299744725227356,
      "epoch": 1.4278846153846154,
      "grad_norm": 0.02226838655769825,
      "learning_rate": 1e-06,
      "loss": -0.0129,
      "step": 891
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.23174694180488586,
      "epoch": 1.4294871794871795,
      "grad_norm": 0.02325165644288063,
      "learning_rate": 1e-06,
      "loss": 0.0171,
      "step": 892
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3854.0,
      "completions/max_terminated_length": 3854.0,
      "completions/mean_length": 1381.82421875,
      "completions/mean_terminated_length": 1381.82421875,
      "completions/min_length": 453.0,
      "completions/min_terminated_length": 453.0,
      "entropy": 0.23948068916797638,
      "epoch": 1.4310897435897436,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.02791544795036316,
      "learning_rate": 1e-06,
      "loss": -0.0007,
      "num_tokens": 758422935.0,
      "reward": 0.23517410457134247,
      "reward_std": 0.05936092138290405,
      "rewards/progression_diversity/mean": -0.00016881769988685846,
      "rewards/progression_diversity/std": 0.0023643465247005224,
      "rewards/symbolic_reward_accuracy/mean": 0.08984375,
      "rewards/symbolic_reward_accuracy/std": 0.2862374484539032,
      "rewards/symbolic_reward_partial_score/mean": 0.6042318344116211,
      "rewards/symbolic_reward_partial_score/std": 0.19526702165603638,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0488243103027344,
      "sampling/importance_sampling_ratio/min": 0.00036482111318036914,
      "sampling/sampling_logp_difference/max": 7.916103363037109,
      "sampling/sampling_logp_difference/mean": 0.10077446699142456,
      "step": 893
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2307603806257248,
      "epoch": 1.4326923076923077,
      "grad_norm": 0.029786186292767525,
      "learning_rate": 1e-06,
      "loss": 0.0158,
      "step": 894
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2346210777759552,
      "epoch": 1.4342948717948718,
      "grad_norm": 0.030796803534030914,
      "learning_rate": 1e-06,
      "loss": -0.0004,
      "step": 895
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.24034950882196426,
      "epoch": 1.435897435897436,
      "grad_norm": 0.01626007631421089,
      "learning_rate": 1e-06,
      "loss": -0.0021,
      "step": 896
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4579.0,
      "completions/max_terminated_length": 4579.0,
      "completions/mean_length": 1346.220703125,
      "completions/mean_terminated_length": 1346.220703125,
      "completions/min_length": 402.0,
      "completions/min_terminated_length": 402.0,
      "entropy": 0.2430427074432373,
      "epoch": 1.4375,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 0.030239736661314964,
      "learning_rate": 1e-06,
      "loss": 0.0065,
      "num_tokens": 759944376.0,
      "reward": 0.2594153881072998,
      "reward_std": 0.09177584946155548,
      "rewards/progression_diversity/mean": -0.0003554845170583576,
      "rewards/progression_diversity/std": 0.004576821345835924,
      "rewards/symbolic_reward_accuracy/mean": 0.119140625,
      "rewards/symbolic_reward_accuracy/std": 0.32427072525024414,
      "rewards/symbolic_reward_partial_score/mean": 0.6264485120773315,
      "rewards/symbolic_reward_partial_score/std": 0.23991042375564575,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0497946739196777,
      "sampling/importance_sampling_ratio/min": 0.003207254223525524,
      "sampling/sampling_logp_difference/max": 5.742340087890625,
      "sampling/sampling_logp_difference/mean": 0.1022864505648613,
      "step": 897
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.24024366587400436,
      "epoch": 1.439102564102564,
      "grad_norm": 0.02961251139640808,
      "learning_rate": 1e-06,
      "loss": 0.0027,
      "step": 898
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.24130822718143463,
      "epoch": 1.4407051282051282,
      "grad_norm": 0.02147165685892105,
      "learning_rate": 1e-06,
      "loss": 0.0061,
      "step": 899
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.3203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2432633340358734,
      "epoch": 1.4423076923076923,
      "grad_norm": 0.027493350207805634,
      "learning_rate": 1e-06,
      "loss": 0.0033,
      "step": 900
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4780.0,
      "completions/max_terminated_length": 4780.0,
      "completions/mean_length": 1268.705078125,
      "completions/mean_terminated_length": 1268.705078125,
      "completions/min_length": 417.0,
      "completions/min_terminated_length": 417.0,
      "entropy": 0.24526084959506989,
      "epoch": 1.4439102564102564,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.02393464371562004,
      "learning_rate": 1e-06,
      "loss": 0.0016,
      "num_tokens": 761386065.0,
      "reward": 0.23188182711601257,
      "reward_std": 0.0820198506116867,
      "rewards/progression_diversity/mean": -0.0002951324568130076,
      "rewards/progression_diversity/std": 0.0035472013987600803,
      "rewards/symbolic_reward_accuracy/mean": 0.05078125,
      "rewards/symbolic_reward_accuracy/std": 0.21976542472839355,
      "rewards/symbolic_reward_partial_score/mean": 0.67138671875,
      "rewards/symbolic_reward_partial_score/std": 0.20371752977371216,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0510923862457275,
      "sampling/importance_sampling_ratio/min": 4.374550189822912e-05,
      "sampling/sampling_logp_difference/max": 10.037121772766113,
      "sampling/sampling_logp_difference/mean": 0.103520967066288,
      "step": 901
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.24283086508512497,
      "epoch": 1.4455128205128205,
      "grad_norm": 0.02904311753809452,
      "learning_rate": 1e-06,
      "loss": 0.0,
      "step": 902
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.24352674186229706,
      "epoch": 1.4471153846153846,
      "grad_norm": 0.01613425463438034,
      "learning_rate": 1e-06,
      "loss": 0.0014,
      "step": 903
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.25162430852651596,
      "epoch": 1.4487179487179487,
      "grad_norm": 0.02424633502960205,
      "learning_rate": 1e-06,
      "loss": 0.0034,
      "step": 904
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4364.0,
      "completions/max_terminated_length": 4364.0,
      "completions/mean_length": 1284.42578125,
      "completions/mean_terminated_length": 1284.42578125,
      "completions/min_length": 519.0,
      "completions/min_terminated_length": 519.0,
      "entropy": 0.24817583709955215,
      "epoch": 1.4503205128205128,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 0.025694923475384712,
      "learning_rate": 1e-06,
      "loss": 0.0016,
      "num_tokens": 762867995.0,
      "reward": 0.30710139870643616,
      "reward_std": 0.05024458467960358,
      "rewards/progression_diversity/mean": -0.0003102348418906331,
      "rewards/progression_diversity/std": 0.003935862332582474,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.656494140625,
      "rewards/symbolic_reward_partial_score/std": 0.2331247478723526,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0509289503097534,
      "sampling/importance_sampling_ratio/min": 0.002624225802719593,
      "sampling/sampling_logp_difference/max": 5.94296932220459,
      "sampling/sampling_logp_difference/mean": 0.10329769551753998,
      "step": 905
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.24394653737545013,
      "epoch": 1.4519230769230769,
      "grad_norm": 0.020293056964874268,
      "learning_rate": 1e-06,
      "loss": -0.0066,
      "step": 906
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.24355324357748032,
      "epoch": 1.453525641025641,
      "grad_norm": 0.01819128543138504,
      "learning_rate": 1e-06,
      "loss": 0.0068,
      "step": 907
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.241835318505764,
      "epoch": 1.455128205128205,
      "grad_norm": 0.01816965639591217,
      "learning_rate": 1e-06,
      "loss": -0.0012,
      "step": 908
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3849.0,
      "completions/max_terminated_length": 3849.0,
      "completions/mean_length": 1251.197265625,
      "completions/mean_terminated_length": 1251.197265625,
      "completions/min_length": 485.0,
      "completions/min_terminated_length": 485.0,
      "entropy": 0.24188891798257828,
      "epoch": 1.4567307692307692,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 0.02820868045091629,
      "learning_rate": 1e-06,
      "loss": -0.0056,
      "num_tokens": 764313808.0,
      "reward": 0.31225496530532837,
      "reward_std": 0.08842569589614868,
      "rewards/progression_diversity/mean": -9.080560994334519e-05,
      "rewards/progression_diversity/std": 0.001025109551846981,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.6658528447151184,
      "rewards/symbolic_reward_partial_score/std": 0.24876438081264496,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0511972904205322,
      "sampling/importance_sampling_ratio/min": 0.0021847328171133995,
      "sampling/sampling_logp_difference/max": 6.1262617111206055,
      "sampling/sampling_logp_difference/mean": 0.1035478338599205,
      "step": 909
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.24276289343833923,
      "epoch": 1.4583333333333333,
      "grad_norm": 0.023914597928524017,
      "learning_rate": 1e-06,
      "loss": -0.0036,
      "step": 910
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2428332194685936,
      "epoch": 1.4599358974358974,
      "grad_norm": 0.018325267359614372,
      "learning_rate": 1e-06,
      "loss": 0.0122,
      "step": 911
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.24748840928077698,
      "epoch": 1.4615384615384617,
      "grad_norm": 0.021704666316509247,
      "learning_rate": 1e-06,
      "loss": -0.0035,
      "step": 912
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3406.0,
      "completions/max_terminated_length": 3406.0,
      "completions/mean_length": 1280.796875,
      "completions/mean_terminated_length": 1280.796875,
      "completions/min_length": 428.0,
      "completions/min_terminated_length": 428.0,
      "entropy": 0.24230319261550903,
      "epoch": 1.4631410256410255,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 0.026809457689523697,
      "learning_rate": 1e-06,
      "loss": -0.0139,
      "num_tokens": 765825992.0,
      "reward": 0.29785943031311035,
      "reward_std": 0.1095178872346878,
      "rewards/progression_diversity/mean": -0.00018848836771212518,
      "rewards/progression_diversity/std": 0.0023818761110305786,
      "rewards/symbolic_reward_accuracy/mean": 0.169921875,
      "rewards/symbolic_reward_accuracy/std": 0.3759314715862274,
      "rewards/symbolic_reward_partial_score/mean": 0.653027355670929,
      "rewards/symbolic_reward_partial_score/std": 0.24102628231048584,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0500684976577759,
      "sampling/importance_sampling_ratio/min": 0.00195193977560848,
      "sampling/sampling_logp_difference/max": 6.238931655883789,
      "sampling/sampling_logp_difference/mean": 0.10224372148513794,
      "step": 913
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.24049583077430725,
      "epoch": 1.4647435897435899,
      "grad_norm": 0.019987255334854126,
      "learning_rate": 1e-06,
      "loss": 0.0159,
      "step": 914
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.23697714507579803,
      "epoch": 1.4663461538461537,
      "grad_norm": 0.014981388114392757,
      "learning_rate": 1e-06,
      "loss": 0.0016,
      "step": 915
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.23524734377861023,
      "epoch": 1.467948717948718,
      "grad_norm": 0.019518760964274406,
      "learning_rate": 1e-06,
      "loss": 0.0036,
      "step": 916
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3382.0,
      "completions/max_terminated_length": 3382.0,
      "completions/mean_length": 1324.177734375,
      "completions/mean_terminated_length": 1324.177734375,
      "completions/min_length": 383.0,
      "completions/min_terminated_length": 383.0,
      "entropy": 0.23860549181699753,
      "epoch": 1.469551282051282,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 0.034368306398391724,
      "learning_rate": 1e-06,
      "loss": 0.0,
      "num_tokens": 767335715.0,
      "reward": 0.2800864577293396,
      "reward_std": 0.11742740869522095,
      "rewards/progression_diversity/mean": -0.0001446699898224324,
      "rewards/progression_diversity/std": 0.0014527636812999845,
      "rewards/symbolic_reward_accuracy/mean": 0.146484375,
      "rewards/symbolic_reward_accuracy/std": 0.35393697023391724,
      "rewards/symbolic_reward_partial_score/mean": 0.6406575441360474,
      "rewards/symbolic_reward_partial_score/std": 0.218194380402565,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0496015548706055,
      "sampling/importance_sampling_ratio/min": 2.004770264726896e-12,
      "sampling/sampling_logp_difference/max": 26.93549156188965,
      "sampling/sampling_logp_difference/mean": 0.10103441774845123,
      "step": 917
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2368185818195343,
      "epoch": 1.4711538461538463,
      "grad_norm": 0.018526000902056694,
      "learning_rate": 1e-06,
      "loss": -0.0033,
      "step": 918
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.24254272133111954,
      "epoch": 1.4727564102564101,
      "grad_norm": 0.021521111950278282,
      "learning_rate": 1e-06,
      "loss": 0.0077,
      "step": 919
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.24013221263885498,
      "epoch": 1.4743589743589745,
      "grad_norm": 0.019323555752635002,
      "learning_rate": 1e-06,
      "loss": 0.0033,
      "step": 920
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3175.0,
      "completions/max_terminated_length": 3175.0,
      "completions/mean_length": 1178.45703125,
      "completions/mean_terminated_length": 1178.45703125,
      "completions/min_length": 421.0,
      "completions/min_terminated_length": 421.0,
      "entropy": 0.24065189808607101,
      "epoch": 1.4759615384615383,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 0.027268243953585625,
      "learning_rate": 1e-06,
      "loss": 0.0063,
      "num_tokens": 768849229.0,
      "reward": 0.2435724139213562,
      "reward_std": 0.0985773503780365,
      "rewards/progression_diversity/mean": -0.00018066662596538663,
      "rewards/progression_diversity/std": 0.0020558543037623167,
      "rewards/symbolic_reward_accuracy/mean": 0.103515625,
      "rewards/symbolic_reward_accuracy/std": 0.30492907762527466,
      "rewards/symbolic_reward_partial_score/mean": 0.6048828363418579,
      "rewards/symbolic_reward_partial_score/std": 0.21721047163009644,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0502285957336426,
      "sampling/importance_sampling_ratio/min": 0.0018710603471845388,
      "sampling/sampling_logp_difference/max": 6.28125,
      "sampling/sampling_logp_difference/mean": 0.1014840379357338,
      "step": 921
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2395830601453781,
      "epoch": 1.4775641025641026,
      "grad_norm": 0.017518319189548492,
      "learning_rate": 1e-06,
      "loss": 0.0038,
      "step": 922
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.23873654007911682,
      "epoch": 1.4791666666666667,
      "grad_norm": 0.015559080988168716,
      "learning_rate": 1e-06,
      "loss": 0.01,
      "step": 923
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2355535551905632,
      "epoch": 1.4807692307692308,
      "grad_norm": 0.019503874704241753,
      "learning_rate": 1e-06,
      "loss": 0.003,
      "step": 924
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3569.0,
      "completions/max_terminated_length": 3569.0,
      "completions/mean_length": 1200.4765625,
      "completions/mean_terminated_length": 1200.4765625,
      "completions/min_length": 435.0,
      "completions/min_terminated_length": 435.0,
      "entropy": 0.2354103922843933,
      "epoch": 1.482371794871795,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 0.025359977036714554,
      "learning_rate": 1e-06,
      "loss": 0.0109,
      "num_tokens": 770377745.0,
      "reward": 0.29045236110687256,
      "reward_std": 0.059886179864406586,
      "rewards/progression_diversity/mean": -0.00017451572057325393,
      "rewards/progression_diversity/std": 0.002117466414347291,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.6556802988052368,
      "rewards/symbolic_reward_partial_score/std": 0.22488947212696075,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0493230819702148,
      "sampling/importance_sampling_ratio/min": 0.0001748588983900845,
      "sampling/sampling_logp_difference/max": 8.651531219482422,
      "sampling/sampling_logp_difference/mean": 0.10150042176246643,
      "step": 925
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.23772598803043365,
      "epoch": 1.483974358974359,
      "grad_norm": 0.019928786903619766,
      "learning_rate": 1e-06,
      "loss": 0.0116,
      "step": 926
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.23272281885147095,
      "epoch": 1.4855769230769231,
      "grad_norm": 0.020350804552435875,
      "learning_rate": 1e-06,
      "loss": 0.0017,
      "step": 927
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2364245057106018,
      "epoch": 1.4871794871794872,
      "grad_norm": 0.023986948654055595,
      "learning_rate": 1e-06,
      "loss": -0.0081,
      "step": 928
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4100.0,
      "completions/mean_length": 1294.845703125,
      "completions/mean_terminated_length": 1235.672607421875,
      "completions/min_length": 356.0,
      "completions/min_terminated_length": 356.0,
      "entropy": 0.23016077280044556,
      "epoch": 1.4887820512820513,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 171.7614288330078,
      "learning_rate": 1e-06,
      "loss": 0.0327,
      "num_tokens": 771956514.0,
      "reward": 0.22332197427749634,
      "reward_std": 0.07956110686063766,
      "rewards/progression_diversity/mean": -0.0012982608750462532,
      "rewards/progression_diversity/std": 0.026022804901003838,
      "rewards/symbolic_reward_accuracy/mean": 0.05859375,
      "rewards/symbolic_reward_accuracy/std": 0.23509246110916138,
      "rewards/symbolic_reward_partial_score/mean": 0.6285644769668579,
      "rewards/symbolic_reward_partial_score/std": 0.20716184377670288,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0451966524124146,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 1.1087749004364014,
      "step": 929
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2362649217247963,
      "epoch": 1.4903846153846154,
      "grad_norm": 0.02201710268855095,
      "learning_rate": 1e-06,
      "loss": -0.0087,
      "step": 930
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.22993260622024536,
      "epoch": 1.4919871794871795,
      "grad_norm": 0.020292161032557487,
      "learning_rate": 1e-06,
      "loss": 0.0475,
      "step": 931
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.230661079287529,
      "epoch": 1.4935897435897436,
      "grad_norm": 0.021966397762298584,
      "learning_rate": 1e-06,
      "loss": -0.0086,
      "step": 932
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3575.0,
      "completions/max_terminated_length": 3575.0,
      "completions/mean_length": 1369.162109375,
      "completions/mean_terminated_length": 1369.162109375,
      "completions/min_length": 501.0,
      "completions/min_terminated_length": 501.0,
      "entropy": 0.23234926909208298,
      "epoch": 1.4951923076923077,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 0.03726167976856232,
      "learning_rate": 1e-06,
      "loss": 0.0064,
      "num_tokens": 773450389.0,
      "reward": 0.29813235998153687,
      "reward_std": 0.12365701794624329,
      "rewards/progression_diversity/mean": -0.0002430165041005239,
      "rewards/progression_diversity/std": 0.00278676301240921,
      "rewards/symbolic_reward_accuracy/mean": 0.173828125,
      "rewards/symbolic_reward_accuracy/std": 0.3793322443962097,
      "rewards/symbolic_reward_partial_score/mean": 0.6461262702941895,
      "rewards/symbolic_reward_partial_score/std": 0.24039986729621887,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0482527017593384,
      "sampling/importance_sampling_ratio/min": 5.24304415801171e-15,
      "sampling/sampling_logp_difference/max": 32.881874084472656,
      "sampling/sampling_logp_difference/mean": 0.10041865706443787,
      "step": 933
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.23114058375358582,
      "epoch": 1.4967948717948718,
      "grad_norm": 0.02173478528857231,
      "learning_rate": 1e-06,
      "loss": -0.0046,
      "step": 934
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2359921634197235,
      "epoch": 1.498397435897436,
      "grad_norm": 0.01734345592558384,
      "learning_rate": 1e-06,
      "loss": 0.009,
      "step": 935
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.296875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2320852428674698,
      "epoch": 1.5,
      "grad_norm": 0.025080712512135506,
      "learning_rate": 1e-06,
      "loss": 0.0043,
      "step": 936
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3237.0,
      "completions/max_terminated_length": 3237.0,
      "completions/mean_length": 1222.326171875,
      "completions/mean_terminated_length": 1222.326171875,
      "completions/min_length": 378.0,
      "completions/min_terminated_length": 378.0,
      "entropy": 0.23251069337129593,
      "epoch": 1.501602564102564,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 0.03220761939883232,
      "learning_rate": 1e-06,
      "loss": -0.0103,
      "num_tokens": 774980300.0,
      "reward": 0.26821303367614746,
      "reward_std": 0.10178233683109283,
      "rewards/progression_diversity/mean": -0.00047397619346156716,
      "rewards/progression_diversity/std": 0.004046800546348095,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.6284343004226685,
      "rewards/symbolic_reward_partial_score/std": 0.20864459872245789,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0485594272613525,
      "sampling/importance_sampling_ratio/min": 0.0010224799625575542,
      "sampling/sampling_logp_difference/max": 6.885524272918701,
      "sampling/sampling_logp_difference/mean": 0.10064580291509628,
      "step": 937
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.23433350771665573,
      "epoch": 1.5032051282051282,
      "grad_norm": 0.027498142793774605,
      "learning_rate": 1e-06,
      "loss": -0.0059,
      "step": 938
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23485945910215378,
      "epoch": 1.5048076923076923,
      "grad_norm": 0.022965963929891586,
      "learning_rate": 1e-06,
      "loss": 0.0029,
      "step": 939
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2323073446750641,
      "epoch": 1.5064102564102564,
      "grad_norm": 0.017642127349972725,
      "learning_rate": 1e-06,
      "loss": 0.011,
      "step": 940
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3077.0,
      "completions/max_terminated_length": 3077.0,
      "completions/mean_length": 1306.53515625,
      "completions/mean_terminated_length": 1306.53515625,
      "completions/min_length": 482.0,
      "completions/min_terminated_length": 482.0,
      "entropy": 0.23624175786972046,
      "epoch": 1.5080128205128205,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.032414428889751434,
      "learning_rate": 1e-06,
      "loss": -0.0088,
      "num_tokens": 776444462.0,
      "reward": 0.35686036944389343,
      "reward_std": 0.1280536651611328,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.240234375,
      "rewards/symbolic_reward_accuracy/std": 0.4276435375213623,
      "rewards/symbolic_reward_partial_score/mean": 0.7090657949447632,
      "rewards/symbolic_reward_partial_score/std": 0.2242516130208969,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0489134788513184,
      "sampling/importance_sampling_ratio/min": 0.0029864604584872723,
      "sampling/sampling_logp_difference/max": 5.813666343688965,
      "sampling/sampling_logp_difference/mean": 0.10092984884977341,
      "step": 941
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2372562438249588,
      "epoch": 1.5096153846153846,
      "grad_norm": 0.0257510244846344,
      "learning_rate": 1e-06,
      "loss": 0.014,
      "step": 942
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.23343972861766815,
      "epoch": 1.5112179487179487,
      "grad_norm": 0.018904410302639008,
      "learning_rate": 1e-06,
      "loss": 0.0061,
      "step": 943
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.23894614726305008,
      "epoch": 1.5128205128205128,
      "grad_norm": 0.023683864623308182,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "step": 944
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3415.0,
      "completions/mean_length": 1307.0390625,
      "completions/mean_terminated_length": 1277.5341796875,
      "completions/min_length": 383.0,
      "completions/min_terminated_length": 383.0,
      "entropy": 0.23835492879152298,
      "epoch": 1.5144230769230769,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.02669195644557476,
      "learning_rate": 1e-06,
      "loss": -0.0048,
      "num_tokens": 777998866.0,
      "reward": 0.27208733558654785,
      "reward_std": 0.10921524465084076,
      "rewards/progression_diversity/mean": -0.0007387942750938237,
      "rewards/progression_diversity/std": 0.014222451485693455,
      "rewards/symbolic_reward_accuracy/mean": 0.13671875,
      "rewards/symbolic_reward_accuracy/std": 0.3438861668109894,
      "rewards/symbolic_reward_partial_score/mean": 0.6341959834098816,
      "rewards/symbolic_reward_partial_score/std": 0.22311989963054657,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0479480028152466,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.576566219329834,
      "step": 945
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.23516834527254105,
      "epoch": 1.516025641025641,
      "grad_norm": 0.02226245403289795,
      "learning_rate": 1e-06,
      "loss": 0.017,
      "step": 946
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.23766741156578064,
      "epoch": 1.5176282051282053,
      "grad_norm": 0.018796978518366814,
      "learning_rate": 1e-06,
      "loss": 0.0079,
      "step": 947
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.24504876136779785,
      "epoch": 1.5192307692307692,
      "grad_norm": 0.019867926836013794,
      "learning_rate": 1e-06,
      "loss": 0.0013,
      "step": 948
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3256.0,
      "completions/max_terminated_length": 3256.0,
      "completions/mean_length": 1339.611328125,
      "completions/mean_terminated_length": 1339.611328125,
      "completions/min_length": 402.0,
      "completions/min_terminated_length": 402.0,
      "entropy": 0.24304132908582687,
      "epoch": 1.5208333333333335,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 0.0299557913094759,
      "learning_rate": 1e-06,
      "loss": 0.0,
      "num_tokens": 779544475.0,
      "reward": 0.23521313071250916,
      "reward_std": 0.07730136811733246,
      "rewards/progression_diversity/mean": -0.00017136070528067648,
      "rewards/progression_diversity/std": 0.0020358404144644737,
      "rewards/symbolic_reward_accuracy/mean": 0.07421875,
      "rewards/symbolic_reward_accuracy/std": 0.2623828947544098,
      "rewards/symbolic_reward_partial_score/mean": 0.6356120109558105,
      "rewards/symbolic_reward_partial_score/std": 0.2210559844970703,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0506515502929688,
      "sampling/importance_sampling_ratio/min": 7.503909728256986e-05,
      "sampling/sampling_logp_difference/max": 9.497501373291016,
      "sampling/sampling_logp_difference/mean": 0.10373318195343018,
      "step": 949
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.24009553343057632,
      "epoch": 1.5224358974358974,
      "grad_norm": 0.01935042440891266,
      "learning_rate": 1e-06,
      "loss": -0.0066,
      "step": 950
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.2450866997241974,
      "epoch": 1.5240384615384617,
      "grad_norm": 0.02386186644434929,
      "learning_rate": 1e-06,
      "loss": 0.0068,
      "step": 951
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.24486373364925385,
      "epoch": 1.5256410256410255,
      "grad_norm": 0.026845736429095268,
      "learning_rate": 1e-06,
      "loss": -0.0011,
      "step": 952
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3384.0,
      "completions/mean_length": 1382.8046875,
      "completions/mean_terminated_length": 1353.4481201171875,
      "completions/min_length": 478.0,
      "completions/min_terminated_length": 478.0,
      "entropy": 0.24295812845230103,
      "epoch": 1.5272435897435899,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 0.022905530408024788,
      "learning_rate": 1e-06,
      "loss": -0.0003,
      "num_tokens": 781073095.0,
      "reward": 0.2498174011707306,
      "reward_std": 0.08474698662757874,
      "rewards/progression_diversity/mean": -0.00019435372087173164,
      "rewards/progression_diversity/std": 0.0025657941587269306,
      "rewards/symbolic_reward_accuracy/mean": 0.10546875,
      "rewards/symbolic_reward_accuracy/std": 0.3074568510055542,
      "rewards/symbolic_reward_partial_score/mean": 0.6224446296691895,
      "rewards/symbolic_reward_partial_score/std": 0.231918066740036,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.049942970275879,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.46404212713241577,
      "step": 953
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.24674228578805923,
      "epoch": 1.5288461538461537,
      "grad_norm": 0.01813218928873539,
      "learning_rate": 1e-06,
      "loss": -0.0004,
      "step": 954
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2422228530049324,
      "epoch": 1.530448717948718,
      "grad_norm": 1672.424560546875,
      "learning_rate": 1e-06,
      "loss": 0.0588,
      "step": 955
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2425452098250389,
      "epoch": 1.532051282051282,
      "grad_norm": 0.017666012048721313,
      "learning_rate": 1e-06,
      "loss": -0.0026,
      "step": 956
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3996.0,
      "completions/max_terminated_length": 3996.0,
      "completions/mean_length": 1283.716796875,
      "completions/mean_terminated_length": 1283.716796875,
      "completions/min_length": 427.0,
      "completions/min_terminated_length": 427.0,
      "entropy": 0.2416147142648697,
      "epoch": 1.5336538461538463,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 0.022795159369707108,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "num_tokens": 782631542.0,
      "reward": 0.29501813650131226,
      "reward_std": 0.07359684258699417,
      "rewards/progression_diversity/mean": -0.00013884622603654861,
      "rewards/progression_diversity/std": 0.0017234663246199489,
      "rewards/symbolic_reward_accuracy/mean": 0.154296875,
      "rewards/symbolic_reward_accuracy/std": 0.36158639192581177,
      "rewards/symbolic_reward_partial_score/mean": 0.6748046875,
      "rewards/symbolic_reward_partial_score/std": 0.21226514875888824,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0523028373718262,
      "sampling/importance_sampling_ratio/min": 1.8659156353351136e-07,
      "sampling/sampling_logp_difference/max": 15.494343757629395,
      "sampling/sampling_logp_difference/mean": 0.10692096501588821,
      "step": 957
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.24522508680820465,
      "epoch": 1.5352564102564101,
      "grad_norm": 0.01961701549589634,
      "learning_rate": 1e-06,
      "loss": 0.004,
      "step": 958
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.25065000355243683,
      "epoch": 1.5368589743589745,
      "grad_norm": 0.020117729902267456,
      "learning_rate": 1e-06,
      "loss": -0.0044,
      "step": 959
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.25352972745895386,
      "epoch": 1.5384615384615383,
      "grad_norm": 0.019687386229634285,
      "learning_rate": 1e-06,
      "loss": -0.0013,
      "step": 960
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3396.0,
      "completions/mean_length": 1496.544921875,
      "completions/mean_terminated_length": 1467.410888671875,
      "completions/min_length": 343.0,
      "completions/min_terminated_length": 343.0,
      "entropy": 0.23719095438718796,
      "epoch": 1.5400641025641026,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 0.03074975125491619,
      "learning_rate": 1e-06,
      "loss": -0.0022,
      "num_tokens": 784215197.0,
      "reward": 0.3043617606163025,
      "reward_std": 0.09115153551101685,
      "rewards/progression_diversity/mean": -0.0003482637694105506,
      "rewards/progression_diversity/std": 0.006979175843298435,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.6395508050918579,
      "rewards/symbolic_reward_partial_score/std": 0.25730493664741516,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0488170385360718,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 1.163615345954895,
      "step": 961
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2416258454322815,
      "epoch": 1.5416666666666665,
      "grad_norm": 0.02943342924118042,
      "learning_rate": 1e-06,
      "loss": 0.0171,
      "step": 962
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.24121008068323135,
      "epoch": 1.5432692307692308,
      "grad_norm": 0.022504808381199837,
      "learning_rate": 1e-06,
      "loss": -0.0017,
      "step": 963
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.23982763290405273,
      "epoch": 1.5448717948717947,
      "grad_norm": 0.026652559638023376,
      "learning_rate": 1e-06,
      "loss": -0.0056,
      "step": 964
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3516.0,
      "completions/max_terminated_length": 3516.0,
      "completions/mean_length": 1295.953125,
      "completions/mean_terminated_length": 1295.953125,
      "completions/min_length": 420.0,
      "completions/min_terminated_length": 420.0,
      "entropy": 0.2383585274219513,
      "epoch": 1.546474358974359,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 0.03317641466856003,
      "learning_rate": 1e-06,
      "loss": -0.0043,
      "num_tokens": 785827077.0,
      "reward": 0.2608430087566376,
      "reward_std": 0.09105145186185837,
      "rewards/progression_diversity/mean": -0.00017204870528075844,
      "rewards/progression_diversity/std": 0.002580570289865136,
      "rewards/symbolic_reward_accuracy/mean": 0.091796875,
      "rewards/symbolic_reward_accuracy/std": 0.289021372795105,
      "rewards/symbolic_reward_partial_score/mean": 0.6858886480331421,
      "rewards/symbolic_reward_partial_score/std": 0.18577834963798523,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0511445999145508,
      "sampling/importance_sampling_ratio/min": 2.0922935606293436e-10,
      "sampling/sampling_logp_difference/max": 22.28759002685547,
      "sampling/sampling_logp_difference/mean": 0.10452830791473389,
      "step": 965
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2417585551738739,
      "epoch": 1.5480769230769231,
      "grad_norm": 0.015128690749406815,
      "learning_rate": 1e-06,
      "loss": 0.0021,
      "step": 966
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.24052689224481583,
      "epoch": 1.5496794871794872,
      "grad_norm": 0.016625335440039635,
      "learning_rate": 1e-06,
      "loss": 0.0104,
      "step": 967
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.23810868710279465,
      "epoch": 1.5512820512820513,
      "grad_norm": 0.024602653458714485,
      "learning_rate": 1e-06,
      "loss": -0.0002,
      "step": 968
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3115.0,
      "completions/max_terminated_length": 3115.0,
      "completions/mean_length": 1370.482421875,
      "completions/mean_terminated_length": 1370.482421875,
      "completions/min_length": 480.0,
      "completions/min_terminated_length": 480.0,
      "entropy": 0.24552591890096664,
      "epoch": 1.5528846153846154,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.030604401603341103,
      "learning_rate": 1e-06,
      "loss": -0.0049,
      "num_tokens": 787405868.0,
      "reward": 0.22608765959739685,
      "reward_std": 0.0656827911734581,
      "rewards/progression_diversity/mean": -0.00012242203229106963,
      "rewards/progression_diversity/std": 0.0018643095390871167,
      "rewards/symbolic_reward_accuracy/mean": 0.060546875,
      "rewards/symbolic_reward_accuracy/std": 0.2387305200099945,
      "rewards/symbolic_reward_partial_score/mean": 0.6325358152389526,
      "rewards/symbolic_reward_partial_score/std": 0.17896120250225067,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.051476240158081,
      "sampling/importance_sampling_ratio/min": 0.003860935801640153,
      "sampling/sampling_logp_difference/max": 5.556845664978027,
      "sampling/sampling_logp_difference/mean": 0.10471828281879425,
      "step": 969
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.24233480542898178,
      "epoch": 1.5544871794871795,
      "grad_norm": 0.017392568290233612,
      "learning_rate": 1e-06,
      "loss": 0.0187,
      "step": 970
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2429162785410881,
      "epoch": 1.5560897435897436,
      "grad_norm": 0.02391919493675232,
      "learning_rate": 1e-06,
      "loss": -0.0042,
      "step": 971
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.24405669420957565,
      "epoch": 1.5576923076923077,
      "grad_norm": 0.020459329709410667,
      "learning_rate": 1e-06,
      "loss": 0.0036,
      "step": 972
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3250.0,
      "completions/max_terminated_length": 3250.0,
      "completions/mean_length": 1407.92578125,
      "completions/mean_terminated_length": 1407.92578125,
      "completions/min_length": 384.0,
      "completions/min_terminated_length": 384.0,
      "entropy": 0.2465820536017418,
      "epoch": 1.5592948717948718,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.02617330104112625,
      "learning_rate": 1e-06,
      "loss": 0.0132,
      "num_tokens": 788955078.0,
      "reward": 0.2446797788143158,
      "reward_std": 0.07382390648126602,
      "rewards/progression_diversity/mean": -0.00028444104827940464,
      "rewards/progression_diversity/std": 0.002962089842185378,
      "rewards/symbolic_reward_accuracy/mean": 0.078125,
      "rewards/symbolic_reward_accuracy/std": 0.26863065361976624,
      "rewards/symbolic_reward_partial_score/mean": 0.6593587398529053,
      "rewards/symbolic_reward_partial_score/std": 0.18929307162761688,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0518171787261963,
      "sampling/importance_sampling_ratio/min": 7.697512046433985e-06,
      "sampling/sampling_logp_difference/max": 11.774613380432129,
      "sampling/sampling_logp_difference/mean": 0.10526497662067413,
      "step": 973
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.24588342010974884,
      "epoch": 1.560897435897436,
      "grad_norm": 0.022333713248372078,
      "learning_rate": 1e-06,
      "loss": 0.0055,
      "step": 974
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.24611911922693253,
      "epoch": 1.5625,
      "grad_norm": 0.01700982078909874,
      "learning_rate": 1e-06,
      "loss": 0.0166,
      "step": 975
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2488018348813057,
      "epoch": 1.564102564102564,
      "grad_norm": 0.01537299808114767,
      "learning_rate": 1e-06,
      "loss": -0.0055,
      "step": 976
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3609.0,
      "completions/max_terminated_length": 3609.0,
      "completions/mean_length": 1228.7265625,
      "completions/mean_terminated_length": 1228.7265625,
      "completions/min_length": 404.0,
      "completions/min_terminated_length": 404.0,
      "entropy": 0.2481737807393074,
      "epoch": 1.5657051282051282,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 0.03226266801357269,
      "learning_rate": 1e-06,
      "loss": 0.0027,
      "num_tokens": 790546554.0,
      "reward": 0.19851306080818176,
      "reward_std": 0.0660344660282135,
      "rewards/progression_diversity/mean": -0.0002565660688560456,
      "rewards/progression_diversity/std": 0.0029936530627310276,
      "rewards/symbolic_reward_accuracy/mean": 0.044921875,
      "rewards/symbolic_reward_accuracy/std": 0.20733514428138733,
      "rewards/symbolic_reward_partial_score/mean": 0.5718749761581421,
      "rewards/symbolic_reward_partial_score/std": 0.20530447363853455,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0529671907424927,
      "sampling/importance_sampling_ratio/min": 0.0003155683516524732,
      "sampling/sampling_logp_difference/max": 8.061135292053223,
      "sampling/sampling_logp_difference/mean": 0.10595399141311646,
      "step": 977
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.24593167752027512,
      "epoch": 1.5673076923076923,
      "grad_norm": 0.017432967200875282,
      "learning_rate": 1e-06,
      "loss": -0.0049,
      "step": 978
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.24341944605112076,
      "epoch": 1.5689102564102564,
      "grad_norm": 0.01746542379260063,
      "learning_rate": 1e-06,
      "loss": 0.0044,
      "step": 979
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2890625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.24650365114212036,
      "epoch": 1.5705128205128205,
      "grad_norm": 0.015472476370632648,
      "learning_rate": 1e-06,
      "loss": -0.003,
      "step": 980
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3315.0,
      "completions/max_terminated_length": 3315.0,
      "completions/mean_length": 1509.9140625,
      "completions/mean_terminated_length": 1509.9140625,
      "completions/min_length": 491.0,
      "completions/min_terminated_length": 491.0,
      "entropy": 0.2508416026830673,
      "epoch": 1.5721153846153846,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 0.025885384529829025,
      "learning_rate": 1e-06,
      "loss": 0.0112,
      "num_tokens": 792077166.0,
      "reward": 0.3290412425994873,
      "reward_std": 0.11779160797595978,
      "rewards/progression_diversity/mean": -0.00017644937906879932,
      "rewards/progression_diversity/std": 0.0018998431041836739,
      "rewards/symbolic_reward_accuracy/mean": 0.208984375,
      "rewards/symbolic_reward_accuracy/std": 0.40698084235191345,
      "rewards/symbolic_reward_partial_score/mean": 0.6788411140441895,
      "rewards/symbolic_reward_partial_score/std": 0.22582782804965973,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0522162914276123,
      "sampling/importance_sampling_ratio/min": 0.00034631905145943165,
      "sampling/sampling_logp_difference/max": 7.9681501388549805,
      "sampling/sampling_logp_difference/mean": 0.10585089027881622,
      "step": 981
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.24679741263389587,
      "epoch": 1.5737179487179487,
      "grad_norm": 0.025917481631040573,
      "learning_rate": 1e-06,
      "loss": -0.0061,
      "step": 982
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.24447974562644958,
      "epoch": 1.5753205128205128,
      "grad_norm": 0.020733220502734184,
      "learning_rate": 1e-06,
      "loss": -0.0044,
      "step": 983
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2511764466762543,
      "epoch": 1.5769230769230769,
      "grad_norm": 0.022386424243450165,
      "learning_rate": 1e-06,
      "loss": 0.0115,
      "step": 984
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3325.0,
      "completions/max_terminated_length": 3325.0,
      "completions/mean_length": 1401.8359375,
      "completions/mean_terminated_length": 1401.8359375,
      "completions/min_length": 396.0,
      "completions/min_terminated_length": 396.0,
      "entropy": 0.25129084289073944,
      "epoch": 1.578525641025641,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.03222787380218506,
      "learning_rate": 1e-06,
      "loss": 0.0093,
      "num_tokens": 793645338.0,
      "reward": 0.23190167546272278,
      "reward_std": 0.05650953948497772,
      "rewards/progression_diversity/mean": -0.0002622118918225169,
      "rewards/progression_diversity/std": 0.0023251688107848167,
      "rewards/symbolic_reward_accuracy/mean": 0.0703125,
      "rewards/symbolic_reward_accuracy/std": 0.25592297315597534,
      "rewards/symbolic_reward_partial_score/mean": 0.6323893666267395,
      "rewards/symbolic_reward_partial_score/std": 0.19433899223804474,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0537452697753906,
      "sampling/importance_sampling_ratio/min": 2.0533148514800814e-08,
      "sampling/sampling_logp_difference/max": 17.70122528076172,
      "sampling/sampling_logp_difference/mean": 0.10701755434274673,
      "step": 985
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.2517654523253441,
      "epoch": 1.5801282051282053,
      "grad_norm": 0.023150594905018806,
      "learning_rate": 1e-06,
      "loss": -0.0046,
      "step": 986
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2524363100528717,
      "epoch": 1.5817307692307692,
      "grad_norm": 0.017600167542696,
      "learning_rate": 1e-06,
      "loss": 0.0146,
      "step": 987
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2514464929699898,
      "epoch": 1.5833333333333335,
      "grad_norm": 0.014767357148230076,
      "learning_rate": 1e-06,
      "loss": 0.0002,
      "step": 988
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3305.0,
      "completions/mean_length": 1364.2421875,
      "completions/mean_terminated_length": 1334.849365234375,
      "completions/min_length": 428.0,
      "completions/min_terminated_length": 428.0,
      "entropy": 0.24437803030014038,
      "epoch": 1.5849358974358974,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.026894228532910347,
      "learning_rate": 1e-06,
      "loss": 0.0183,
      "num_tokens": 795228758.0,
      "reward": 0.25398755073547363,
      "reward_std": 0.0541648268699646,
      "rewards/progression_diversity/mean": -0.00017180124996230006,
      "rewards/progression_diversity/std": 0.0016380609013140202,
      "rewards/symbolic_reward_accuracy/mean": 0.095703125,
      "rewards/symbolic_reward_accuracy/std": 0.2944713830947876,
      "rewards/symbolic_reward_partial_score/mean": 0.6558756828308105,
      "rewards/symbolic_reward_partial_score/std": 0.18879389762878418,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0525712966918945,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 0.19978845119476318,
      "step": 989
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.24415334314107895,
      "epoch": 1.5865384615384617,
      "grad_norm": 0.02484998293220997,
      "learning_rate": 1e-06,
      "loss": -0.0023,
      "step": 990
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.25222089886665344,
      "epoch": 1.5881410256410255,
      "grad_norm": 0.01625809818506241,
      "learning_rate": 1e-06,
      "loss": 0.004,
      "step": 991
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2524196058511734,
      "epoch": 1.5897435897435899,
      "grad_norm": 0.019578304141759872,
      "learning_rate": 1e-06,
      "loss": 0.003,
      "step": 992
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3574.0,
      "completions/max_terminated_length": 3574.0,
      "completions/mean_length": 1432.29296875,
      "completions/mean_terminated_length": 1432.29296875,
      "completions/min_length": 513.0,
      "completions/min_terminated_length": 513.0,
      "entropy": 0.25225862860679626,
      "epoch": 1.5913461538461537,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 0.02162688970565796,
      "learning_rate": 1e-06,
      "loss": 0.0063,
      "num_tokens": 796763980.0,
      "reward": 0.2408090978860855,
      "reward_std": 0.06670250743627548,
      "rewards/progression_diversity/mean": -0.00014496369112748653,
      "rewards/progression_diversity/std": 0.0017526200972497463,
      "rewards/symbolic_reward_accuracy/mean": 0.080078125,
      "rewards/symbolic_reward_accuracy/std": 0.271679550409317,
      "rewards/symbolic_reward_partial_score/mean": 0.6425455808639526,
      "rewards/symbolic_reward_partial_score/std": 0.1895546019077301,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0530476570129395,
      "sampling/importance_sampling_ratio/min": 0.003646494820713997,
      "sampling/sampling_logp_difference/max": 5.613988876342773,
      "sampling/sampling_logp_difference/mean": 0.10609988868236542,
      "step": 993
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2456853985786438,
      "epoch": 1.592948717948718,
      "grad_norm": 0.019648630172014236,
      "learning_rate": 1e-06,
      "loss": -0.0031,
      "step": 994
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2553592026233673,
      "epoch": 1.594551282051282,
      "grad_norm": 0.025570129975676537,
      "learning_rate": 1e-06,
      "loss": 0.0049,
      "step": 995
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.24990757554769516,
      "epoch": 1.5961538461538463,
      "grad_norm": 0.04156876355409622,
      "learning_rate": 1e-06,
      "loss": 0.0087,
      "step": 996
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3606.0,
      "completions/max_terminated_length": 3606.0,
      "completions/mean_length": 1444.166015625,
      "completions/mean_terminated_length": 1444.166015625,
      "completions/min_length": 420.0,
      "completions/min_terminated_length": 420.0,
      "entropy": 0.2480572983622551,
      "epoch": 1.5977564102564101,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 0.029209570959210396,
      "learning_rate": 1e-06,
      "loss": -0.0069,
      "num_tokens": 798317441.0,
      "reward": 0.2392357736825943,
      "reward_std": 0.06654238700866699,
      "rewards/progression_diversity/mean": -0.0002508866600692272,
      "rewards/progression_diversity/std": 0.0026269752997905016,
      "rewards/symbolic_reward_accuracy/mean": 0.091796875,
      "rewards/symbolic_reward_accuracy/std": 0.289021372795105,
      "rewards/symbolic_reward_partial_score/mean": 0.6138671636581421,
      "rewards/symbolic_reward_partial_score/std": 0.2098008245229721,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0531139373779297,
      "sampling/importance_sampling_ratio/min": 0.003923820797353983,
      "sampling/sampling_logp_difference/max": 5.540689468383789,
      "sampling/sampling_logp_difference/mean": 0.10546484589576721,
      "step": 997
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.24528241902589798,
      "epoch": 1.5993589743589745,
      "grad_norm": 0.02387697622179985,
      "learning_rate": 1e-06,
      "loss": 0.0125,
      "step": 998
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.24644892662763596,
      "epoch": 1.6009615384615383,
      "grad_norm": 0.023926518857479095,
      "learning_rate": 1e-06,
      "loss": 0.0259,
      "step": 999
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.24555939435958862,
      "epoch": 1.6025641025641026,
      "grad_norm": 0.02301941066980362,
      "learning_rate": 1e-06,
      "loss": -0.0124,
      "step": 1000
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3619.0,
      "completions/mean_length": 1375.71875,
      "completions/mean_terminated_length": 1346.3482666015625,
      "completions/min_length": 456.0,
      "completions/min_terminated_length": 456.0,
      "entropy": 0.23569654673337936,
      "epoch": 1.6041666666666665,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.03722744435071945,
      "learning_rate": 1e-06,
      "loss": -0.0089,
      "num_tokens": 800005953.0,
      "reward": 0.23674920201301575,
      "reward_std": 0.0885818749666214,
      "rewards/progression_diversity/mean": -0.0008616495179012418,
      "rewards/progression_diversity/std": 0.013127473182976246,
      "rewards/symbolic_reward_accuracy/mean": 0.095703125,
      "rewards/symbolic_reward_accuracy/std": 0.2944713830947876,
      "rewards/symbolic_reward_partial_score/mean": 0.598437488079071,
      "rewards/symbolic_reward_partial_score/std": 0.22310422360897064,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0491344928741455,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 1.0696396827697754,
      "step": 1001
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.2391332983970642,
      "epoch": 1.6057692307692308,
      "grad_norm": 0.023793857544660568,
      "learning_rate": 1e-06,
      "loss": 0.0327,
      "step": 1002
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.2433808520436287,
      "epoch": 1.6073717948717947,
      "grad_norm": 0.01809019409120083,
      "learning_rate": 1e-06,
      "loss": 0.0041,
      "step": 1003
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.23954389244318008,
      "epoch": 1.608974358974359,
      "grad_norm": 0.017016278579831123,
      "learning_rate": 1e-06,
      "loss": -0.001,
      "step": 1004
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3544.0,
      "completions/max_terminated_length": 3544.0,
      "completions/mean_length": 1332.548828125,
      "completions/mean_terminated_length": 1332.548828125,
      "completions/min_length": 454.0,
      "completions/min_terminated_length": 454.0,
      "entropy": 0.24541258811950684,
      "epoch": 1.6105769230769231,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 0.024081578478217125,
      "learning_rate": 1e-06,
      "loss": 0.0043,
      "num_tokens": 801514282.0,
      "reward": 0.2921689450740814,
      "reward_std": 0.07876162230968475,
      "rewards/progression_diversity/mean": -0.0003911844396498054,
      "rewards/progression_diversity/std": 0.0043034846894443035,
      "rewards/symbolic_reward_accuracy/mean": 0.140625,
      "rewards/symbolic_reward_accuracy/std": 0.3479743003845215,
      "rewards/symbolic_reward_partial_score/mean": 0.6926594972610474,
      "rewards/symbolic_reward_partial_score/std": 0.20360209047794342,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.052016019821167,
      "sampling/importance_sampling_ratio/min": 8.38755295262672e-06,
      "sampling/sampling_logp_difference/max": 11.688761711120605,
      "sampling/sampling_logp_difference/mean": 0.10484174638986588,
      "step": 1005
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.23762197047472,
      "epoch": 1.6121794871794872,
      "grad_norm": 0.017401108518242836,
      "learning_rate": 1e-06,
      "loss": 0.0039,
      "step": 1006
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.24435139447450638,
      "epoch": 1.6137820512820513,
      "grad_norm": 0.02926545962691307,
      "learning_rate": 1e-06,
      "loss": 0.0016,
      "step": 1007
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.24484486877918243,
      "epoch": 1.6153846153846154,
      "grad_norm": 0.023529132828116417,
      "learning_rate": 1e-06,
      "loss": -0.0068,
      "step": 1008
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3515.0,
      "completions/mean_length": 1400.970703125,
      "completions/mean_terminated_length": 1371.649658203125,
      "completions/min_length": 404.0,
      "completions/min_terminated_length": 404.0,
      "entropy": 0.23839521408081055,
      "epoch": 1.6169871794871795,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 0.036472517997026443,
      "learning_rate": 1e-06,
      "loss": 0.0033,
      "num_tokens": 803169707.0,
      "reward": 0.23677846789360046,
      "reward_std": 0.052502475678920746,
      "rewards/progression_diversity/mean": -0.0003774126525968313,
      "rewards/progression_diversity/std": 0.0034933658316731453,
      "rewards/symbolic_reward_accuracy/mean": 0.0703125,
      "rewards/symbolic_reward_accuracy/std": 0.25592297315597534,
      "rewards/symbolic_reward_partial_score/mean": 0.6493000984191895,
      "rewards/symbolic_reward_partial_score/std": 0.23349793255329132,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0512499809265137,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 228.0,
      "sampling/sampling_logp_difference/mean": 0.11092744767665863,
      "step": 1009
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.24148935079574585,
      "epoch": 1.6185897435897436,
      "grad_norm": 0.029029127210378647,
      "learning_rate": 1e-06,
      "loss": 0.0061,
      "step": 1010
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.24070608615875244,
      "epoch": 1.6201923076923077,
      "grad_norm": 0.020563537254929543,
      "learning_rate": 1e-06,
      "loss": 0.009,
      "step": 1011
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.24124725908041,
      "epoch": 1.6217948717948718,
      "grad_norm": 0.016905367374420166,
      "learning_rate": 1e-06,
      "loss": 0.0065,
      "step": 1012
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3733.0,
      "completions/mean_length": 1487.85546875,
      "completions/mean_terminated_length": 1400.0589599609375,
      "completions/min_length": 485.0,
      "completions/min_terminated_length": 485.0,
      "entropy": 0.2362624704837799,
      "epoch": 1.623397435897436,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 0.025372762233018875,
      "learning_rate": 1e-06,
      "loss": 0.0274,
      "num_tokens": 804794641.0,
      "reward": 0.2821791172027588,
      "reward_std": 0.1007528007030487,
      "rewards/progression_diversity/mean": -0.0003515832358971238,
      "rewards/progression_diversity/std": 0.004287549760192633,
      "rewards/symbolic_reward_accuracy/mean": 0.150390625,
      "rewards/symbolic_reward_accuracy/std": 0.35780346393585205,
      "rewards/symbolic_reward_partial_score/mean": 0.6417806148529053,
      "rewards/symbolic_reward_partial_score/std": 0.22223784029483795,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0479587316513062,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 0.5552178621292114,
      "step": 1013
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.237822063267231,
      "epoch": 1.625,
      "grad_norm": 0.017621813341975212,
      "learning_rate": 1e-06,
      "loss": -0.0144,
      "step": 1014
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.24151387810707092,
      "epoch": 1.626602564102564,
      "grad_norm": 0.01545657031238079,
      "learning_rate": 1e-06,
      "loss": -0.0025,
      "step": 1015
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23863951116800308,
      "epoch": 1.6282051282051282,
      "grad_norm": 153.45928955078125,
      "learning_rate": 1e-06,
      "loss": 0.0558,
      "step": 1016
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4848.0,
      "completions/mean_length": 1533.744140625,
      "completions/mean_terminated_length": 1504.6829833984375,
      "completions/min_length": 334.0,
      "completions/min_terminated_length": 334.0,
      "entropy": 0.23886650800704956,
      "epoch": 1.6298076923076923,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.03264259546995163,
      "learning_rate": 1e-06,
      "loss": -0.0017,
      "num_tokens": 806450974.0,
      "reward": 0.1910867989063263,
      "reward_std": 0.055806443095207214,
      "rewards/progression_diversity/mean": -0.0006959763122722507,
      "rewards/progression_diversity/std": 0.00666604982689023,
      "rewards/symbolic_reward_accuracy/mean": 0.025390625,
      "rewards/symbolic_reward_accuracy/std": 0.15746226906776428,
      "rewards/symbolic_reward_partial_score/mean": 0.5868489742279053,
      "rewards/symbolic_reward_partial_score/std": 0.21801504492759705,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0492140054702759,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.3248835504055023,
      "step": 1017
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.23825782537460327,
      "epoch": 1.6314102564102564,
      "grad_norm": 0.02487725391983986,
      "learning_rate": 1e-06,
      "loss": 0.0279,
      "step": 1018
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23552914708852768,
      "epoch": 1.6330128205128205,
      "grad_norm": 0.029635349288582802,
      "learning_rate": 1e-06,
      "loss": 0.0094,
      "step": 1019
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2386602759361267,
      "epoch": 1.6346153846153846,
      "grad_norm": 0.02634367346763611,
      "learning_rate": 1e-06,
      "loss": 0.0005,
      "step": 1020
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3918.0,
      "completions/mean_length": 1762.474609375,
      "completions/mean_terminated_length": 1676.2967529296875,
      "completions/min_length": 469.0,
      "completions/min_terminated_length": 469.0,
      "entropy": 0.23500239849090576,
      "epoch": 1.6362179487179487,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.02729848027229309,
      "learning_rate": 1e-06,
      "loss": -0.0103,
      "num_tokens": 808198689.0,
      "reward": 0.29712605476379395,
      "reward_std": 0.09707895666360855,
      "rewards/progression_diversity/mean": -0.0002862706023734063,
      "rewards/progression_diversity/std": 0.004292279481887817,
      "rewards/symbolic_reward_accuracy/mean": 0.1484375,
      "rewards/symbolic_reward_accuracy/std": 0.35588082671165466,
      "rewards/symbolic_reward_partial_score/mean": 0.695507824420929,
      "rewards/symbolic_reward_partial_score/std": 0.23922735452651978,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.047842025756836,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 0.4406740665435791,
      "step": 1021
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.23327183723449707,
      "epoch": 1.6378205128205128,
      "grad_norm": 20.64282989501953,
      "learning_rate": 1e-06,
      "loss": 0.0436,
      "step": 1022
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.22972074151039124,
      "epoch": 1.6394230769230769,
      "grad_norm": 0.02955857291817665,
      "learning_rate": 1e-06,
      "loss": 0.0107,
      "step": 1023
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.23090046644210815,
      "epoch": 1.641025641025641,
      "grad_norm": 0.025243405252695084,
      "learning_rate": 1e-06,
      "loss": -0.0164,
      "step": 1024
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4233.0,
      "completions/mean_length": 1425.66796875,
      "completions/mean_terminated_length": 1396.395263671875,
      "completions/min_length": 325.0,
      "completions/min_terminated_length": 325.0,
      "entropy": 0.24533987045288086,
      "epoch": 1.6426282051282053,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 0.031315580010414124,
      "learning_rate": 1e-06,
      "loss": -0.0027,
      "num_tokens": 809735287.0,
      "reward": 0.2972439229488373,
      "reward_std": 0.09149158000946045,
      "rewards/progression_diversity/mean": -0.00021706035477109253,
      "rewards/progression_diversity/std": 0.0027459680568426847,
      "rewards/symbolic_reward_accuracy/mean": 0.1640625,
      "rewards/symbolic_reward_accuracy/std": 0.37069445848464966,
      "rewards/symbolic_reward_partial_score/mean": 0.6633463501930237,
      "rewards/symbolic_reward_partial_score/std": 0.2075018286705017,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0503005981445312,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 704.0,
      "sampling/sampling_logp_difference/mean": 0.17904753983020782,
      "step": 1025
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.23708289116621017,
      "epoch": 1.6442307692307692,
      "grad_norm": 0.04043427109718323,
      "learning_rate": 1e-06,
      "loss": 0.0122,
      "step": 1026
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2370290458202362,
      "epoch": 1.6458333333333335,
      "grad_norm": 0.019201675429940224,
      "learning_rate": 1e-06,
      "loss": 0.0346,
      "step": 1027
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23796197026968002,
      "epoch": 1.6474358974358974,
      "grad_norm": 0.021262312307953835,
      "learning_rate": 1e-06,
      "loss": 0.0163,
      "step": 1028
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4151.0,
      "completions/max_terminated_length": 4151.0,
      "completions/mean_length": 1320.484375,
      "completions/mean_terminated_length": 1320.484375,
      "completions/min_length": 276.0,
      "completions/min_terminated_length": 276.0,
      "entropy": 0.22943055629730225,
      "epoch": 1.6490384615384617,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 0.03114836849272251,
      "learning_rate": 1e-06,
      "loss": 0.0025,
      "num_tokens": 811322847.0,
      "reward": 0.31148767471313477,
      "reward_std": 0.11784271150827408,
      "rewards/progression_diversity/mean": -0.00015914140385575593,
      "rewards/progression_diversity/std": 0.0020391715224832296,
      "rewards/symbolic_reward_accuracy/mean": 0.181640625,
      "rewards/symbolic_reward_accuracy/std": 0.38592514395713806,
      "rewards/symbolic_reward_partial_score/mean": 0.6750162839889526,
      "rewards/symbolic_reward_partial_score/std": 0.22747963666915894,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0495221614837646,
      "sampling/importance_sampling_ratio/min": 0.0016271222848445177,
      "sampling/sampling_logp_difference/max": 6.420942306518555,
      "sampling/sampling_logp_difference/mean": 0.10276812314987183,
      "step": 1029
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.23415642231702805,
      "epoch": 1.6506410256410255,
      "grad_norm": 0.017106305807828903,
      "learning_rate": 1e-06,
      "loss": -0.007,
      "step": 1030
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2420409992337227,
      "epoch": 1.6522435897435899,
      "grad_norm": 0.021166684105992317,
      "learning_rate": 1e-06,
      "loss": 0.0165,
      "step": 1031
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.22991270571947098,
      "epoch": 1.6538461538461537,
      "grad_norm": 0.026932181790471077,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "step": 1032
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3803.0,
      "completions/mean_length": 1430.66015625,
      "completions/mean_terminated_length": 1372.0196533203125,
      "completions/min_length": 468.0,
      "completions/min_terminated_length": 468.0,
      "entropy": 0.23568789660930634,
      "epoch": 1.655448717948718,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 0.02257698029279709,
      "learning_rate": 1e-06,
      "loss": 0.0309,
      "num_tokens": 812878337.0,
      "reward": 0.372047483921051,
      "reward_std": 0.05590501427650452,
      "rewards/progression_diversity/mean": -0.0008204940240830183,
      "rewards/progression_diversity/std": 0.008694627322256565,
      "rewards/symbolic_reward_accuracy/mean": 0.255859375,
      "rewards/symbolic_reward_accuracy/std": 0.43676990270614624,
      "rewards/symbolic_reward_partial_score/mean": 0.7297688722610474,
      "rewards/symbolic_reward_partial_score/std": 0.2233889102935791,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0493979454040527,
      "sampling/importance_sampling_ratio/min": 1.542988610145346e-12,
      "sampling/sampling_logp_difference/max": 27.19729995727539,
      "sampling/sampling_logp_difference/mean": 0.10240708291530609,
      "step": 1033
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2355349287390709,
      "epoch": 1.657051282051282,
      "grad_norm": 0.017844535410404205,
      "learning_rate": 1e-06,
      "loss": 0.0074,
      "step": 1034
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.23197440803050995,
      "epoch": 1.6586538461538463,
      "grad_norm": 0.020765669643878937,
      "learning_rate": 1e-06,
      "loss": 0.0046,
      "step": 1035
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.23954713344573975,
      "epoch": 1.6602564102564101,
      "grad_norm": 0.02009597420692444,
      "learning_rate": 1e-06,
      "loss": -0.0043,
      "step": 1036
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4692.0,
      "completions/max_terminated_length": 4692.0,
      "completions/mean_length": 1252.927734375,
      "completions/mean_terminated_length": 1252.927734375,
      "completions/min_length": 400.0,
      "completions/min_terminated_length": 400.0,
      "entropy": 0.23540305346250534,
      "epoch": 1.6618589743589745,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.022153835743665695,
      "learning_rate": 1e-06,
      "loss": 0.0036,
      "num_tokens": 814426508.0,
      "reward": 0.29029542207717896,
      "reward_std": 0.04895063489675522,
      "rewards/progression_diversity/mean": -0.00024394365027546883,
      "rewards/progression_diversity/std": 0.0027705603279173374,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.7176594734191895,
      "rewards/symbolic_reward_partial_score/std": 0.19320081174373627,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0504940748214722,
      "sampling/importance_sampling_ratio/min": 0.0024849912151694298,
      "sampling/sampling_logp_difference/max": 5.997486114501953,
      "sampling/sampling_logp_difference/mean": 0.10428472608327866,
      "step": 1037
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.24068902432918549,
      "epoch": 1.6634615384615383,
      "grad_norm": 0.021785540506243706,
      "learning_rate": 1e-06,
      "loss": 0.0068,
      "step": 1038
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.23235713690519333,
      "epoch": 1.6650641025641026,
      "grad_norm": 0.020368505269289017,
      "learning_rate": 1e-06,
      "loss": 0.0047,
      "step": 1039
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.236878864467144,
      "epoch": 1.6666666666666665,
      "grad_norm": 0.0215722918510437,
      "learning_rate": 1e-06,
      "loss": 0.0003,
      "step": 1040
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4650.0,
      "completions/mean_length": 1380.171875,
      "completions/mean_terminated_length": 1291.74072265625,
      "completions/min_length": 391.0,
      "completions/min_terminated_length": 391.0,
      "entropy": 0.23111912608146667,
      "epoch": 1.6682692307692308,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 2.0536277294158936,
      "learning_rate": 1e-06,
      "loss": 0.0259,
      "num_tokens": 816040052.0,
      "reward": 0.26442086696624756,
      "reward_std": 0.0863037258386612,
      "rewards/progression_diversity/mean": -0.0022489791736006737,
      "rewards/progression_diversity/std": 0.027088049799203873,
      "rewards/symbolic_reward_accuracy/mean": 0.123046875,
      "rewards/symbolic_reward_accuracy/std": 0.32881227135658264,
      "rewards/symbolic_reward_partial_score/mean": 0.6373372077941895,
      "rewards/symbolic_reward_partial_score/std": 0.2190377116203308,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.047485589981079,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 0.2174200415611267,
      "step": 1041
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.23076869547367096,
      "epoch": 1.6698717948717947,
      "grad_norm": 0.016981706023216248,
      "learning_rate": 1e-06,
      "loss": -0.0025,
      "step": 1042
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.22752298414707184,
      "epoch": 1.671474358974359,
      "grad_norm": 0.019416002556681633,
      "learning_rate": 1e-06,
      "loss": 0.0329,
      "step": 1043
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.23331021517515182,
      "epoch": 1.6730769230769231,
      "grad_norm": 0.015192318707704544,
      "learning_rate": 1e-06,
      "loss": 0.0003,
      "step": 1044
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3838.0,
      "completions/max_terminated_length": 3838.0,
      "completions/mean_length": 1214.56640625,
      "completions/mean_terminated_length": 1214.56640625,
      "completions/min_length": 375.0,
      "completions/min_terminated_length": 375.0,
      "entropy": 0.24394644051790237,
      "epoch": 1.6746794871794872,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 0.03186223655939102,
      "learning_rate": 1e-06,
      "loss": 0.0053,
      "num_tokens": 817475990.0,
      "reward": 0.31599071621894836,
      "reward_std": 0.08919353783130646,
      "rewards/progression_diversity/mean": -4.969753717887215e-05,
      "rewards/progression_diversity/std": 0.0011245269561186433,
      "rewards/symbolic_reward_accuracy/mean": 0.19140625,
      "rewards/symbolic_reward_accuracy/std": 0.3937928080558777,
      "rewards/symbolic_reward_partial_score/mean": 0.6704915165901184,
      "rewards/symbolic_reward_partial_score/std": 0.22778773307800293,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0511565208435059,
      "sampling/importance_sampling_ratio/min": 0.00036043746513314545,
      "sampling/sampling_logp_difference/max": 7.928192138671875,
      "sampling/sampling_logp_difference/mean": 0.1060151681303978,
      "step": 1045
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.23804038763046265,
      "epoch": 1.6762820512820513,
      "grad_norm": 0.012359860353171825,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "step": 1046
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.24049938470125198,
      "epoch": 1.6778846153846154,
      "grad_norm": 0.01780606620013714,
      "learning_rate": 1e-06,
      "loss": 0.0041,
      "step": 1047
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2433236464858055,
      "epoch": 1.6794871794871795,
      "grad_norm": 0.018220573663711548,
      "learning_rate": 1e-06,
      "loss": -0.0022,
      "step": 1048
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3815.0,
      "completions/max_terminated_length": 3815.0,
      "completions/mean_length": 1282.876953125,
      "completions/mean_terminated_length": 1282.876953125,
      "completions/min_length": 351.0,
      "completions/min_terminated_length": 351.0,
      "entropy": 0.24433332681655884,
      "epoch": 1.6810897435897436,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 0.032106902450323105,
      "learning_rate": 1e-06,
      "loss": -0.0046,
      "num_tokens": 818907495.0,
      "reward": 0.29857733845710754,
      "reward_std": 0.13391146063804626,
      "rewards/progression_diversity/mean": -0.0001772599498508498,
      "rewards/progression_diversity/std": 0.0024098155554383993,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.6280761957168579,
      "rewards/symbolic_reward_partial_score/std": 0.25627195835113525,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0512542724609375,
      "sampling/importance_sampling_ratio/min": 9.631957800593227e-05,
      "sampling/sampling_logp_difference/max": 9.247838973999023,
      "sampling/sampling_logp_difference/mean": 0.10506170988082886,
      "step": 1049
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.24064549058675766,
      "epoch": 1.6826923076923077,
      "grad_norm": 0.01762561686336994,
      "learning_rate": 1e-06,
      "loss": 0.0004,
      "step": 1050
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.24004538357257843,
      "epoch": 1.6842948717948718,
      "grad_norm": 0.022497214376926422,
      "learning_rate": 1e-06,
      "loss": 0.0018,
      "step": 1051
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2305016741156578,
      "epoch": 1.685897435897436,
      "grad_norm": 0.02127472683787346,
      "learning_rate": 1e-06,
      "loss": -0.0066,
      "step": 1052
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3644.0,
      "completions/max_terminated_length": 3644.0,
      "completions/mean_length": 1173.380859375,
      "completions/mean_terminated_length": 1173.380859375,
      "completions/min_length": 352.0,
      "completions/min_terminated_length": 352.0,
      "entropy": 0.23441405594348907,
      "epoch": 1.6875,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 0.02767036482691765,
      "learning_rate": 1e-06,
      "loss": 0.0067,
      "num_tokens": 820447722.0,
      "reward": 0.2577317953109741,
      "reward_std": 0.04780818521976471,
      "rewards/progression_diversity/mean": -0.0002602554450277239,
      "rewards/progression_diversity/std": 0.0028319996781647205,
      "rewards/symbolic_reward_accuracy/mean": 0.11328125,
      "rewards/symbolic_reward_accuracy/std": 0.3172462284564972,
      "rewards/symbolic_reward_partial_score/mean": 0.6325520873069763,
      "rewards/symbolic_reward_partial_score/std": 0.20778940618038177,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0493583679199219,
      "sampling/importance_sampling_ratio/min": 0.0004806618671864271,
      "sampling/sampling_logp_difference/max": 7.640346527099609,
      "sampling/sampling_logp_difference/mean": 0.10264922678470612,
      "step": 1053
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.22839296609163284,
      "epoch": 1.689102564102564,
      "grad_norm": 0.01796301268041134,
      "learning_rate": 1e-06,
      "loss": -0.0,
      "step": 1054
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.22674311697483063,
      "epoch": 1.6907051282051282,
      "grad_norm": 0.01843862235546112,
      "learning_rate": 1e-06,
      "loss": 0.0064,
      "step": 1055
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.22671975195407867,
      "epoch": 1.6923076923076923,
      "grad_norm": 0.02962394803762436,
      "learning_rate": 1e-06,
      "loss": 0.0027,
      "step": 1056
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3951.0,
      "completions/mean_length": 1224.1875,
      "completions/mean_terminated_length": 1194.5205078125,
      "completions/min_length": 374.0,
      "completions/min_terminated_length": 374.0,
      "entropy": 0.22822780907154083,
      "epoch": 1.6939102564102564,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 0.06622759997844696,
      "learning_rate": 1e-06,
      "loss": 0.0265,
      "num_tokens": 821993514.0,
      "reward": 0.26173463463783264,
      "reward_std": 0.07937778532505035,
      "rewards/progression_diversity/mean": -0.00036702307988889515,
      "rewards/progression_diversity/std": 0.0035130290780216455,
      "rewards/symbolic_reward_accuracy/mean": 0.1171875,
      "rewards/symbolic_reward_accuracy/std": 0.32195815443992615,
      "rewards/symbolic_reward_partial_score/mean": 0.6387370228767395,
      "rewards/symbolic_reward_partial_score/std": 0.2266356348991394,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0486118793487549,
      "sampling/importance_sampling_ratio/min": 1.0303001507505627e-11,
      "sampling/sampling_logp_difference/max": 25.298585891723633,
      "sampling/sampling_logp_difference/mean": 0.10055477917194366,
      "step": 1057
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.23284170031547546,
      "epoch": 1.6955128205128205,
      "grad_norm": 0.02792915143072605,
      "learning_rate": 1e-06,
      "loss": -0.0146,
      "step": 1058
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.22962763160467148,
      "epoch": 1.6971153846153846,
      "grad_norm": 0.018218589946627617,
      "learning_rate": 1e-06,
      "loss": -0.0005,
      "step": 1059
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2237945944070816,
      "epoch": 1.6987179487179487,
      "grad_norm": 0.01743151806294918,
      "learning_rate": 1e-06,
      "loss": 0.0005,
      "step": 1060
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3416.0,
      "completions/mean_length": 1289.6484375,
      "completions/mean_terminated_length": 1260.109619140625,
      "completions/min_length": 376.0,
      "completions/min_terminated_length": 376.0,
      "entropy": 0.22563714534044266,
      "epoch": 1.7003205128205128,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 0.027078071609139442,
      "learning_rate": 1e-06,
      "loss": 0.0019,
      "num_tokens": 823552902.0,
      "reward": 0.27887141704559326,
      "reward_std": 0.07552332431077957,
      "rewards/progression_diversity/mean": -6.769368337700143e-05,
      "rewards/progression_diversity/std": 0.0008789710118435323,
      "rewards/symbolic_reward_accuracy/mean": 0.146484375,
      "rewards/symbolic_reward_accuracy/std": 0.35393697023391724,
      "rewards/symbolic_reward_partial_score/mean": 0.637255847454071,
      "rewards/symbolic_reward_partial_score/std": 0.2156410962343216,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0477631092071533,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 0.3541214168071747,
      "step": 1061
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.22320057451725006,
      "epoch": 1.7019230769230769,
      "grad_norm": 614.7477416992188,
      "learning_rate": 1e-06,
      "loss": 0.0147,
      "step": 1062
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2247406244277954,
      "epoch": 1.703525641025641,
      "grad_norm": 0.016136281192302704,
      "learning_rate": 1e-06,
      "loss": 0.0032,
      "step": 1063
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.22823335230350494,
      "epoch": 1.7051282051282053,
      "grad_norm": 0.02105732634663582,
      "learning_rate": 1e-06,
      "loss": -0.004,
      "step": 1064
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4059.0,
      "completions/mean_length": 1367.603515625,
      "completions/mean_terminated_length": 1338.2171630859375,
      "completions/min_length": 352.0,
      "completions/min_terminated_length": 352.0,
      "entropy": 0.2308506816625595,
      "epoch": 1.7067307692307692,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 0.027784336358308792,
      "learning_rate": 1e-06,
      "loss": -0.003,
      "num_tokens": 825156523.0,
      "reward": 0.2746616005897522,
      "reward_std": 0.09659157693386078,
      "rewards/progression_diversity/mean": -0.0001522963575553149,
      "rewards/progression_diversity/std": 0.0019277379615232348,
      "rewards/symbolic_reward_accuracy/mean": 0.134765625,
      "rewards/symbolic_reward_accuracy/std": 0.3418070077896118,
      "rewards/symbolic_reward_partial_score/mean": 0.6466634273529053,
      "rewards/symbolic_reward_partial_score/std": 0.22976163029670715,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0482929944992065,
      "sampling/importance_sampling_ratio/min": 4.196563168079592e-06,
      "sampling/sampling_logp_difference/max": 12.381244659423828,
      "sampling/sampling_logp_difference/mean": 0.10019037127494812,
      "step": 1065
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.22658158838748932,
      "epoch": 1.7083333333333335,
      "grad_norm": 0.019185129553079605,
      "learning_rate": 1e-06,
      "loss": -0.004,
      "step": 1066
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2332712933421135,
      "epoch": 1.7099358974358974,
      "grad_norm": 0.0282441433519125,
      "learning_rate": 1e-06,
      "loss": 0.0069,
      "step": 1067
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.23132744431495667,
      "epoch": 1.7115384615384617,
      "grad_norm": 0.160556823015213,
      "learning_rate": 1e-06,
      "loss": 0.0127,
      "step": 1068
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4672.0,
      "completions/max_terminated_length": 4672.0,
      "completions/mean_length": 1349.27734375,
      "completions/mean_terminated_length": 1349.27734375,
      "completions/min_length": 402.0,
      "completions/min_terminated_length": 402.0,
      "entropy": 0.22879090905189514,
      "epoch": 1.7131410256410255,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 0.10462523251771927,
      "learning_rate": 1e-06,
      "loss": 0.0044,
      "num_tokens": 826677881.0,
      "reward": 0.2972085177898407,
      "reward_std": 0.11050209403038025,
      "rewards/progression_diversity/mean": -0.0008295955485664308,
      "rewards/progression_diversity/std": 0.008723512291908264,
      "rewards/symbolic_reward_accuracy/mean": 0.158203125,
      "rewards/symbolic_reward_accuracy/std": 0.36528825759887695,
      "rewards/symbolic_reward_partial_score/mean": 0.67431640625,
      "rewards/symbolic_reward_partial_score/std": 0.23070181906223297,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0478163957595825,
      "sampling/importance_sampling_ratio/min": 0.0023437354248017073,
      "sampling/sampling_logp_difference/max": 6.056009292602539,
      "sampling/sampling_logp_difference/mean": 0.09988612681627274,
      "step": 1069
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.22955607622861862,
      "epoch": 1.7147435897435899,
      "grad_norm": 0.022891266271471977,
      "learning_rate": 1e-06,
      "loss": 0.0089,
      "step": 1070
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.23325985670089722,
      "epoch": 1.7163461538461537,
      "grad_norm": 0.021752603352069855,
      "learning_rate": 1e-06,
      "loss": -0.004,
      "step": 1071
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.3046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.2312638759613037,
      "epoch": 1.717948717948718,
      "grad_norm": 0.016995754092931747,
      "learning_rate": 1e-06,
      "loss": 0.0081,
      "step": 1072
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3701.0,
      "completions/mean_length": 1393.966796875,
      "completions/mean_terminated_length": 1364.632080078125,
      "completions/min_length": 432.0,
      "completions/min_terminated_length": 432.0,
      "entropy": 0.23279593884944916,
      "epoch": 1.719551282051282,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 0.02495507150888443,
      "learning_rate": 1e-06,
      "loss": 0.0083,
      "num_tokens": 828169640.0,
      "reward": 0.2605937719345093,
      "reward_std": 0.053633954375982285,
      "rewards/progression_diversity/mean": -0.0006842931034043431,
      "rewards/progression_diversity/std": 0.006281269248574972,
      "rewards/symbolic_reward_accuracy/mean": 0.111328125,
      "rewards/symbolic_reward_accuracy/std": 0.31484565138816833,
      "rewards/symbolic_reward_partial_score/mean": 0.6466634273529053,
      "rewards/symbolic_reward_partial_score/std": 0.21451237797737122,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0482239723205566,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 149.9818115234375,
      "sampling/sampling_logp_difference/mean": 0.10541543364524841,
      "step": 1073
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2331898808479309,
      "epoch": 1.7211538461538463,
      "grad_norm": 0.02485121600329876,
      "learning_rate": 1e-06,
      "loss": 0.007,
      "step": 1074
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2273246869444847,
      "epoch": 1.7227564102564101,
      "grad_norm": 0.025766994804143906,
      "learning_rate": 1e-06,
      "loss": -0.0,
      "step": 1075
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2262198030948639,
      "epoch": 1.7243589743589745,
      "grad_norm": 70.85974884033203,
      "learning_rate": 1e-06,
      "loss": 0.013,
      "step": 1076
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 12415.0,
      "completions/mean_length": 1387.005859375,
      "completions/mean_terminated_length": 1298.614990234375,
      "completions/min_length": 303.0,
      "completions/min_terminated_length": 303.0,
      "entropy": 0.22221150994300842,
      "epoch": 1.7259615384615383,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.019594384357333183,
      "learning_rate": 1e-06,
      "loss": 0.0331,
      "num_tokens": 829798859.0,
      "reward": 0.3287256360054016,
      "reward_std": 0.0677950382232666,
      "rewards/progression_diversity/mean": -0.0004827457305509597,
      "rewards/progression_diversity/std": 0.005075244233012199,
      "rewards/symbolic_reward_accuracy/mean": 0.208984375,
      "rewards/symbolic_reward_accuracy/std": 0.40698084235191345,
      "rewards/symbolic_reward_partial_score/mean": 0.6797525882720947,
      "rewards/symbolic_reward_partial_score/std": 0.23522421717643738,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0458087921142578,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.522271990776062,
      "step": 1077
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2273356318473816,
      "epoch": 1.7275641025641026,
      "grad_norm": 0.018257969990372658,
      "learning_rate": 1e-06,
      "loss": -0.0013,
      "step": 1078
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.22496186941862106,
      "epoch": 1.7291666666666665,
      "grad_norm": 0.01670248992741108,
      "learning_rate": 1e-06,
      "loss": -0.002,
      "step": 1079
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2231815606355667,
      "epoch": 1.7307692307692308,
      "grad_norm": 0.021122116595506668,
      "learning_rate": 1e-06,
      "loss": 0.061,
      "step": 1080
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4567.0,
      "completions/max_terminated_length": 4567.0,
      "completions/mean_length": 1337.560546875,
      "completions/mean_terminated_length": 1337.560546875,
      "completions/min_length": 488.0,
      "completions/min_terminated_length": 488.0,
      "entropy": 0.2179974913597107,
      "epoch": 1.7323717948717947,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 0.02882332168519497,
      "learning_rate": 1e-06,
      "loss": 0.0089,
      "num_tokens": 831412682.0,
      "reward": 0.25469720363616943,
      "reward_std": 0.042546242475509644,
      "rewards/progression_diversity/mean": -0.0004942620871588588,
      "rewards/progression_diversity/std": 0.004545349627733231,
      "rewards/symbolic_reward_accuracy/mean": 0.12109375,
      "rewards/symbolic_reward_accuracy/std": 0.3265552520751953,
      "rewards/symbolic_reward_partial_score/mean": 0.6068196892738342,
      "rewards/symbolic_reward_partial_score/std": 0.2223115861415863,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0473459959030151,
      "sampling/importance_sampling_ratio/min": 2.902779897340224e-06,
      "sampling/sampling_logp_difference/max": 12.749841690063477,
      "sampling/sampling_logp_difference/mean": 0.09889820963144302,
      "step": 1081
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.22881120443344116,
      "epoch": 1.733974358974359,
      "grad_norm": 0.024050477892160416,
      "learning_rate": 1e-06,
      "loss": 0.0022,
      "step": 1082
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.22975780814886093,
      "epoch": 1.7355769230769231,
      "grad_norm": 0.014919934794306755,
      "learning_rate": 1e-06,
      "loss": 0.0094,
      "step": 1083
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.3125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.22842169553041458,
      "epoch": 1.7371794871794872,
      "grad_norm": 0.015330510213971138,
      "learning_rate": 1e-06,
      "loss": -0.0049,
      "step": 1084
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3636.0,
      "completions/max_terminated_length": 3636.0,
      "completions/mean_length": 1231.79296875,
      "completions/mean_terminated_length": 1231.79296875,
      "completions/min_length": 393.0,
      "completions/min_terminated_length": 393.0,
      "entropy": 0.23932421952486038,
      "epoch": 1.7387820512820513,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.02318231202661991,
      "learning_rate": 1e-06,
      "loss": -0.0044,
      "num_tokens": 832881792.0,
      "reward": 0.25613856315612793,
      "reward_std": 0.03916171193122864,
      "rewards/progression_diversity/mean": -0.0004032338911201805,
      "rewards/progression_diversity/std": 0.0032340919133275747,
      "rewards/symbolic_reward_accuracy/mean": 0.09765625,
      "rewards/symbolic_reward_accuracy/std": 0.29713961482048035,
      "rewards/symbolic_reward_partial_score/mean": 0.658496081829071,
      "rewards/symbolic_reward_partial_score/std": 0.20491082966327667,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.050029993057251,
      "sampling/importance_sampling_ratio/min": 0.0007102126837708056,
      "sampling/sampling_logp_difference/max": 7.249946117401123,
      "sampling/sampling_logp_difference/mean": 0.10211381316184998,
      "step": 1085
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.23092254996299744,
      "epoch": 1.7403846153846154,
      "grad_norm": 0.013799590058624744,
      "learning_rate": 1e-06,
      "loss": -0.0024,
      "step": 1086
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.23684141784906387,
      "epoch": 1.7419871794871795,
      "grad_norm": 0.018218791112303734,
      "learning_rate": 1e-06,
      "loss": 0.0023,
      "step": 1087
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2406981661915779,
      "epoch": 1.7435897435897436,
      "grad_norm": 0.012886385433375835,
      "learning_rate": 1e-06,
      "loss": 0.0037,
      "step": 1088
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4425.0,
      "completions/mean_length": 1411.501953125,
      "completions/mean_terminated_length": 1233.9625244140625,
      "completions/min_length": 299.0,
      "completions/min_terminated_length": 299.0,
      "entropy": 0.22861982136964798,
      "epoch": 1.7451923076923077,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 152.24722290039062,
      "learning_rate": 1e-06,
      "loss": 0.0197,
      "num_tokens": 834562321.0,
      "reward": 0.2795882225036621,
      "reward_std": 0.03494984656572342,
      "rewards/progression_diversity/mean": -0.0011390319559723139,
      "rewards/progression_diversity/std": 0.019081035628914833,
      "rewards/symbolic_reward_accuracy/mean": 0.154296875,
      "rewards/symbolic_reward_accuracy/std": 0.36158639192581177,
      "rewards/symbolic_reward_partial_score/mean": 0.6273112297058105,
      "rewards/symbolic_reward_partial_score/std": 0.21833449602127075,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0453113317489624,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 0.9983885884284973,
      "step": 1089
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2288360819220543,
      "epoch": 1.7467948717948718,
      "grad_norm": 0.01483655534684658,
      "learning_rate": 1e-06,
      "loss": 0.0598,
      "step": 1090
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.2326875552535057,
      "epoch": 1.748397435897436,
      "grad_norm": 136.22293090820312,
      "learning_rate": 1e-06,
      "loss": 0.0291,
      "step": 1091
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.233309805393219,
      "epoch": 1.75,
      "grad_norm": 0.023925775662064552,
      "learning_rate": 1e-06,
      "loss": 0.0074,
      "step": 1092
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3879.0,
      "completions/mean_length": 1282.169921875,
      "completions/mean_terminated_length": 1252.616455078125,
      "completions/min_length": 338.0,
      "completions/min_terminated_length": 338.0,
      "entropy": 0.23763851076364517,
      "epoch": 1.751602564102564,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 0.028567634522914886,
      "learning_rate": 1e-06,
      "loss": 0.0098,
      "num_tokens": 836112856.0,
      "reward": 0.2890186309814453,
      "reward_std": 0.06066850945353508,
      "rewards/progression_diversity/mean": -0.00048197241267189384,
      "rewards/progression_diversity/std": 0.0038850673008710146,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.6984374523162842,
      "rewards/symbolic_reward_partial_score/std": 0.21642756462097168,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0489928722381592,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.8276915550231934,
      "step": 1093
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23289386928081512,
      "epoch": 1.7532051282051282,
      "grad_norm": 0.015758074820041656,
      "learning_rate": 1e-06,
      "loss": -0.0056,
      "step": 1094
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.23799997568130493,
      "epoch": 1.7548076923076923,
      "grad_norm": 0.026696138083934784,
      "learning_rate": 1e-06,
      "loss": 0.0046,
      "step": 1095
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.2384975180029869,
      "epoch": 1.7564102564102564,
      "grad_norm": 0.02093910239636898,
      "learning_rate": 1e-06,
      "loss": 0.014,
      "step": 1096
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3248.0,
      "completions/max_terminated_length": 3248.0,
      "completions/mean_length": 1161.291015625,
      "completions/mean_terminated_length": 1161.291015625,
      "completions/min_length": 327.0,
      "completions/min_terminated_length": 327.0,
      "entropy": 0.23988765478134155,
      "epoch": 1.7580128205128205,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.025087056681513786,
      "learning_rate": 1e-06,
      "loss": 0.0066,
      "num_tokens": 837526381.0,
      "reward": 0.32568255066871643,
      "reward_std": 0.07381178438663483,
      "rewards/progression_diversity/mean": -0.00010529650171520188,
      "rewards/progression_diversity/std": 0.0017732558771967888,
      "rewards/symbolic_reward_accuracy/mean": 0.21484375,
      "rewards/symbolic_reward_accuracy/std": 0.4111155867576599,
      "rewards/symbolic_reward_partial_score/mean": 0.6559244990348816,
      "rewards/symbolic_reward_partial_score/std": 0.2370387762784958,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.051661729812622,
      "sampling/importance_sampling_ratio/min": 8.779976212736074e-08,
      "sampling/sampling_logp_difference/max": 16.248207092285156,
      "sampling/sampling_logp_difference/mean": 0.10443771630525589,
      "step": 1097
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.24236983805894852,
      "epoch": 1.7596153846153846,
      "grad_norm": 0.01639835350215435,
      "learning_rate": 1e-06,
      "loss": -0.0026,
      "step": 1098
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.23802829533815384,
      "epoch": 1.7612179487179487,
      "grad_norm": 0.01632961444556713,
      "learning_rate": 1e-06,
      "loss": -0.0023,
      "step": 1099
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.23437372595071793,
      "epoch": 1.7628205128205128,
      "grad_norm": 0.013542949222028255,
      "learning_rate": 1e-06,
      "loss": 0.0084,
      "step": 1100
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3300.0,
      "completions/mean_length": 1334.92578125,
      "completions/mean_terminated_length": 1275.909912109375,
      "completions/min_length": 345.0,
      "completions/min_terminated_length": 345.0,
      "entropy": 0.22916942834854126,
      "epoch": 1.7644230769230769,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.020054111257195473,
      "learning_rate": 1e-06,
      "loss": -0.0065,
      "num_tokens": 839122695.0,
      "reward": 0.2580447793006897,
      "reward_std": 0.033093225210905075,
      "rewards/progression_diversity/mean": -0.00021209442638792098,
      "rewards/progression_diversity/std": 0.0026572090573608875,
      "rewards/symbolic_reward_accuracy/mean": 0.107421875,
      "rewards/symbolic_reward_accuracy/std": 0.30995169281959534,
      "rewards/symbolic_reward_partial_score/mean": 0.6466145515441895,
      "rewards/symbolic_reward_partial_score/std": 0.1929846704006195,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0464985370635986,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.6539708971977234,
      "step": 1101
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.22496525943279266,
      "epoch": 1.766025641025641,
      "grad_norm": 698.2822265625,
      "learning_rate": 1e-06,
      "loss": 0.0585,
      "step": 1102
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.22553813457489014,
      "epoch": 1.7676282051282053,
      "grad_norm": 0.027719808742403984,
      "learning_rate": 1e-06,
      "loss": 0.0128,
      "step": 1103
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.2250175029039383,
      "epoch": 1.7692307692307692,
      "grad_norm": 0.015375426970422268,
      "learning_rate": 1e-06,
      "loss": 0.0297,
      "step": 1104
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3192.0,
      "completions/mean_length": 1293.3046875,
      "completions/mean_terminated_length": 1174.4803466796875,
      "completions/min_length": 362.0,
      "completions/min_terminated_length": 362.0,
      "entropy": 0.22259868681430817,
      "epoch": 1.7708333333333335,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 692.8626098632812,
      "learning_rate": 1e-06,
      "loss": 0.0203,
      "num_tokens": 840688755.0,
      "reward": 0.3711351156234741,
      "reward_std": 0.09630028158426285,
      "rewards/progression_diversity/mean": -0.001723826164379716,
      "rewards/progression_diversity/std": 0.026012076064944267,
      "rewards/symbolic_reward_accuracy/mean": 0.271484375,
      "rewards/symbolic_reward_accuracy/std": 0.44516023993492126,
      "rewards/symbolic_reward_partial_score/mean": 0.6968098878860474,
      "rewards/symbolic_reward_partial_score/std": 0.23859074711799622,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0450929403305054,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.8819183111190796,
      "step": 1105
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.22188594937324524,
      "epoch": 1.7724358974358974,
      "grad_norm": 0.017745744436979294,
      "learning_rate": 1e-06,
      "loss": -0.0115,
      "step": 1106
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22368201613426208,
      "epoch": 1.7740384615384617,
      "grad_norm": 0.01989768259227276,
      "learning_rate": 1e-06,
      "loss": 0.037,
      "step": 1107
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.22869247198104858,
      "epoch": 1.7756410256410255,
      "grad_norm": 0.013347904197871685,
      "learning_rate": 1e-06,
      "loss": 0.0261,
      "step": 1108
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3432.0,
      "completions/mean_length": 1230.427734375,
      "completions/mean_terminated_length": 1171.0020751953125,
      "completions/min_length": 256.0,
      "completions/min_terminated_length": 256.0,
      "entropy": 0.22080733627080917,
      "epoch": 1.7772435897435899,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.01532725803554058,
      "learning_rate": 1e-06,
      "loss": 0.0232,
      "num_tokens": 842301870.0,
      "reward": 0.2819012403488159,
      "reward_std": 0.061507582664489746,
      "rewards/progression_diversity/mean": -0.0003044310724362731,
      "rewards/progression_diversity/std": 0.005076899658888578,
      "rewards/symbolic_reward_accuracy/mean": 0.142578125,
      "rewards/symbolic_reward_accuracy/std": 0.3499840497970581,
      "rewards/symbolic_reward_partial_score/mean": 0.6558268070220947,
      "rewards/symbolic_reward_partial_score/std": 0.21851463615894318,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0459144115447998,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.5158374905586243,
      "step": 1109
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2195032313466072,
      "epoch": 1.7788461538461537,
      "grad_norm": 0.027773037552833557,
      "learning_rate": 1e-06,
      "loss": -0.0005,
      "step": 1110
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.2160264179110527,
      "epoch": 1.780448717948718,
      "grad_norm": 0.016807833686470985,
      "learning_rate": 1e-06,
      "loss": 0.0238,
      "step": 1111
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21882987767457962,
      "epoch": 1.782051282051282,
      "grad_norm": 0.017015958204865456,
      "learning_rate": 1e-06,
      "loss": -0.0025,
      "step": 1112
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3504.0,
      "completions/mean_length": 1344.291015625,
      "completions/mean_terminated_length": 1285.3118896484375,
      "completions/min_length": 361.0,
      "completions/min_terminated_length": 361.0,
      "entropy": 0.22031641751527786,
      "epoch": 1.7836538461538463,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.02029530704021454,
      "learning_rate": 1e-06,
      "loss": -0.0052,
      "num_tokens": 843814051.0,
      "reward": 0.3037402629852295,
      "reward_std": 0.06527281552553177,
      "rewards/progression_diversity/mean": -1.7595344559140358e-07,
      "rewards/progression_diversity/std": 3.981372174166609e-06,
      "rewards/symbolic_reward_accuracy/mean": 0.1640625,
      "rewards/symbolic_reward_accuracy/std": 0.37069445848464966,
      "rewards/symbolic_reward_partial_score/mean": 0.6856445074081421,
      "rewards/symbolic_reward_partial_score/std": 0.2108502984046936,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0456528663635254,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.39048463106155396,
      "step": 1113
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21349480748176575,
      "epoch": 1.7852564102564101,
      "grad_norm": 0.019900523126125336,
      "learning_rate": 1e-06,
      "loss": 0.0008,
      "step": 1114
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.21848982572555542,
      "epoch": 1.7868589743589745,
      "grad_norm": 0.010702534578740597,
      "learning_rate": 1e-06,
      "loss": 0.0025,
      "step": 1115
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.22711623460054398,
      "epoch": 1.7884615384615383,
      "grad_norm": 0.021489204838871956,
      "learning_rate": 1e-06,
      "loss": 0.0038,
      "step": 1116
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3124.0,
      "completions/max_terminated_length": 3124.0,
      "completions/mean_length": 1099.703125,
      "completions/mean_terminated_length": 1099.703125,
      "completions/min_length": 324.0,
      "completions/min_terminated_length": 324.0,
      "entropy": 0.2246062159538269,
      "epoch": 1.7900641025641026,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.01770607940852642,
      "learning_rate": 1e-06,
      "loss": 0.0047,
      "num_tokens": 845220171.0,
      "reward": 0.3277079463005066,
      "reward_std": 0.05349338799715042,
      "rewards/progression_diversity/mean": -0.00020373196457512677,
      "rewards/progression_diversity/std": 0.002731936750933528,
      "rewards/symbolic_reward_accuracy/mean": 0.1953125,
      "rewards/symbolic_reward_accuracy/std": 0.3968288004398346,
      "rewards/symbolic_reward_partial_score/mean": 0.7017415761947632,
      "rewards/symbolic_reward_partial_score/std": 0.196936696767807,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0474233627319336,
      "sampling/importance_sampling_ratio/min": 1.3281011845833746e-08,
      "sampling/sampling_logp_difference/max": 18.136930465698242,
      "sampling/sampling_logp_difference/mean": 0.09797333925962448,
      "step": 1117
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21827614307403564,
      "epoch": 1.7916666666666665,
      "grad_norm": 0.011651732958853245,
      "learning_rate": 1e-06,
      "loss": 0.0005,
      "step": 1118
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.21932896971702576,
      "epoch": 1.7932692307692308,
      "grad_norm": 0.010273098014295101,
      "learning_rate": 1e-06,
      "loss": -0.0024,
      "step": 1119
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.2230270653963089,
      "epoch": 1.7948717948717947,
      "grad_norm": 0.02401752397418022,
      "learning_rate": 1e-06,
      "loss": 0.0044,
      "step": 1120
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3138.0,
      "completions/max_terminated_length": 3138.0,
      "completions/mean_length": 1054.958984375,
      "completions/mean_terminated_length": 1054.958984375,
      "completions/min_length": 302.0,
      "completions/min_terminated_length": 302.0,
      "entropy": 0.21830987185239792,
      "epoch": 1.796474358974359,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.015644196420907974,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "num_tokens": 846662438.0,
      "reward": 0.3441348671913147,
      "reward_std": 0.06147376075387001,
      "rewards/progression_diversity/mean": -8.792350126896054e-05,
      "rewards/progression_diversity/std": 0.00172401312738657,
      "rewards/symbolic_reward_accuracy/mean": 0.216796875,
      "rewards/symbolic_reward_accuracy/std": 0.4124660789966583,
      "rewards/symbolic_reward_partial_score/mean": 0.7135253548622131,
      "rewards/symbolic_reward_partial_score/std": 0.2144842892885208,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0474189519882202,
      "sampling/importance_sampling_ratio/min": 4.838908353121951e-05,
      "sampling/sampling_logp_difference/max": 9.936236381530762,
      "sampling/sampling_logp_difference/mean": 0.0984996110200882,
      "step": 1121
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.22250553965568542,
      "epoch": 1.7980769230769231,
      "grad_norm": 0.020056243985891342,
      "learning_rate": 1e-06,
      "loss": 0.007,
      "step": 1122
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.22270743548870087,
      "epoch": 1.7996794871794872,
      "grad_norm": 0.018208522349596024,
      "learning_rate": 1e-06,
      "loss": 0.0082,
      "step": 1123
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.22166861593723297,
      "epoch": 1.8012820512820513,
      "grad_norm": 0.014248647727072239,
      "learning_rate": 1e-06,
      "loss": -0.0053,
      "step": 1124
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 5645.0,
      "completions/max_terminated_length": 5645.0,
      "completions/mean_length": 1187.763671875,
      "completions/mean_terminated_length": 1187.763671875,
      "completions/min_length": 364.0,
      "completions/min_terminated_length": 364.0,
      "entropy": 0.21534562855958939,
      "epoch": 1.8028846153846154,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.024967024102807045,
      "learning_rate": 1e-06,
      "loss": 0.0209,
      "num_tokens": 848091485.0,
      "reward": 0.4134520888328552,
      "reward_std": 0.04205480217933655,
      "rewards/progression_diversity/mean": -0.0004985664272680879,
      "rewards/progression_diversity/std": 0.007882904261350632,
      "rewards/symbolic_reward_accuracy/mean": 0.30078125,
      "rewards/symbolic_reward_accuracy/std": 0.45904624462127686,
      "rewards/symbolic_reward_partial_score/mean": 0.7766276001930237,
      "rewards/symbolic_reward_partial_score/std": 0.2088257223367691,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0464260578155518,
      "sampling/importance_sampling_ratio/min": 1.6780650184955448e-05,
      "sampling/sampling_logp_difference/max": 10.995284080505371,
      "sampling/sampling_logp_difference/mean": 0.09675593674182892,
      "step": 1125
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.22125839442014694,
      "epoch": 1.8044871794871795,
      "grad_norm": 0.012069555930793285,
      "learning_rate": 1e-06,
      "loss": -0.0023,
      "step": 1126
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.22001603245735168,
      "epoch": 1.8060897435897436,
      "grad_norm": 0.014061706140637398,
      "learning_rate": 1e-06,
      "loss": -0.0032,
      "step": 1127
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.22054609656333923,
      "epoch": 1.8076923076923077,
      "grad_norm": 0.011219343170523643,
      "learning_rate": 1e-06,
      "loss": -0.0031,
      "step": 1128
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 8772.0,
      "completions/max_terminated_length": 8772.0,
      "completions/mean_length": 1137.666015625,
      "completions/mean_terminated_length": 1137.666015625,
      "completions/min_length": 287.0,
      "completions/min_terminated_length": 287.0,
      "entropy": 0.21648870408535004,
      "epoch": 1.8092948717948718,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.022340567782521248,
      "learning_rate": 1e-06,
      "loss": 0.005,
      "num_tokens": 849531970.0,
      "reward": 0.3116188943386078,
      "reward_std": 0.024817654862999916,
      "rewards/progression_diversity/mean": -0.00022150327276904136,
      "rewards/progression_diversity/std": 0.004627527203410864,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.6793619394302368,
      "rewards/symbolic_reward_partial_score/std": 0.1976500004529953,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0465974807739258,
      "sampling/importance_sampling_ratio/min": 5.26275156570658e-19,
      "sampling/sampling_logp_difference/max": 42.088462829589844,
      "sampling/sampling_logp_difference/mean": 0.09620961546897888,
      "step": 1129
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.03125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.21649357676506042,
      "epoch": 1.810897435897436,
      "grad_norm": 0.01756151393055916,
      "learning_rate": 1e-06,
      "loss": 0.0014,
      "step": 1130
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21943299472332,
      "epoch": 1.8125,
      "grad_norm": 0.015073378570377827,
      "learning_rate": 1e-06,
      "loss": -0.007,
      "step": 1131
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2232174053788185,
      "epoch": 1.814102564102564,
      "grad_norm": 0.01907176896929741,
      "learning_rate": 1e-06,
      "loss": 0.0046,
      "step": 1132
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3176.0,
      "completions/max_terminated_length": 3176.0,
      "completions/mean_length": 1107.029296875,
      "completions/mean_terminated_length": 1107.029296875,
      "completions/min_length": 350.0,
      "completions/min_terminated_length": 350.0,
      "entropy": 0.2129407450556755,
      "epoch": 1.8157051282051282,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.025003377348184586,
      "learning_rate": 1e-06,
      "loss": 0.0111,
      "num_tokens": 850939665.0,
      "reward": 0.28454291820526123,
      "reward_std": 0.03294537216424942,
      "rewards/progression_diversity/mean": -0.00029976246878504753,
      "rewards/progression_diversity/std": 0.002749260514974594,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.682861328125,
      "rewards/symbolic_reward_partial_score/std": 0.1956394761800766,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0451654195785522,
      "sampling/importance_sampling_ratio/min": 8.991118960466338e-08,
      "sampling/sampling_logp_difference/max": 16.224443435668945,
      "sampling/sampling_logp_difference/mean": 0.09530235081911087,
      "step": 1133
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.21520614624023438,
      "epoch": 1.8173076923076923,
      "grad_norm": 0.008025935851037502,
      "learning_rate": 1e-06,
      "loss": -0.004,
      "step": 1134
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.21551941335201263,
      "epoch": 1.8189102564102564,
      "grad_norm": 0.021345248445868492,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "step": 1135
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.21580560505390167,
      "epoch": 1.8205128205128205,
      "grad_norm": 0.0155746815726161,
      "learning_rate": 1e-06,
      "loss": 0.0006,
      "step": 1136
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3362.0,
      "completions/max_terminated_length": 3362.0,
      "completions/mean_length": 1122.947265625,
      "completions/mean_terminated_length": 1122.947265625,
      "completions/min_length": 323.0,
      "completions/min_terminated_length": 323.0,
      "entropy": 0.21716423332691193,
      "epoch": 1.8221153846153846,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 0.02500241994857788,
      "learning_rate": 1e-06,
      "loss": -0.0001,
      "num_tokens": 852408278.0,
      "reward": 0.28805452585220337,
      "reward_std": 0.041980091482400894,
      "rewards/progression_diversity/mean": -0.00020892325846944004,
      "rewards/progression_diversity/std": 0.001957833534106612,
      "rewards/symbolic_reward_accuracy/mean": 0.13671875,
      "rewards/symbolic_reward_accuracy/std": 0.3438861668109894,
      "rewards/symbolic_reward_partial_score/mean": 0.6867513656616211,
      "rewards/symbolic_reward_partial_score/std": 0.19679629802703857,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045684576034546,
      "sampling/importance_sampling_ratio/min": 4.51097449324922e-12,
      "sampling/sampling_logp_difference/max": 26.124507904052734,
      "sampling/sampling_logp_difference/mean": 0.09778503328561783,
      "step": 1137
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.21945731341838837,
      "epoch": 1.8237179487179487,
      "grad_norm": 0.013288813643157482,
      "learning_rate": 1e-06,
      "loss": -0.0022,
      "step": 1138
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.22035419940948486,
      "epoch": 1.8253205128205128,
      "grad_norm": 0.025754936039447784,
      "learning_rate": 1e-06,
      "loss": 0.0074,
      "step": 1139
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.22086691856384277,
      "epoch": 1.8269230769230769,
      "grad_norm": 0.018110549077391624,
      "learning_rate": 1e-06,
      "loss": 0.0034,
      "step": 1140
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3428.0,
      "completions/mean_length": 1185.529296875,
      "completions/mean_terminated_length": 1155.7867431640625,
      "completions/min_length": 354.0,
      "completions/min_terminated_length": 354.0,
      "entropy": 0.22858010977506638,
      "epoch": 1.828525641025641,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.029191119596362114,
      "learning_rate": 1e-06,
      "loss": 0.0041,
      "num_tokens": 853835237.0,
      "reward": 0.38153761625289917,
      "reward_std": 0.07013949751853943,
      "rewards/progression_diversity/mean": -4.912294389214367e-05,
      "rewards/progression_diversity/std": 0.0007894500158727169,
      "rewards/symbolic_reward_accuracy/mean": 0.275390625,
      "rewards/symbolic_reward_accuracy/std": 0.44714778661727905,
      "rewards/symbolic_reward_partial_score/mean": 0.7216634154319763,
      "rewards/symbolic_reward_partial_score/std": 0.22831861674785614,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.046138048171997,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 472.0,
      "sampling/sampling_logp_difference/mean": 0.10319283604621887,
      "step": 1141
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2211167812347412,
      "epoch": 1.8301282051282053,
      "grad_norm": 0.02721370942890644,
      "learning_rate": 1e-06,
      "loss": 0.0007,
      "step": 1142
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2238641008734703,
      "epoch": 1.8317307692307692,
      "grad_norm": 0.02170705609023571,
      "learning_rate": 1e-06,
      "loss": 0.0196,
      "step": 1143
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.23243413865566254,
      "epoch": 1.8333333333333335,
      "grad_norm": 0.014753867872059345,
      "learning_rate": 1e-06,
      "loss": -0.0014,
      "step": 1144
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3178.0,
      "completions/mean_length": 1113.748046875,
      "completions/mean_terminated_length": 1083.864990234375,
      "completions/min_length": 354.0,
      "completions/min_terminated_length": 354.0,
      "entropy": 0.22820258885622025,
      "epoch": 1.8349358974358974,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.017704647034406662,
      "learning_rate": 1e-06,
      "loss": -0.0001,
      "num_tokens": 855269764.0,
      "reward": 0.2694726586341858,
      "reward_std": 0.026778917759656906,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.126953125,
      "rewards/symbolic_reward_accuracy/std": 0.33324605226516724,
      "rewards/symbolic_reward_partial_score/mean": 0.6449869871139526,
      "rewards/symbolic_reward_partial_score/std": 0.22033336758613586,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0469450950622559,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 720.0,
      "sampling/sampling_logp_difference/mean": 0.3227856755256653,
      "step": 1145
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.22650671005249023,
      "epoch": 1.8365384615384617,
      "grad_norm": 199.25045776367188,
      "learning_rate": 1e-06,
      "loss": 0.0266,
      "step": 1146
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22604414075613022,
      "epoch": 1.8381410256410255,
      "grad_norm": 0.009444105438888073,
      "learning_rate": 1e-06,
      "loss": -0.0039,
      "step": 1147
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.22440266609191895,
      "epoch": 1.8397435897435899,
      "grad_norm": 0.01778390072286129,
      "learning_rate": 1e-06,
      "loss": -0.0054,
      "step": 1148
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3424.0,
      "completions/mean_length": 1209.474609375,
      "completions/mean_terminated_length": 1120.037353515625,
      "completions/min_length": 283.0,
      "completions/min_terminated_length": 283.0,
      "entropy": 0.22518840432167053,
      "epoch": 1.8413461538461537,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 323.1822814941406,
      "learning_rate": 1e-06,
      "loss": 0.0188,
      "num_tokens": 856870839.0,
      "reward": 0.28326651453971863,
      "reward_std": 0.05484289303421974,
      "rewards/progression_diversity/mean": -0.0004978624056093395,
      "rewards/progression_diversity/std": 0.009513349272310734,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.6402018070220947,
      "rewards/symbolic_reward_partial_score/std": 0.2216087281703949,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0432510375976562,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.484938621520996,
      "step": 1149
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.21288639307022095,
      "epoch": 1.842948717948718,
      "grad_norm": 1597.748046875,
      "learning_rate": 1e-06,
      "loss": 0.155,
      "step": 1150
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22490354627370834,
      "epoch": 1.844551282051282,
      "grad_norm": 0.015957126393914223,
      "learning_rate": 1e-06,
      "loss": -0.004,
      "step": 1151
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22101520746946335,
      "epoch": 1.8461538461538463,
      "grad_norm": 0.017633656039834023,
      "learning_rate": 1e-06,
      "loss": 0.0068,
      "step": 1152
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3059.0,
      "completions/max_terminated_length": 3059.0,
      "completions/mean_length": 1138.349609375,
      "completions/mean_terminated_length": 1138.349609375,
      "completions/min_length": 356.0,
      "completions/min_terminated_length": 356.0,
      "entropy": 0.23107396066188812,
      "epoch": 1.8477564102564101,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.022070132195949554,
      "learning_rate": 1e-06,
      "loss": 0.0089,
      "num_tokens": 858210874.0,
      "reward": 0.38855424523353577,
      "reward_std": 0.0762196034193039,
      "rewards/progression_diversity/mean": -4.658957914216444e-05,
      "rewards/progression_diversity/std": 0.0010542018571868539,
      "rewards/symbolic_reward_accuracy/mean": 0.267578125,
      "rewards/symbolic_reward_accuracy/std": 0.4431293308734894,
      "rewards/symbolic_reward_partial_score/mean": 0.7600260972976685,
      "rewards/symbolic_reward_partial_score/std": 0.22460278868675232,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0477694272994995,
      "sampling/importance_sampling_ratio/min": 7.687971810810268e-05,
      "sampling/sampling_logp_difference/max": 9.473268508911133,
      "sampling/sampling_logp_difference/mean": 0.10008453577756882,
      "step": 1153
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22674327343702316,
      "epoch": 1.8493589743589745,
      "grad_norm": 0.015364538878202438,
      "learning_rate": 1e-06,
      "loss": 0.0082,
      "step": 1154
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.22742107510566711,
      "epoch": 1.8509615384615383,
      "grad_norm": 0.012995360419154167,
      "learning_rate": 1e-06,
      "loss": 0.0062,
      "step": 1155
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.22633031755685806,
      "epoch": 1.8525641025641026,
      "grad_norm": 0.015840977430343628,
      "learning_rate": 1e-06,
      "loss": -0.0093,
      "step": 1156
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3445.0,
      "completions/mean_length": 1179.333984375,
      "completions/mean_terminated_length": 1119.7078857421875,
      "completions/min_length": 349.0,
      "completions/min_terminated_length": 349.0,
      "entropy": 0.22518686205148697,
      "epoch": 1.8541666666666665,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.025853393599390984,
      "learning_rate": 1e-06,
      "loss": 0.0031,
      "num_tokens": 859689013.0,
      "reward": 0.35142046213150024,
      "reward_std": 0.06578496098518372,
      "rewards/progression_diversity/mean": -0.0005352114676497877,
      "rewards/progression_diversity/std": 0.010669847950339317,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.7508463263511658,
      "rewards/symbolic_reward_partial_score/std": 0.1848256140947342,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0448198318481445,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.6326540112495422,
      "step": 1157
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.21813210099935532,
      "epoch": 1.8557692307692308,
      "grad_norm": 0.018281714990735054,
      "learning_rate": 1e-06,
      "loss": 0.0262,
      "step": 1158
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.22226067632436752,
      "epoch": 1.8573717948717947,
      "grad_norm": 0.018198495730757713,
      "learning_rate": 1e-06,
      "loss": 0.0277,
      "step": 1159
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.22661340981721878,
      "epoch": 1.858974358974359,
      "grad_norm": 0.010464332066476345,
      "learning_rate": 1e-06,
      "loss": 0.0035,
      "step": 1160
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3748.0,
      "completions/mean_length": 1216.416015625,
      "completions/mean_terminated_length": 1186.73388671875,
      "completions/min_length": 417.0,
      "completions/min_terminated_length": 417.0,
      "entropy": 0.22897808998823166,
      "epoch": 1.8605769230769231,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.018317364156246185,
      "learning_rate": 1e-06,
      "loss": 0.0018,
      "num_tokens": 861117626.0,
      "reward": 0.27255719900131226,
      "reward_std": 0.025438256561756134,
      "rewards/progression_diversity/mean": -0.00014033068146090955,
      "rewards/progression_diversity/std": 0.002014033030718565,
      "rewards/symbolic_reward_accuracy/mean": 0.12890625,
      "rewards/symbolic_reward_accuracy/std": 0.33542385697364807,
      "rewards/symbolic_reward_partial_score/mean": 0.6513671875,
      "rewards/symbolic_reward_partial_score/std": 0.2093709260225296,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0455472469329834,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 0.44371387362480164,
      "step": 1161
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.22330108284950256,
      "epoch": 1.8621794871794872,
      "grad_norm": 0.010753365233540535,
      "learning_rate": 1e-06,
      "loss": 0.0224,
      "step": 1162
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.22619932144880295,
      "epoch": 1.8637820512820513,
      "grad_norm": 0.012639653868973255,
      "learning_rate": 1e-06,
      "loss": -0.0069,
      "step": 1163
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.2241613268852234,
      "epoch": 1.8653846153846154,
      "grad_norm": 0.02455468475818634,
      "learning_rate": 1e-06,
      "loss": 0.0086,
      "step": 1164
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3400.0,
      "completions/max_terminated_length": 3400.0,
      "completions/mean_length": 1240.51171875,
      "completions/mean_terminated_length": 1240.51171875,
      "completions/min_length": 299.0,
      "completions/min_terminated_length": 299.0,
      "entropy": 0.23062631487846375,
      "epoch": 1.8669871794871795,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.010606328025460243,
      "learning_rate": 1e-06,
      "loss": -0.0041,
      "num_tokens": 862543008.0,
      "reward": 0.3666699528694153,
      "reward_std": 0.06977473199367523,
      "rewards/progression_diversity/mean": -1.885928213596344e-07,
      "rewards/progression_diversity/std": 4.2673682401073165e-06,
      "rewards/symbolic_reward_accuracy/mean": 0.251953125,
      "rewards/symbolic_reward_accuracy/std": 0.43455907702445984,
      "rewards/symbolic_reward_partial_score/mean": 0.7183268070220947,
      "rewards/symbolic_reward_partial_score/std": 0.23471878468990326,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.047785997390747,
      "sampling/importance_sampling_ratio/min": 6.825915477293165e-08,
      "sampling/sampling_logp_difference/max": 16.499954223632812,
      "sampling/sampling_logp_difference/mean": 0.09932725131511688,
      "step": 1165
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.224430114030838,
      "epoch": 1.8685897435897436,
      "grad_norm": 0.021581292152404785,
      "learning_rate": 1e-06,
      "loss": -0.0059,
      "step": 1166
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22753822803497314,
      "epoch": 1.8701923076923077,
      "grad_norm": 0.014603802002966404,
      "learning_rate": 1e-06,
      "loss": 0.0018,
      "step": 1167
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.22805190086364746,
      "epoch": 1.8717948717948718,
      "grad_norm": 0.010613877326250076,
      "learning_rate": 1e-06,
      "loss": 0.0065,
      "step": 1168
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3325.0,
      "completions/mean_length": 1174.26953125,
      "completions/mean_terminated_length": 1144.5048828125,
      "completions/min_length": 268.0,
      "completions/min_terminated_length": 268.0,
      "entropy": 0.2323755919933319,
      "epoch": 1.873397435897436,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.02169196680188179,
      "learning_rate": 1e-06,
      "loss": 0.0052,
      "num_tokens": 863941530.0,
      "reward": 0.3810667097568512,
      "reward_std": 0.07798746973276138,
      "rewards/progression_diversity/mean": -0.0002637706929817796,
      "rewards/progression_diversity/std": 0.003273366717621684,
      "rewards/symbolic_reward_accuracy/mean": 0.279296875,
      "rewards/symbolic_reward_accuracy/std": 0.44909247756004333,
      "rewards/symbolic_reward_partial_score/mean": 0.7122883796691895,
      "rewards/symbolic_reward_partial_score/std": 0.2458869218826294,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0482022762298584,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 712.0,
      "sampling/sampling_logp_difference/mean": 0.4974533021450043,
      "step": 1169
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.23275116831064224,
      "epoch": 1.875,
      "grad_norm": 0.011001808568835258,
      "learning_rate": 1e-06,
      "loss": 0.0016,
      "step": 1170
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.23044289648532867,
      "epoch": 1.876602564102564,
      "grad_norm": 0.016635850071907043,
      "learning_rate": 1e-06,
      "loss": -0.0069,
      "step": 1171
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.23118175566196442,
      "epoch": 1.8782051282051282,
      "grad_norm": 96598.3203125,
      "learning_rate": 1e-06,
      "loss": 4.3159,
      "step": 1172
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4005.0,
      "completions/mean_length": 1170.333984375,
      "completions/mean_terminated_length": 1140.5616455078125,
      "completions/min_length": 369.0,
      "completions/min_terminated_length": 369.0,
      "entropy": 0.22949066758155823,
      "epoch": 1.8798076923076923,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.020497862249612808,
      "learning_rate": 1e-06,
      "loss": -0.0043,
      "num_tokens": 865447381.0,
      "reward": 0.44851863384246826,
      "reward_std": 0.04487219825387001,
      "rewards/progression_diversity/mean": -0.00019038662139791995,
      "rewards/progression_diversity/std": 0.0023014748003333807,
      "rewards/symbolic_reward_accuracy/mean": 0.357421875,
      "rewards/symbolic_reward_accuracy/std": 0.4797092080116272,
      "rewards/symbolic_reward_partial_score/mean": 0.7808756828308105,
      "rewards/symbolic_reward_partial_score/std": 0.20934894680976868,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0473530292510986,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.7223914265632629,
      "step": 1173
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.23133239150047302,
      "epoch": 1.8814102564102564,
      "grad_norm": 0.010313909500837326,
      "learning_rate": 1e-06,
      "loss": -0.0036,
      "step": 1174
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.23350611329078674,
      "epoch": 1.8830128205128205,
      "grad_norm": 0.019430264830589294,
      "learning_rate": 1e-06,
      "loss": 0.0015,
      "step": 1175
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2238868772983551,
      "epoch": 1.8846153846153846,
      "grad_norm": 0.013273504562675953,
      "learning_rate": 1e-06,
      "loss": 0.0231,
      "step": 1176
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3514.0,
      "completions/mean_length": 1243.806640625,
      "completions/mean_terminated_length": 1154.57177734375,
      "completions/min_length": 389.0,
      "completions/min_terminated_length": 389.0,
      "entropy": 0.2321956604719162,
      "epoch": 1.8862179487179487,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.014576968736946583,
      "learning_rate": 1e-06,
      "loss": 0.006,
      "num_tokens": 866961778.0,
      "reward": 0.20335838198661804,
      "reward_std": 0.03265673667192459,
      "rewards/progression_diversity/mean": -0.0005896041984669864,
      "rewards/progression_diversity/std": 0.008867304772138596,
      "rewards/symbolic_reward_accuracy/mean": 0.03125,
      "rewards/symbolic_reward_accuracy/std": 0.17416280508041382,
      "rewards/symbolic_reward_partial_score/mean": 0.6173340082168579,
      "rewards/symbolic_reward_partial_score/std": 0.1687249392271042,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0458412170410156,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.8149086236953735,
      "step": 1177
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.22551760077476501,
      "epoch": 1.8878205128205128,
      "grad_norm": 0.017989974468946457,
      "learning_rate": 1e-06,
      "loss": 0.0529,
      "step": 1178
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.2227119281888008,
      "epoch": 1.8894230769230769,
      "grad_norm": 0.02180834859609604,
      "learning_rate": 1e-06,
      "loss": 0.005,
      "step": 1179
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.22573918104171753,
      "epoch": 1.891025641025641,
      "grad_norm": 126.00947570800781,
      "learning_rate": 1e-06,
      "loss": 0.0381,
      "step": 1180
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3086.0,
      "completions/mean_length": 1103.10546875,
      "completions/mean_terminated_length": 1073.2015380859375,
      "completions/min_length": 345.0,
      "completions/min_terminated_length": 345.0,
      "entropy": 0.22882883995771408,
      "epoch": 1.8926282051282053,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.019178207963705063,
      "learning_rate": 1e-06,
      "loss": -0.0016,
      "num_tokens": 868483272.0,
      "reward": 0.3014499545097351,
      "reward_std": 0.037228312343358994,
      "rewards/progression_diversity/mean": -0.001001341617666185,
      "rewards/progression_diversity/std": 0.01981303095817566,
      "rewards/symbolic_reward_accuracy/mean": 0.171875,
      "rewards/symbolic_reward_accuracy/std": 0.3776407241821289,
      "rewards/symbolic_reward_partial_score/mean": 0.6617676019668579,
      "rewards/symbolic_reward_partial_score/std": 0.21480776369571686,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0480608940124512,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.56421959400177,
      "step": 1181
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.22795479744672775,
      "epoch": 1.8942307692307692,
      "grad_norm": 0.011316453106701374,
      "learning_rate": 1e-06,
      "loss": -0.0061,
      "step": 1182
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.22722984105348587,
      "epoch": 1.8958333333333335,
      "grad_norm": 0.016038935631513596,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "step": 1183
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.22345466166734695,
      "epoch": 1.8974358974358974,
      "grad_norm": 0.0241796113550663,
      "learning_rate": 1e-06,
      "loss": 0.0253,
      "step": 1184
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3469.0,
      "completions/max_terminated_length": 3469.0,
      "completions/mean_length": 1094.568359375,
      "completions/mean_terminated_length": 1094.568359375,
      "completions/min_length": 364.0,
      "completions/min_terminated_length": 364.0,
      "entropy": 0.23068898171186447,
      "epoch": 1.8990384615384617,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.025562407448887825,
      "learning_rate": 1e-06,
      "loss": 0.0018,
      "num_tokens": 869886539.0,
      "reward": 0.2939690053462982,
      "reward_std": 0.05816289782524109,
      "rewards/progression_diversity/mean": -7.342880417127162e-05,
      "rewards/progression_diversity/std": 0.0009783387649804354,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.7142741084098816,
      "rewards/symbolic_reward_partial_score/std": 0.16729366779327393,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0498249530792236,
      "sampling/importance_sampling_ratio/min": 0.002071201568469405,
      "sampling/sampling_logp_difference/max": 6.17962646484375,
      "sampling/sampling_logp_difference/mean": 0.10211198776960373,
      "step": 1185
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.2281019389629364,
      "epoch": 1.9006410256410255,
      "grad_norm": 0.014062655158340931,
      "learning_rate": 1e-06,
      "loss": -0.0028,
      "step": 1186
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0390625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.22974391281604767,
      "epoch": 1.9022435897435899,
      "grad_norm": 0.014239110052585602,
      "learning_rate": 1e-06,
      "loss": 0.0121,
      "step": 1187
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.22985130548477173,
      "epoch": 1.9038461538461537,
      "grad_norm": 0.015234572812914848,
      "learning_rate": 1e-06,
      "loss": -0.0055,
      "step": 1188
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3333.0,
      "completions/max_terminated_length": 3333.0,
      "completions/mean_length": 1139.83984375,
      "completions/mean_terminated_length": 1139.83984375,
      "completions/min_length": 372.0,
      "completions/min_terminated_length": 372.0,
      "entropy": 0.22518738359212875,
      "epoch": 1.905448717948718,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.030008932575583458,
      "learning_rate": 1e-06,
      "loss": 0.0136,
      "num_tokens": 871431865.0,
      "reward": 0.3191036581993103,
      "reward_std": 0.04677165299654007,
      "rewards/progression_diversity/mean": -0.0002812910242937505,
      "rewards/progression_diversity/std": 0.0030721058137714863,
      "rewards/symbolic_reward_accuracy/mean": 0.193359375,
      "rewards/symbolic_reward_accuracy/std": 0.39531853795051575,
      "rewards/symbolic_reward_partial_score/mean": 0.6769694089889526,
      "rewards/symbolic_reward_partial_score/std": 0.21019434928894043,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.048149585723877,
      "sampling/importance_sampling_ratio/min": 7.317296608670226e-17,
      "sampling/sampling_logp_difference/max": 37.15370559692383,
      "sampling/sampling_logp_difference/mean": 0.09954197704792023,
      "step": 1189
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.22382111847400665,
      "epoch": 1.907051282051282,
      "grad_norm": 0.015485767275094986,
      "learning_rate": 1e-06,
      "loss": 0.0095,
      "step": 1190
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22165128588676453,
      "epoch": 1.9086538461538463,
      "grad_norm": 0.019756661728024483,
      "learning_rate": 1e-06,
      "loss": -0.0091,
      "step": 1191
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.22612737119197845,
      "epoch": 1.9102564102564101,
      "grad_norm": 0.020691078156232834,
      "learning_rate": 1e-06,
      "loss": -0.0043,
      "step": 1192
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3072.0,
      "completions/mean_length": 1140.0625,
      "completions/mean_terminated_length": 1140.0625,
      "completions/min_length": 410.0,
      "completions/min_terminated_length": 410.0,
      "entropy": 0.22674325108528137,
      "epoch": 1.9118589743589745,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.016689619049429893,
      "learning_rate": 1e-06,
      "loss": -0.0015,
      "num_tokens": 872902937.0,
      "reward": 0.26588261127471924,
      "reward_std": 0.033439576625823975,
      "rewards/progression_diversity/mean": -0.0001180602703243494,
      "rewards/progression_diversity/std": 0.0016232561320066452,
      "rewards/symbolic_reward_accuracy/mean": 0.12109375,
      "rewards/symbolic_reward_accuracy/std": 0.3265552520751953,
      "rewards/symbolic_reward_partial_score/mean": 0.644091784954071,
      "rewards/symbolic_reward_partial_score/std": 0.22011272609233856,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.048423171043396,
      "sampling/importance_sampling_ratio/min": 0.0005005777347832918,
      "sampling/sampling_logp_difference/max": 7.599747657775879,
      "sampling/sampling_logp_difference/mean": 0.10049919784069061,
      "step": 1193
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22232681512832642,
      "epoch": 1.9134615384615383,
      "grad_norm": 0.019354144111275673,
      "learning_rate": 1e-06,
      "loss": -0.0067,
      "step": 1194
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.22689828276634216,
      "epoch": 1.9150641025641026,
      "grad_norm": 0.015032010152935982,
      "learning_rate": 1e-06,
      "loss": 0.0079,
      "step": 1195
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.2218078002333641,
      "epoch": 1.9166666666666665,
      "grad_norm": 0.015426430851221085,
      "learning_rate": 1e-06,
      "loss": -0.0044,
      "step": 1196
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3017.0,
      "completions/mean_length": 1175.25,
      "completions/mean_terminated_length": 1115.60791015625,
      "completions/min_length": 349.0,
      "completions/min_terminated_length": 349.0,
      "entropy": 0.22400058060884476,
      "epoch": 1.9182692307692308,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 994.896484375,
      "learning_rate": 1e-06,
      "loss": 0.0255,
      "num_tokens": 874303465.0,
      "reward": 0.30390554666519165,
      "reward_std": 0.03565283119678497,
      "rewards/progression_diversity/mean": -7.151110912673175e-05,
      "rewards/progression_diversity/std": 0.0012204337399452925,
      "rewards/symbolic_reward_accuracy/mean": 0.169921875,
      "rewards/symbolic_reward_accuracy/std": 0.3759314715862274,
      "rewards/symbolic_reward_partial_score/mean": 0.6744791269302368,
      "rewards/symbolic_reward_partial_score/std": 0.20566026866436005,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0463134050369263,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 724.0,
      "sampling/sampling_logp_difference/mean": 0.4934929311275482,
      "step": 1197
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.2278030663728714,
      "epoch": 1.9198717948717947,
      "grad_norm": 0.0072776079177856445,
      "learning_rate": 1e-06,
      "loss": -0.0058,
      "step": 1198
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.22660139203071594,
      "epoch": 1.921474358974359,
      "grad_norm": 0.011884551495313644,
      "learning_rate": 1e-06,
      "loss": 0.0251,
      "step": 1199
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.22588203847408295,
      "epoch": 1.9230769230769231,
      "grad_norm": 0.008959163911640644,
      "learning_rate": 1e-06,
      "loss": 0.0028,
      "step": 1200
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3548.0,
      "completions/max_terminated_length": 3548.0,
      "completions/mean_length": 1056.767578125,
      "completions/mean_terminated_length": 1056.767578125,
      "completions/min_length": 370.0,
      "completions/min_terminated_length": 370.0,
      "entropy": 0.22032135725021362,
      "epoch": 1.9246794871794872,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.01947578601539135,
      "learning_rate": 1e-06,
      "loss": -0.001,
      "num_tokens": 875797554.0,
      "reward": 0.24959278106689453,
      "reward_std": 0.034925058484077454,
      "rewards/progression_diversity/mean": -0.00019414318376220763,
      "rewards/progression_diversity/std": 0.0018961295718327165,
      "rewards/symbolic_reward_accuracy/mean": 0.09765625,
      "rewards/symbolic_reward_accuracy/std": 0.29713961482048035,
      "rewards/symbolic_reward_partial_score/mean": 0.6366698741912842,
      "rewards/symbolic_reward_partial_score/std": 0.21680741012096405,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0476453304290771,
      "sampling/importance_sampling_ratio/min": 1.6844656158809812e-07,
      "sampling/sampling_logp_difference/max": 15.596647262573242,
      "sampling/sampling_logp_difference/mean": 0.0985308587551117,
      "step": 1201
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.21852584183216095,
      "epoch": 1.9262820512820513,
      "grad_norm": 0.02164105512201786,
      "learning_rate": 1e-06,
      "loss": -0.0057,
      "step": 1202
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.21969875693321228,
      "epoch": 1.9278846153846154,
      "grad_norm": 0.013584469445049763,
      "learning_rate": 1e-06,
      "loss": 0.0022,
      "step": 1203
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2164856493473053,
      "epoch": 1.9294871794871795,
      "grad_norm": 0.01743365451693535,
      "learning_rate": 1e-06,
      "loss": 0.0023,
      "step": 1204
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3204.0,
      "completions/max_terminated_length": 3204.0,
      "completions/mean_length": 1141.201171875,
      "completions/mean_terminated_length": 1141.201171875,
      "completions/min_length": 349.0,
      "completions/min_terminated_length": 349.0,
      "entropy": 0.2209923416376114,
      "epoch": 1.9310897435897436,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 0.022249111905694008,
      "learning_rate": 1e-06,
      "loss": 0.0014,
      "num_tokens": 877274857.0,
      "reward": 0.3205749988555908,
      "reward_std": 0.028882907703518867,
      "rewards/progression_diversity/mean": -0.00011996681860182434,
      "rewards/progression_diversity/std": 0.0014987426111474633,
      "rewards/symbolic_reward_accuracy/mean": 0.17578125,
      "rewards/symbolic_reward_accuracy/std": 0.3810062110424042,
      "rewards/symbolic_reward_partial_score/mean": 0.7170247435569763,
      "rewards/symbolic_reward_partial_score/std": 0.19960245490074158,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0480538606643677,
      "sampling/importance_sampling_ratio/min": 6.5602007333354084e-12,
      "sampling/sampling_logp_difference/max": 25.75,
      "sampling/sampling_logp_difference/mean": 0.09933799505233765,
      "step": 1205
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.21989618986845016,
      "epoch": 1.9326923076923077,
      "grad_norm": 0.008175384253263474,
      "learning_rate": 1e-06,
      "loss": 0.0013,
      "step": 1206
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.22836829721927643,
      "epoch": 1.9342948717948718,
      "grad_norm": 0.016739025712013245,
      "learning_rate": 1e-06,
      "loss": -0.0014,
      "step": 1207
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.2236020788550377,
      "epoch": 1.935897435897436,
      "grad_norm": 0.013731605373322964,
      "learning_rate": 1e-06,
      "loss": 0.0002,
      "step": 1208
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3453.0,
      "completions/mean_length": 1140.861328125,
      "completions/mean_terminated_length": 1111.03125,
      "completions/min_length": 269.0,
      "completions/min_terminated_length": 269.0,
      "entropy": 0.2177734076976776,
      "epoch": 1.9375,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 574.372314453125,
      "learning_rate": 1e-06,
      "loss": 0.0217,
      "num_tokens": 878788738.0,
      "reward": 0.20117032527923584,
      "reward_std": 0.02102627232670784,
      "rewards/progression_diversity/mean": -0.00015734444605186582,
      "rewards/progression_diversity/std": 0.002003313507884741,
      "rewards/symbolic_reward_accuracy/mean": 0.0234375,
      "rewards/symbolic_reward_accuracy/std": 0.15143637359142303,
      "rewards/symbolic_reward_partial_score/mean": 0.6243489980697632,
      "rewards/symbolic_reward_partial_score/std": 0.15533775091171265,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0462570190429688,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.6281495690345764,
      "step": 1209
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.22087281942367554,
      "epoch": 1.939102564102564,
      "grad_norm": 0.0154855502769351,
      "learning_rate": 1e-06,
      "loss": 0.0046,
      "step": 1210
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22198112308979034,
      "epoch": 1.9407051282051282,
      "grad_norm": 0.018509745597839355,
      "learning_rate": 1e-06,
      "loss": -0.0006,
      "step": 1211
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.22084571421146393,
      "epoch": 1.9423076923076923,
      "grad_norm": 0.021920692175626755,
      "learning_rate": 1e-06,
      "loss": -0.0027,
      "step": 1212
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3133.0,
      "completions/mean_length": 1260.185546875,
      "completions/mean_terminated_length": 1200.8765869140625,
      "completions/min_length": 391.0,
      "completions/min_terminated_length": 391.0,
      "entropy": 0.22076184302568436,
      "epoch": 1.9439102564102564,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.016443662345409393,
      "learning_rate": 1e-06,
      "loss": -0.0056,
      "num_tokens": 880305873.0,
      "reward": 0.2951447665691376,
      "reward_std": 0.037818461656570435,
      "rewards/progression_diversity/mean": -0.00017372408183291554,
      "rewards/progression_diversity/std": 0.002410082146525383,
      "rewards/symbolic_reward_accuracy/mean": 0.1484375,
      "rewards/symbolic_reward_accuracy/std": 0.35588082671165466,
      "rewards/symbolic_reward_partial_score/mean": 0.6882486939430237,
      "rewards/symbolic_reward_partial_score/std": 0.20156751573085785,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0463939905166626,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.4395545721054077,
      "step": 1213
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.22432439774274826,
      "epoch": 1.9455128205128205,
      "grad_norm": 752.1104125976562,
      "learning_rate": 1e-06,
      "loss": 0.0734,
      "step": 1214
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.22369728982448578,
      "epoch": 1.9471153846153846,
      "grad_norm": 0.005198925733566284,
      "learning_rate": 1e-06,
      "loss": 0.0301,
      "step": 1215
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.22393931448459625,
      "epoch": 1.9487179487179487,
      "grad_norm": 0.01498700212687254,
      "learning_rate": 1e-06,
      "loss": 0.0054,
      "step": 1216
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2988.0,
      "completions/max_terminated_length": 2988.0,
      "completions/mean_length": 1089.60546875,
      "completions/mean_terminated_length": 1089.60546875,
      "completions/min_length": 313.0,
      "completions/min_terminated_length": 313.0,
      "entropy": 0.2190534770488739,
      "epoch": 1.9503205128205128,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.02252928353846073,
      "learning_rate": 1e-06,
      "loss": -0.0032,
      "num_tokens": 881729991.0,
      "reward": 0.28866058588027954,
      "reward_std": 0.05070298910140991,
      "rewards/progression_diversity/mean": -0.00015254000027198344,
      "rewards/progression_diversity/std": 0.0022601871751248837,
      "rewards/symbolic_reward_accuracy/mean": 0.1171875,
      "rewards/symbolic_reward_accuracy/std": 0.32195815443992615,
      "rewards/symbolic_reward_partial_score/mean": 0.7278319597244263,
      "rewards/symbolic_reward_partial_score/std": 0.17303742468357086,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0459058284759521,
      "sampling/importance_sampling_ratio/min": 2.218274985921198e-08,
      "sampling/sampling_logp_difference/max": 17.623950958251953,
      "sampling/sampling_logp_difference/mean": 0.09728400409221649,
      "step": 1217
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21607843786478043,
      "epoch": 1.9519230769230769,
      "grad_norm": 0.030551020056009293,
      "learning_rate": 1e-06,
      "loss": -0.0004,
      "step": 1218
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2140772044658661,
      "epoch": 1.953525641025641,
      "grad_norm": 0.009000630117952824,
      "learning_rate": 1e-06,
      "loss": 0.0062,
      "step": 1219
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.21863539516925812,
      "epoch": 1.9551282051282053,
      "grad_norm": 0.01365700364112854,
      "learning_rate": 1e-06,
      "loss": 0.0039,
      "step": 1220
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2953.0,
      "completions/max_terminated_length": 2953.0,
      "completions/mean_length": 1145.857421875,
      "completions/mean_terminated_length": 1145.857421875,
      "completions/min_length": 401.0,
      "completions/min_terminated_length": 401.0,
      "entropy": 0.2172141596674919,
      "epoch": 1.9567307692307692,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.024405311793088913,
      "learning_rate": 1e-06,
      "loss": -0.0044,
      "num_tokens": 883216462.0,
      "reward": 0.3235936760902405,
      "reward_std": 0.029494676738977432,
      "rewards/progression_diversity/mean": -8.07223841547966e-06,
      "rewards/progression_diversity/std": 0.00018265389371663332,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.7192708253860474,
      "rewards/symbolic_reward_partial_score/std": 0.19232387840747833,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.046884298324585,
      "sampling/importance_sampling_ratio/min": 1.05600235755735e-20,
      "sampling/sampling_logp_difference/max": 45.99721145629883,
      "sampling/sampling_logp_difference/mean": 0.09837117791175842,
      "step": 1221
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.22011933475732803,
      "epoch": 1.9583333333333335,
      "grad_norm": 0.00895681418478489,
      "learning_rate": 1e-06,
      "loss": 0.0046,
      "step": 1222
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.21496566385030746,
      "epoch": 1.9599358974358974,
      "grad_norm": 0.014537299983203411,
      "learning_rate": 1e-06,
      "loss": 0.0008,
      "step": 1223
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.21989809721708298,
      "epoch": 1.9615384615384617,
      "grad_norm": 0.008813065476715565,
      "learning_rate": 1e-06,
      "loss": 0.0009,
      "step": 1224
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3093.0,
      "completions/max_terminated_length": 3093.0,
      "completions/mean_length": 1287.013671875,
      "completions/mean_terminated_length": 1287.013671875,
      "completions/min_length": 458.0,
      "completions/min_terminated_length": 458.0,
      "entropy": 0.21527937054634094,
      "epoch": 1.9631410256410255,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.017107542604207993,
      "learning_rate": 1e-06,
      "loss": 0.0119,
      "num_tokens": 884739141.0,
      "reward": 0.28098633885383606,
      "reward_std": 0.0465182401239872,
      "rewards/progression_diversity/mean": -1.3702083379030228e-06,
      "rewards/progression_diversity/std": 3.1004274205770344e-05,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.6866210699081421,
      "rewards/symbolic_reward_partial_score/std": 0.21217504143714905,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0459423065185547,
      "sampling/importance_sampling_ratio/min": 9.484894411184965e-19,
      "sampling/sampling_logp_difference/max": 41.49941635131836,
      "sampling/sampling_logp_difference/mean": 0.0953700914978981,
      "step": 1225
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.21589966118335724,
      "epoch": 1.9647435897435899,
      "grad_norm": 0.01544854324311018,
      "learning_rate": 1e-06,
      "loss": -0.0047,
      "step": 1226
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.21224521100521088,
      "epoch": 1.9663461538461537,
      "grad_norm": 0.016370629891753197,
      "learning_rate": 1e-06,
      "loss": -0.0017,
      "step": 1227
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.21077106148004532,
      "epoch": 1.967948717948718,
      "grad_norm": 0.009457322768867016,
      "learning_rate": 1e-06,
      "loss": 0.0047,
      "step": 1228
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3146.0,
      "completions/max_terminated_length": 3146.0,
      "completions/mean_length": 1206.787109375,
      "completions/mean_terminated_length": 1206.787109375,
      "completions/min_length": 437.0,
      "completions/min_terminated_length": 437.0,
      "entropy": 0.21696603298187256,
      "epoch": 1.969551282051282,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 0.01737239398062229,
      "learning_rate": 1e-06,
      "loss": 0.0056,
      "num_tokens": 886241192.0,
      "reward": 0.41241925954818726,
      "reward_std": 0.028087418526411057,
      "rewards/progression_diversity/mean": -0.00026385916862636805,
      "rewards/progression_diversity/std": 0.002647166606038809,
      "rewards/symbolic_reward_accuracy/mean": 0.322265625,
      "rewards/symbolic_reward_accuracy/std": 0.46780112385749817,
      "rewards/symbolic_reward_partial_score/mean": 0.7302083969116211,
      "rewards/symbolic_reward_partial_score/std": 0.22925032675266266,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0460162162780762,
      "sampling/importance_sampling_ratio/min": 0.00035160055267624557,
      "sampling/sampling_logp_difference/max": 7.953014850616455,
      "sampling/sampling_logp_difference/mean": 0.09595166146755219,
      "step": 1229
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.21299895644187927,
      "epoch": 1.9711538461538463,
      "grad_norm": 0.013916675001382828,
      "learning_rate": 1e-06,
      "loss": 0.0016,
      "step": 1230
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.2122596800327301,
      "epoch": 1.9727564102564101,
      "grad_norm": 0.013204848393797874,
      "learning_rate": 1e-06,
      "loss": 0.001,
      "step": 1231
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.2141348496079445,
      "epoch": 1.9743589743589745,
      "grad_norm": 0.015932990238070488,
      "learning_rate": 1e-06,
      "loss": -0.0001,
      "step": 1232
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3633.0,
      "completions/mean_length": 1235.24609375,
      "completions/mean_terminated_length": 1205.600830078125,
      "completions/min_length": 452.0,
      "completions/min_terminated_length": 452.0,
      "entropy": 0.21588806062936783,
      "epoch": 1.9759615384615383,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.016500849276781082,
      "learning_rate": 1e-06,
      "loss": 0.0061,
      "num_tokens": 887740710.0,
      "reward": 0.3222983777523041,
      "reward_std": 0.06374066323041916,
      "rewards/progression_diversity/mean": -0.00014570857456419617,
      "rewards/progression_diversity/std": 0.0015578385209664702,
      "rewards/symbolic_reward_accuracy/mean": 0.19140625,
      "rewards/symbolic_reward_accuracy/std": 0.3937928080558777,
      "rewards/symbolic_reward_partial_score/mean": 0.6921712160110474,
      "rewards/symbolic_reward_partial_score/std": 0.2026739865541458,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045090913772583,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.3644917607307434,
      "step": 1233
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.21097960323095322,
      "epoch": 1.9775641025641026,
      "grad_norm": 0.0278862826526165,
      "learning_rate": 1e-06,
      "loss": 0.0195,
      "step": 1234
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.2159006968140602,
      "epoch": 1.9791666666666665,
      "grad_norm": 0.014327644370496273,
      "learning_rate": 1e-06,
      "loss": 0.0054,
      "step": 1235
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2141767367720604,
      "epoch": 1.9807692307692308,
      "grad_norm": 0.017790069803595543,
      "learning_rate": 1e-06,
      "loss": -0.0013,
      "step": 1236
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3288.0,
      "completions/max_terminated_length": 3288.0,
      "completions/mean_length": 1284.09375,
      "completions/mean_terminated_length": 1284.09375,
      "completions/min_length": 427.0,
      "completions/min_terminated_length": 427.0,
      "entropy": 0.210280142724514,
      "epoch": 1.9823717948717947,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 0.021001746878027916,
      "learning_rate": 1e-06,
      "loss": -0.0052,
      "num_tokens": 889325798.0,
      "reward": 0.2985340356826782,
      "reward_std": 0.05770270898938179,
      "rewards/progression_diversity/mean": -0.00011500486289151013,
      "rewards/progression_diversity/std": 0.0018274638568982482,
      "rewards/symbolic_reward_accuracy/mean": 0.142578125,
      "rewards/symbolic_reward_accuracy/std": 0.3499840497970581,
      "rewards/symbolic_reward_partial_score/mean": 0.7099609375,
      "rewards/symbolic_reward_partial_score/std": 0.19229154288768768,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0444259643554688,
      "sampling/importance_sampling_ratio/min": 6.935766052899004e-22,
      "sampling/sampling_logp_difference/max": 48.72018051147461,
      "sampling/sampling_logp_difference/mean": 0.09380759298801422,
      "step": 1237
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.20352954417467117,
      "epoch": 1.983974358974359,
      "grad_norm": 0.02166086994111538,
      "learning_rate": 1e-06,
      "loss": -0.0037,
      "step": 1238
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.20602308958768845,
      "epoch": 1.9855769230769231,
      "grad_norm": 0.017689600586891174,
      "learning_rate": 1e-06,
      "loss": 0.01,
      "step": 1239
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.20504777133464813,
      "epoch": 1.9871794871794872,
      "grad_norm": 0.01678495481610298,
      "learning_rate": 1e-06,
      "loss": -0.0135,
      "step": 1240
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3242.0,
      "completions/max_terminated_length": 3242.0,
      "completions/mean_length": 1257.923828125,
      "completions/mean_terminated_length": 1257.923828125,
      "completions/min_length": 450.0,
      "completions/min_terminated_length": 450.0,
      "entropy": 0.20558489859104156,
      "epoch": 1.9887820512820513,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.024906298145651817,
      "learning_rate": 1e-06,
      "loss": -0.0053,
      "num_tokens": 890897599.0,
      "reward": 0.18425101041793823,
      "reward_std": 0.02554313838481903,
      "rewards/progression_diversity/mean": -0.00019268083269707859,
      "rewards/progression_diversity/std": 0.0028432749677449465,
      "rewards/symbolic_reward_accuracy/mean": 0.005859375,
      "rewards/symbolic_reward_accuracy/std": 0.07639661431312561,
      "rewards/symbolic_reward_partial_score/mean": 0.6024577021598816,
      "rewards/symbolic_reward_partial_score/std": 0.18746401369571686,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0438592433929443,
      "sampling/importance_sampling_ratio/min": 0.0005599469877779484,
      "sampling/sampling_logp_difference/max": 7.487668514251709,
      "sampling/sampling_logp_difference/mean": 0.09330444037914276,
      "step": 1241
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.20575018227100372,
      "epoch": 1.9903846153846154,
      "grad_norm": 0.01447804644703865,
      "learning_rate": 1e-06,
      "loss": 0.0037,
      "step": 1242
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.208658367395401,
      "epoch": 1.9919871794871795,
      "grad_norm": 0.026018286123871803,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "step": 1243
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.20609137415885925,
      "epoch": 1.9935897435897436,
      "grad_norm": 0.015114527195692062,
      "learning_rate": 1e-06,
      "loss": -0.0032,
      "step": 1244
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3097.0,
      "completions/mean_length": 1268.546875,
      "completions/mean_terminated_length": 1238.9666748046875,
      "completions/min_length": 388.0,
      "completions/min_terminated_length": 388.0,
      "entropy": 0.19605816900730133,
      "epoch": 1.9951923076923077,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.026206906884908676,
      "learning_rate": 1e-06,
      "loss": 0.0017,
      "num_tokens": 892442183.0,
      "reward": 0.3488447368144989,
      "reward_std": 0.04794444516301155,
      "rewards/progression_diversity/mean": -0.000291308737359941,
      "rewards/progression_diversity/std": 0.0030992666725069284,
      "rewards/symbolic_reward_accuracy/mean": 0.2265625,
      "rewards/symbolic_reward_accuracy/std": 0.4190165400505066,
      "rewards/symbolic_reward_partial_score/mean": 0.7097005248069763,
      "rewards/symbolic_reward_partial_score/std": 0.21007077395915985,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0414423942565918,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.47119155526161194,
      "step": 1245
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.19719059765338898,
      "epoch": 1.9967948717948718,
      "grad_norm": 0.014178547076880932,
      "learning_rate": 1e-06,
      "loss": -0.0034,
      "step": 1246
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.19025534391403198,
      "epoch": 1.998397435897436,
      "grad_norm": 0.00976389180868864,
      "learning_rate": 1e-06,
      "loss": 0.0362,
      "step": 1247
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.19911998510360718,
      "epoch": 2.0,
      "grad_norm": 0.021017316728830338,
      "learning_rate": 1e-06,
      "loss": 0.0021,
      "step": 1248
    },
    {
      "epoch": 2.0,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.00244140625,
      "eval_completions/max_length": 6023.21875,
      "eval_completions/max_terminated_length": 3145.0,
      "eval_completions/mean_length": 1369.302734375,
      "eval_completions/mean_terminated_length": 1332.704662322998,
      "eval_completions/min_length": 460.25,
      "eval_completions/min_terminated_length": 460.25,
      "eval_entropy": 0.19390225457027555,
      "eval_frac_reward_zero_std": 0.28125,
      "eval_loss": 0.00268191983923316,
      "eval_num_tokens": 892442183.0,
      "eval_reward": 0.23870036005973816,
      "eval_reward_std": 0.03971031281980686,
      "eval_rewards/progression_diversity/mean": -0.0003259587915636075,
      "eval_rewards/progression_diversity/std": 0.002640488281031139,
      "eval_rewards/symbolic_reward_accuracy/mean": 0.079345703125,
      "eval_rewards/symbolic_reward_accuracy/std": 0.19223300064913929,
      "eval_rewards/symbolic_reward_partial_score/mean": 0.6377197187393904,
      "eval_rewards/symbolic_reward_partial_score/std": 0.1931128588039428,
      "eval_rewards/tag_count_reward/mean": -0.002197265625,
      "eval_rewards/tag_count_reward/std": 0.019685723586007953,
      "eval_runtime": 1580.0116,
      "eval_samples_per_second": 0.158,
      "eval_sampling/importance_sampling_ratio/max": 2.0,
      "eval_sampling/importance_sampling_ratio/mean": 1.0408074893057346,
      "eval_sampling/importance_sampling_ratio/min": 0.00181254785630403,
      "eval_sampling/sampling_logp_difference/max": 151.67179602384567,
      "eval_sampling/sampling_logp_difference/mean": 0.2478228227701038,
      "eval_steps_per_second": 0.001,
      "step": 1248
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3355.0,
      "completions/mean_length": 1373.986328125,
      "completions/mean_terminated_length": 1315.1236572265625,
      "completions/min_length": 392.0,
      "completions/min_terminated_length": 392.0,
      "entropy": 0.20005454123020172,
      "epoch": 2.0016025641025643,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 233.30934143066406,
      "learning_rate": 1e-06,
      "loss": 0.0216,
      "num_tokens": 893940272.0,
      "reward": 0.36688268184661865,
      "reward_std": 0.056143928319215775,
      "rewards/progression_diversity/mean": -0.001186927082017064,
      "rewards/progression_diversity/std": 0.01717980019748211,
      "rewards/symbolic_reward_accuracy/mean": 0.240234375,
      "rewards/symbolic_reward_accuracy/std": 0.4276435375213623,
      "rewards/symbolic_reward_partial_score/mean": 0.7431640625,
      "rewards/symbolic_reward_partial_score/std": 0.20910829305648804,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0366449356079102,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 2.187889575958252,
      "step": 1249
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.19553004205226898,
      "epoch": 2.003205128205128,
      "grad_norm": 0.019861171022057533,
      "learning_rate": 1e-06,
      "loss": -0.0016,
      "step": 1250
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.19362886995077133,
      "epoch": 2.0048076923076925,
      "grad_norm": 0.02055576629936695,
      "learning_rate": 1e-06,
      "loss": 0.0059,
      "step": 1251
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.19335567951202393,
      "epoch": 2.0064102564102564,
      "grad_norm": 0.017372747883200645,
      "learning_rate": 1e-06,
      "loss": 0.0263,
      "step": 1252
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3457.0,
      "completions/mean_length": 1377.3046875,
      "completions/mean_terminated_length": 1347.9373779296875,
      "completions/min_length": 419.0,
      "completions/min_terminated_length": 419.0,
      "entropy": 0.1914682313799858,
      "epoch": 2.0080128205128207,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 0.020828014239668846,
      "learning_rate": 1e-06,
      "loss": -0.0009,
      "num_tokens": 895523628.0,
      "reward": 0.28705087304115295,
      "reward_std": 0.06678985804319382,
      "rewards/progression_diversity/mean": -0.00047905254177749157,
      "rewards/progression_diversity/std": 0.005628896411508322,
      "rewards/symbolic_reward_accuracy/mean": 0.13671875,
      "rewards/symbolic_reward_accuracy/std": 0.3438861668109894,
      "rewards/symbolic_reward_partial_score/mean": 0.6840657591819763,
      "rewards/symbolic_reward_partial_score/std": 0.20577622950077057,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0396616458892822,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.37476086616516113,
      "step": 1253
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.19290538132190704,
      "epoch": 2.0096153846153846,
      "grad_norm": 1153.7135009765625,
      "learning_rate": 1e-06,
      "loss": 0.144,
      "step": 1254
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.19067180156707764,
      "epoch": 2.011217948717949,
      "grad_norm": 0.014077689498662949,
      "learning_rate": 1e-06,
      "loss": -0.0025,
      "step": 1255
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.19498515874147415,
      "epoch": 2.0128205128205128,
      "grad_norm": 0.017978975549340248,
      "learning_rate": 1e-06,
      "loss": 0.0053,
      "step": 1256
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3308.0,
      "completions/mean_length": 1435.3828125,
      "completions/mean_terminated_length": 1376.7608642578125,
      "completions/min_length": 418.0,
      "completions/min_terminated_length": 418.0,
      "entropy": 0.19777005165815353,
      "epoch": 2.014423076923077,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 0.028694914653897285,
      "learning_rate": 1e-06,
      "loss": 0.0114,
      "num_tokens": 897164496.0,
      "reward": 0.3483375608921051,
      "reward_std": 0.07038730382919312,
      "rewards/progression_diversity/mean": -0.0007183193229138851,
      "rewards/progression_diversity/std": 0.0077925813384354115,
      "rewards/symbolic_reward_accuracy/mean": 0.232421875,
      "rewards/symbolic_reward_accuracy/std": 0.42278963327407837,
      "rewards/symbolic_reward_partial_score/mean": 0.6976073980331421,
      "rewards/symbolic_reward_partial_score/std": 0.23062647879123688,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0391186475753784,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.5552532076835632,
      "step": 1257
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.19494935125112534,
      "epoch": 2.016025641025641,
      "grad_norm": 0.016217295080423355,
      "learning_rate": 1e-06,
      "loss": 0.0099,
      "step": 1258
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.1985015571117401,
      "epoch": 2.0176282051282053,
      "grad_norm": 0.02569577842950821,
      "learning_rate": 1e-06,
      "loss": 0.0176,
      "step": 1259
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.1980082392692566,
      "epoch": 2.019230769230769,
      "grad_norm": 0.01939314603805542,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "step": 1260
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3630.0,
      "completions/mean_length": 1615.064453125,
      "completions/mean_terminated_length": 1469.4141845703125,
      "completions/min_length": 395.0,
      "completions/min_terminated_length": 395.0,
      "entropy": 0.18866483867168427,
      "epoch": 2.0208333333333335,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 20.04486656188965,
      "learning_rate": 1e-06,
      "loss": 0.0255,
      "num_tokens": 898821809.0,
      "reward": 0.36717134714126587,
      "reward_std": 0.06693032383918762,
      "rewards/progression_diversity/mean": -0.0011277215089648962,
      "rewards/progression_diversity/std": 0.00911035481840372,
      "rewards/symbolic_reward_accuracy/mean": 0.244140625,
      "rewards/symbolic_reward_accuracy/std": 0.42999663949012756,
      "rewards/symbolic_reward_partial_score/mean": 0.7389160394668579,
      "rewards/symbolic_reward_partial_score/std": 0.21004585921764374,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0375559329986572,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.6628260612487793,
      "step": 1261
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.18701110780239105,
      "epoch": 2.0224358974358974,
      "grad_norm": 0.018055502325296402,
      "learning_rate": 1e-06,
      "loss": 0.02,
      "step": 1262
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.18735942244529724,
      "epoch": 2.0240384615384617,
      "grad_norm": 0.02284780703485012,
      "learning_rate": 1e-06,
      "loss": 0.0494,
      "step": 1263
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.1921609714627266,
      "epoch": 2.0256410256410255,
      "grad_norm": 0.03690061345696449,
      "learning_rate": 1e-06,
      "loss": 0.0227,
      "step": 1264
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3370.0,
      "completions/mean_length": 1423.083984375,
      "completions/mean_terminated_length": 1364.413818359375,
      "completions/min_length": 419.0,
      "completions/min_terminated_length": 419.0,
      "entropy": 0.1979808807373047,
      "epoch": 2.02724358974359,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.024405112490057945,
      "learning_rate": 1e-06,
      "loss": 0.0033,
      "num_tokens": 900394908.0,
      "reward": 0.3067031800746918,
      "reward_std": 0.04862578958272934,
      "rewards/progression_diversity/mean": -9.318315278505906e-05,
      "rewards/progression_diversity/std": 0.001644739182665944,
      "rewards/symbolic_reward_accuracy/mean": 0.14453125,
      "rewards/symbolic_reward_accuracy/std": 0.35197147727012634,
      "rewards/symbolic_reward_partial_score/mean": 0.7345865964889526,
      "rewards/symbolic_reward_partial_score/std": 0.20190957188606262,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0397248268127441,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.6828402876853943,
      "step": 1265
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.19661108404397964,
      "epoch": 2.0288461538461537,
      "grad_norm": 0.013450069352984428,
      "learning_rate": 1e-06,
      "loss": 0.054,
      "step": 1266
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.19856248795986176,
      "epoch": 2.030448717948718,
      "grad_norm": 0.017992835491895676,
      "learning_rate": 1e-06,
      "loss": -0.0069,
      "step": 1267
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.19879474490880966,
      "epoch": 2.032051282051282,
      "grad_norm": 0.019288472831249237,
      "learning_rate": 1e-06,
      "loss": -0.0147,
      "step": 1268
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 9273.0,
      "completions/mean_length": 1691.78515625,
      "completions/mean_terminated_length": 1546.8914794921875,
      "completions/min_length": 350.0,
      "completions/min_terminated_length": 350.0,
      "entropy": 0.19042234122753143,
      "epoch": 2.0336538461538463,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 47.59299087524414,
      "learning_rate": 1e-06,
      "loss": 0.0107,
      "num_tokens": 902148702.0,
      "reward": 0.3654331862926483,
      "reward_std": 0.061046987771987915,
      "rewards/progression_diversity/mean": -0.0011158745037391782,
      "rewards/progression_diversity/std": 0.014414801262319088,
      "rewards/symbolic_reward_accuracy/mean": 0.232421875,
      "rewards/symbolic_reward_accuracy/std": 0.42278963327407837,
      "rewards/symbolic_reward_partial_score/mean": 0.7565592527389526,
      "rewards/symbolic_reward_partial_score/std": 0.20248234272003174,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0370126962661743,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 1.0821824073791504,
      "step": 1269
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.18952977657318115,
      "epoch": 2.03525641025641,
      "grad_norm": 0.0266401544213295,
      "learning_rate": 1e-06,
      "loss": 0.1359,
      "step": 1270
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.18822447210550308,
      "epoch": 2.0368589743589745,
      "grad_norm": 0.027362341061234474,
      "learning_rate": 1e-06,
      "loss": 0.0679,
      "step": 1271
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.18995289504528046,
      "epoch": 2.0384615384615383,
      "grad_norm": 0.014087039045989513,
      "learning_rate": 1e-06,
      "loss": 0.0075,
      "step": 1272
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3306.0,
      "completions/mean_length": 1566.1796875,
      "completions/mean_terminated_length": 1420.04736328125,
      "completions/min_length": 428.0,
      "completions/min_terminated_length": 428.0,
      "entropy": 0.19420960545539856,
      "epoch": 2.0400641025641026,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.01688203774392605,
      "learning_rate": 1e-06,
      "loss": 0.0064,
      "num_tokens": 903831514.0,
      "reward": 0.36713409423828125,
      "reward_std": 0.03999203443527222,
      "rewards/progression_diversity/mean": -0.00046001013834029436,
      "rewards/progression_diversity/std": 0.004506079014390707,
      "rewards/symbolic_reward_accuracy/mean": 0.240234375,
      "rewards/symbolic_reward_accuracy/std": 0.4276435375213623,
      "rewards/symbolic_reward_partial_score/mean": 0.74658203125,
      "rewards/symbolic_reward_partial_score/std": 0.20671479403972626,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0370961427688599,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 1.4572553634643555,
      "step": 1273
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.1954876109957695,
      "epoch": 2.0416666666666665,
      "grad_norm": 0.018446512520313263,
      "learning_rate": 1e-06,
      "loss": 0.0048,
      "step": 1274
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.1945226937532425,
      "epoch": 2.043269230769231,
      "grad_norm": 324.0447082519531,
      "learning_rate": 1e-06,
      "loss": 0.0717,
      "step": 1275
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.19547558575868607,
      "epoch": 2.0448717948717947,
      "grad_norm": 755.1400756835938,
      "learning_rate": 1e-06,
      "loss": 0.0474,
      "step": 1276
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3276.0,
      "completions/mean_length": 1617.3359375,
      "completions/mean_terminated_length": 1353.1212158203125,
      "completions/min_length": 395.0,
      "completions/min_terminated_length": 395.0,
      "entropy": 0.19297361373901367,
      "epoch": 2.046474358974359,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.018968136981129646,
      "learning_rate": 1e-06,
      "loss": 0.0109,
      "num_tokens": 905556918.0,
      "reward": 0.2657776474952698,
      "reward_std": 0.0448925644159317,
      "rewards/progression_diversity/mean": -0.00036140886368229985,
      "rewards/progression_diversity/std": 0.004768828861415386,
      "rewards/symbolic_reward_accuracy/mean": 0.115234375,
      "rewards/symbolic_reward_accuracy/std": 0.3196168541908264,
      "rewards/symbolic_reward_partial_score/mean": 0.6613280773162842,
      "rewards/symbolic_reward_partial_score/std": 0.21297021210193634,
      "rewards/tag_count_reward/mean": -0.017578125,
      "rewards/tag_count_reward/std": 0.13154059648513794,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0336802005767822,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 2.5898587703704834,
      "step": 1277
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.18827137351036072,
      "epoch": 2.048076923076923,
      "grad_norm": 0.011503035202622414,
      "learning_rate": 1e-06,
      "loss": 0.0706,
      "step": 1278
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.19312509894371033,
      "epoch": 2.0496794871794872,
      "grad_norm": 0.009355909191071987,
      "learning_rate": 1e-06,
      "loss": 0.0425,
      "step": 1279
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.18875424563884735,
      "epoch": 2.051282051282051,
      "grad_norm": 0.016550183296203613,
      "learning_rate": 1e-06,
      "loss": 0.0553,
      "step": 1280
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3837.0,
      "completions/mean_length": 1660.9140625,
      "completions/mean_terminated_length": 1456.831787109375,
      "completions/min_length": 438.0,
      "completions/min_terminated_length": 438.0,
      "entropy": 0.19483474642038345,
      "epoch": 2.0528846153846154,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.02122480981051922,
      "learning_rate": 1e-06,
      "loss": -0.0055,
      "num_tokens": 907271450.0,
      "reward": 0.33715343475341797,
      "reward_std": 0.034450314939022064,
      "rewards/progression_diversity/mean": -0.00047867521061562,
      "rewards/progression_diversity/std": 0.004741494078189135,
      "rewards/symbolic_reward_accuracy/mean": 0.216796875,
      "rewards/symbolic_reward_accuracy/std": 0.4124660789966583,
      "rewards/symbolic_reward_partial_score/mean": 0.69482421875,
      "rewards/symbolic_reward_partial_score/std": 0.21679222583770752,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0364654064178467,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.3800773620605469,
      "step": 1281
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.1928672194480896,
      "epoch": 2.0544871794871793,
      "grad_norm": 1.5037897825241089,
      "learning_rate": 1e-06,
      "loss": 0.0278,
      "step": 1282
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.1988951489329338,
      "epoch": 2.0560897435897436,
      "grad_norm": 0.024000803008675575,
      "learning_rate": 1e-06,
      "loss": 0.0196,
      "step": 1283
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.19203483313322067,
      "epoch": 2.0576923076923075,
      "grad_norm": 0.019834544509649277,
      "learning_rate": 1e-06,
      "loss": 0.0836,
      "step": 1284
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5436.0,
      "completions/mean_length": 1464.345703125,
      "completions/mean_terminated_length": 1405.8372802734375,
      "completions/min_length": 348.0,
      "completions/min_terminated_length": 348.0,
      "entropy": 0.20045289397239685,
      "epoch": 2.059294871794872,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 28.706775665283203,
      "learning_rate": 1e-06,
      "loss": 0.0216,
      "num_tokens": 908900203.0,
      "reward": 0.30004429817199707,
      "reward_std": 0.05181866139173508,
      "rewards/progression_diversity/mean": -0.0009444843744859099,
      "rewards/progression_diversity/std": 0.011486358940601349,
      "rewards/symbolic_reward_accuracy/mean": 0.154296875,
      "rewards/symbolic_reward_accuracy/std": 0.36158639192581177,
      "rewards/symbolic_reward_partial_score/mean": 0.6928874254226685,
      "rewards/symbolic_reward_partial_score/std": 0.2050057202577591,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0416698455810547,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.41028910875320435,
      "step": 1285
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2051420360803604,
      "epoch": 2.0608974358974357,
      "grad_norm": 0.01245288085192442,
      "learning_rate": 1e-06,
      "loss": -0.0077,
      "step": 1286
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.2012268602848053,
      "epoch": 2.0625,
      "grad_norm": 0.025453390553593636,
      "learning_rate": 1e-06,
      "loss": 0.0173,
      "step": 1287
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.20480038225650787,
      "epoch": 2.064102564102564,
      "grad_norm": 0.025336798280477524,
      "learning_rate": 1e-06,
      "loss": 0.0103,
      "step": 1288
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3935.0,
      "completions/mean_length": 1418.521484375,
      "completions/mean_terminated_length": 1389.23486328125,
      "completions/min_length": 429.0,
      "completions/min_terminated_length": 429.0,
      "entropy": 0.2055332362651825,
      "epoch": 2.065705128205128,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 157.3429412841797,
      "learning_rate": 1e-06,
      "loss": 0.0238,
      "num_tokens": 910496934.0,
      "reward": 0.26742783188819885,
      "reward_std": 0.019551731646060944,
      "rewards/progression_diversity/mean": -0.0008687502122484148,
      "rewards/progression_diversity/std": 0.010150086134672165,
      "rewards/symbolic_reward_accuracy/mean": 0.09375,
      "rewards/symbolic_reward_accuracy/std": 0.29176566004753113,
      "rewards/symbolic_reward_partial_score/mean": 0.7046061158180237,
      "rewards/symbolic_reward_partial_score/std": 0.19600564241409302,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0424489974975586,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.4152454733848572,
      "step": 1289
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.2030770629644394,
      "epoch": 2.0673076923076925,
      "grad_norm": 0.015944253653287888,
      "learning_rate": 1e-06,
      "loss": -0.0017,
      "step": 1290
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.20718741416931152,
      "epoch": 2.0689102564102564,
      "grad_norm": 0.014883228577673435,
      "learning_rate": 1e-06,
      "loss": 0.0054,
      "step": 1291
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.20744304358959198,
      "epoch": 2.0705128205128207,
      "grad_norm": 0.014635228551924229,
      "learning_rate": 1e-06,
      "loss": -0.0017,
      "step": 1292
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3403.0,
      "completions/mean_length": 1773.353515625,
      "completions/mean_terminated_length": 1600.1048583984375,
      "completions/min_length": 487.0,
      "completions/min_terminated_length": 487.0,
      "entropy": 0.18955910950899124,
      "epoch": 2.0721153846153846,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 1073.1669921875,
      "learning_rate": 1e-06,
      "loss": 0.058,
      "num_tokens": 912368443.0,
      "reward": 0.2491791993379593,
      "reward_std": 0.03544043377041817,
      "rewards/progression_diversity/mean": -0.0005382450763136148,
      "rewards/progression_diversity/std": 0.008408893831074238,
      "rewards/symbolic_reward_accuracy/mean": 0.08203125,
      "rewards/symbolic_reward_accuracy/std": 0.2746807038784027,
      "rewards/symbolic_reward_partial_score/mean": 0.670458972454071,
      "rewards/symbolic_reward_partial_score/std": 0.19899418950080872,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.036078691482544,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.4867057800292969,
      "step": 1293
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.19430772960186005,
      "epoch": 2.073717948717949,
      "grad_norm": 0.013100269250571728,
      "learning_rate": 1e-06,
      "loss": -0.0115,
      "step": 1294
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.19107140600681305,
      "epoch": 2.0753205128205128,
      "grad_norm": 0.012087260372936726,
      "learning_rate": 1e-06,
      "loss": 0.0252,
      "step": 1295
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.1892467811703682,
      "epoch": 2.076923076923077,
      "grad_norm": 0.01232385728508234,
      "learning_rate": 1e-06,
      "loss": 0.0455,
      "step": 1296
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3224.0,
      "completions/mean_length": 1561.599609375,
      "completions/mean_terminated_length": 1444.8878173828125,
      "completions/min_length": 362.0,
      "completions/min_terminated_length": 362.0,
      "entropy": 0.1971185952425003,
      "epoch": 2.078525641025641,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 33.13689422607422,
      "learning_rate": 1e-06,
      "loss": 0.0168,
      "num_tokens": 914118782.0,
      "reward": 0.27982568740844727,
      "reward_std": 0.04554954543709755,
      "rewards/progression_diversity/mean": -0.00034400858567096293,
      "rewards/progression_diversity/std": 0.0038654166273772717,
      "rewards/symbolic_reward_accuracy/mean": 0.12109375,
      "rewards/symbolic_reward_accuracy/std": 0.3265552520751953,
      "rewards/symbolic_reward_partial_score/mean": 0.6931803226470947,
      "rewards/symbolic_reward_partial_score/std": 0.18553432822227478,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.037705659866333,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 732.0,
      "sampling/sampling_logp_difference/mean": 1.3978204727172852,
      "step": 1297
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.1944933459162712,
      "epoch": 2.0801282051282053,
      "grad_norm": 0.015022831037640572,
      "learning_rate": 1e-06,
      "loss": 0.0251,
      "step": 1298
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.19650224596261978,
      "epoch": 2.081730769230769,
      "grad_norm": 0.022843046113848686,
      "learning_rate": 1e-06,
      "loss": -0.0049,
      "step": 1299
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.19136834889650345,
      "epoch": 2.0833333333333335,
      "grad_norm": 0.02248123101890087,
      "learning_rate": 1e-06,
      "loss": 0.0554,
      "step": 1300
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3264.0,
      "completions/mean_length": 1438.8671875,
      "completions/mean_terminated_length": 1409.620361328125,
      "completions/min_length": 310.0,
      "completions/min_terminated_length": 310.0,
      "entropy": 0.2071860283613205,
      "epoch": 2.0849358974358974,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.022805871441960335,
      "learning_rate": 1e-06,
      "loss": 0.0063,
      "num_tokens": 915655274.0,
      "reward": 0.4092583954334259,
      "reward_std": 0.04942026361823082,
      "rewards/progression_diversity/mean": -0.0004308145144023001,
      "rewards/progression_diversity/std": 0.00391729548573494,
      "rewards/symbolic_reward_accuracy/mean": 0.302734375,
      "rewards/symbolic_reward_accuracy/std": 0.45989060401916504,
      "rewards/symbolic_reward_partial_score/mean": 0.7593912482261658,
      "rewards/symbolic_reward_partial_score/std": 0.2201661616563797,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0432233810424805,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.34774258732795715,
      "step": 1301
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.20583676546812057,
      "epoch": 2.0865384615384617,
      "grad_norm": 0.010654095560312271,
      "learning_rate": 1e-06,
      "loss": -0.0036,
      "step": 1302
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.21007215976715088,
      "epoch": 2.0881410256410255,
      "grad_norm": 0.018026195466518402,
      "learning_rate": 1e-06,
      "loss": -0.002,
      "step": 1303
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.20896370708942413,
      "epoch": 2.08974358974359,
      "grad_norm": 0.018298866227269173,
      "learning_rate": 1e-06,
      "loss": 0.0288,
      "step": 1304
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2887.0,
      "completions/mean_length": 1537.986328125,
      "completions/mean_terminated_length": 1450.4853515625,
      "completions/min_length": 502.0,
      "completions/min_terminated_length": 502.0,
      "entropy": 0.19623306393623352,
      "epoch": 2.0913461538461537,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 455.7239685058594,
      "learning_rate": 1e-06,
      "loss": 0.0324,
      "num_tokens": 917384147.0,
      "reward": 0.27777591347694397,
      "reward_std": 0.05723512917757034,
      "rewards/progression_diversity/mean": -0.00024088792270049453,
      "rewards/progression_diversity/std": 0.0034442872274667025,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.6778808832168579,
      "rewards/symbolic_reward_partial_score/std": 0.21165545284748077,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.040327787399292,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 0.7170519232749939,
      "step": 1305
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.20050545781850815,
      "epoch": 2.092948717948718,
      "grad_norm": 0.024372782558202744,
      "learning_rate": 1e-06,
      "loss": 0.0046,
      "step": 1306
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.20225122570991516,
      "epoch": 2.094551282051282,
      "grad_norm": 0.017451845109462738,
      "learning_rate": 1e-06,
      "loss": -0.0046,
      "step": 1307
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.19906309247016907,
      "epoch": 2.0961538461538463,
      "grad_norm": 0.015806974843144417,
      "learning_rate": 1e-06,
      "loss": 0.0063,
      "step": 1308
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3072.0,
      "completions/mean_length": 1569.970703125,
      "completions/mean_terminated_length": 1569.970703125,
      "completions/min_length": 474.0,
      "completions/min_terminated_length": 474.0,
      "entropy": 0.2078276202082634,
      "epoch": 2.09775641025641,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.030513443052768707,
      "learning_rate": 1e-06,
      "loss": 0.0077,
      "num_tokens": 918996132.0,
      "reward": 0.3342166543006897,
      "reward_std": 0.05483569577336311,
      "rewards/progression_diversity/mean": -0.00021007427130825818,
      "rewards/progression_diversity/std": 0.001884151715785265,
      "rewards/symbolic_reward_accuracy/mean": 0.203125,
      "rewards/symbolic_reward_accuracy/std": 0.4027182459831238,
      "rewards/symbolic_reward_partial_score/mean": 0.7078125476837158,
      "rewards/symbolic_reward_partial_score/std": 0.2104804515838623,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0436408519744873,
      "sampling/importance_sampling_ratio/min": 1.435761191714846e-06,
      "sampling/sampling_logp_difference/max": 13.453815460205078,
      "sampling/sampling_logp_difference/mean": 0.09233014285564423,
      "step": 1309
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.20950071513652802,
      "epoch": 2.0993589743589745,
      "grad_norm": 0.01937035657465458,
      "learning_rate": 1e-06,
      "loss": -0.0061,
      "step": 1310
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.21072106808423996,
      "epoch": 2.1009615384615383,
      "grad_norm": 0.019583554938435555,
      "learning_rate": 1e-06,
      "loss": -0.0035,
      "step": 1311
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.20755978673696518,
      "epoch": 2.1025641025641026,
      "grad_norm": 0.017628004774451256,
      "learning_rate": 1e-06,
      "loss": 0.013,
      "step": 1312
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2935.0,
      "completions/max_terminated_length": 2935.0,
      "completions/mean_length": 1462.236328125,
      "completions/mean_terminated_length": 1462.236328125,
      "completions/min_length": 465.0,
      "completions/min_terminated_length": 465.0,
      "entropy": 0.20946191251277924,
      "epoch": 2.1041666666666665,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.016548819839954376,
      "learning_rate": 1e-06,
      "loss": -0.0005,
      "num_tokens": 920573229.0,
      "reward": 0.3093579113483429,
      "reward_std": 0.030774272978305817,
      "rewards/progression_diversity/mean": -0.0002441601827740669,
      "rewards/progression_diversity/std": 0.0027843969874083996,
      "rewards/symbolic_reward_accuracy/mean": 0.146484375,
      "rewards/symbolic_reward_accuracy/std": 0.35393697023391724,
      "rewards/symbolic_reward_partial_score/mean": 0.7382323741912842,
      "rewards/symbolic_reward_partial_score/std": 0.1788843870162964,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.044257640838623,
      "sampling/importance_sampling_ratio/min": 0.00013327680062502623,
      "sampling/sampling_logp_difference/max": 8.92308235168457,
      "sampling/sampling_logp_difference/mean": 0.09280422329902649,
      "step": 1313
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2082810178399086,
      "epoch": 2.105769230769231,
      "grad_norm": 0.018022574484348297,
      "learning_rate": 1e-06,
      "loss": 0.0031,
      "step": 1314
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.20982997119426727,
      "epoch": 2.1073717948717947,
      "grad_norm": 0.01984243281185627,
      "learning_rate": 1e-06,
      "loss": -0.004,
      "step": 1315
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.21134591847658157,
      "epoch": 2.108974358974359,
      "grad_norm": 0.025971846655011177,
      "learning_rate": 1e-06,
      "loss": 0.0064,
      "step": 1316
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3383.0,
      "completions/mean_length": 1503.8203125,
      "completions/mean_terminated_length": 1474.7005615234375,
      "completions/min_length": 454.0,
      "completions/min_terminated_length": 454.0,
      "entropy": 0.20815780013799667,
      "epoch": 2.110576923076923,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 0.027938535436987877,
      "learning_rate": 1e-06,
      "loss": 0.0123,
      "num_tokens": 922209697.0,
      "reward": 0.3024898171424866,
      "reward_std": 0.08157630264759064,
      "rewards/progression_diversity/mean": -0.0005339247290976346,
      "rewards/progression_diversity/std": 0.009711179882287979,
      "rewards/symbolic_reward_accuracy/mean": 0.162109375,
      "rewards/symbolic_reward_accuracy/std": 0.3689115643501282,
      "rewards/symbolic_reward_partial_score/mean": 0.6847493648529053,
      "rewards/symbolic_reward_partial_score/std": 0.21822915971279144,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0429041385650635,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 0.4195671081542969,
      "step": 1317
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21150559186935425,
      "epoch": 2.1121794871794872,
      "grad_norm": 0.0272557120770216,
      "learning_rate": 1e-06,
      "loss": -0.0185,
      "step": 1318
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.20635531842708588,
      "epoch": 2.113782051282051,
      "grad_norm": 0.0219330545514822,
      "learning_rate": 1e-06,
      "loss": 0.0242,
      "step": 1319
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.20883511006832123,
      "epoch": 2.1153846153846154,
      "grad_norm": 344.6760559082031,
      "learning_rate": 1e-06,
      "loss": 0.0317,
      "step": 1320
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3218.0,
      "completions/mean_length": 1496.935546875,
      "completions/mean_terminated_length": 1467.8023681640625,
      "completions/min_length": 487.0,
      "completions/min_terminated_length": 487.0,
      "entropy": 0.20407634973526,
      "epoch": 2.1169871794871793,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.02108251117169857,
      "learning_rate": 1e-06,
      "loss": 0.0286,
      "num_tokens": 923895312.0,
      "reward": 0.36836761236190796,
      "reward_std": 0.025770656764507294,
      "rewards/progression_diversity/mean": -0.0001522178645245731,
      "rewards/progression_diversity/std": 0.0016012955456972122,
      "rewards/symbolic_reward_accuracy/mean": 0.24609375,
      "rewards/symbolic_reward_accuracy/std": 0.4311550557613373,
      "rewards/symbolic_reward_partial_score/mean": 0.7357096076011658,
      "rewards/symbolic_reward_partial_score/std": 0.22949159145355225,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0432062149047852,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 728.0,
      "sampling/sampling_logp_difference/mean": 0.1866808831691742,
      "step": 1321
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.20261122286319733,
      "epoch": 2.1185897435897436,
      "grad_norm": 0.024469273164868355,
      "learning_rate": 1e-06,
      "loss": 0.006,
      "step": 1322
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.20310993492603302,
      "epoch": 2.1201923076923075,
      "grad_norm": 0.023724382743239403,
      "learning_rate": 1e-06,
      "loss": -0.0056,
      "step": 1323
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.20938430726528168,
      "epoch": 2.121794871794872,
      "grad_norm": 0.012969693168997765,
      "learning_rate": 1e-06,
      "loss": 0.0018,
      "step": 1324
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2685.0,
      "completions/max_terminated_length": 2685.0,
      "completions/mean_length": 1410.677734375,
      "completions/mean_terminated_length": 1410.677734375,
      "completions/min_length": 437.0,
      "completions/min_terminated_length": 437.0,
      "entropy": 0.2097427099943161,
      "epoch": 2.123397435897436,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.01685580424964428,
      "learning_rate": 1e-06,
      "loss": 0.0013,
      "num_tokens": 925466827.0,
      "reward": 0.38571590185165405,
      "reward_std": 0.03916021063923836,
      "rewards/progression_diversity/mean": -0.00018581100448500365,
      "rewards/progression_diversity/std": 0.002618141006678343,
      "rewards/symbolic_reward_accuracy/mean": 0.255859375,
      "rewards/symbolic_reward_accuracy/std": 0.43676990270614624,
      "rewards/symbolic_reward_partial_score/mean": 0.7740071415901184,
      "rewards/symbolic_reward_partial_score/std": 0.18180802464485168,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0450756549835205,
      "sampling/importance_sampling_ratio/min": 1.69496820490167e-06,
      "sampling/sampling_logp_difference/max": 13.287846565246582,
      "sampling/sampling_logp_difference/mean": 0.0947427749633789,
      "step": 1325
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.2110649198293686,
      "epoch": 2.125,
      "grad_norm": 0.019374966621398926,
      "learning_rate": 1e-06,
      "loss": 0.0141,
      "step": 1326
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.20897484570741653,
      "epoch": 2.126602564102564,
      "grad_norm": 0.011842702515423298,
      "learning_rate": 1e-06,
      "loss": -0.0107,
      "step": 1327
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.21269075572490692,
      "epoch": 2.128205128205128,
      "grad_norm": 0.01617637649178505,
      "learning_rate": 1e-06,
      "loss": 0.007,
      "step": 1328
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2945.0,
      "completions/mean_length": 1554.6484375,
      "completions/mean_terminated_length": 1467.24560546875,
      "completions/min_length": 448.0,
      "completions/min_terminated_length": 448.0,
      "entropy": 0.2093176394701004,
      "epoch": 2.1298076923076925,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 3.476539134979248,
      "learning_rate": 1e-06,
      "loss": -0.0011,
      "num_tokens": 927113287.0,
      "reward": 0.372197687625885,
      "reward_std": 0.024160441011190414,
      "rewards/progression_diversity/mean": -0.0009393933578394353,
      "rewards/progression_diversity/std": 0.014015908353030682,
      "rewards/symbolic_reward_accuracy/mean": 0.244140625,
      "rewards/symbolic_reward_accuracy/std": 0.42999663949012756,
      "rewards/symbolic_reward_partial_score/mean": 0.7543619871139526,
      "rewards/symbolic_reward_partial_score/std": 0.20396019518375397,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.041762351989746,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 1.016873836517334,
      "step": 1329
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.21161337196826935,
      "epoch": 2.1314102564102564,
      "grad_norm": 0.009917390532791615,
      "learning_rate": 1e-06,
      "loss": 0.025,
      "step": 1330
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.21297961473464966,
      "epoch": 2.1330128205128207,
      "grad_norm": 0.008365034125745296,
      "learning_rate": 1e-06,
      "loss": 0.0207,
      "step": 1331
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.21263594925403595,
      "epoch": 2.1346153846153846,
      "grad_norm": 0.020000530406832695,
      "learning_rate": 1e-06,
      "loss": 0.001,
      "step": 1332
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2631.0,
      "completions/max_terminated_length": 2631.0,
      "completions/mean_length": 1392.521484375,
      "completions/mean_terminated_length": 1392.521484375,
      "completions/min_length": 438.0,
      "completions/min_terminated_length": 438.0,
      "entropy": 0.21844053268432617,
      "epoch": 2.136217948717949,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.02576032653450966,
      "learning_rate": 1e-06,
      "loss": -0.0026,
      "num_tokens": 928662738.0,
      "reward": 0.39745038747787476,
      "reward_std": 0.04422683268785477,
      "rewards/progression_diversity/mean": -8.176280243787915e-05,
      "rewards/progression_diversity/std": 0.001850081025622785,
      "rewards/symbolic_reward_accuracy/mean": 0.2890625,
      "rewards/symbolic_reward_accuracy/std": 0.45377036929130554,
      "rewards/symbolic_reward_partial_score/mean": 0.7467122673988342,
      "rewards/symbolic_reward_partial_score/std": 0.21178719401359558,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0465729236602783,
      "sampling/importance_sampling_ratio/min": 9.539062739349902e-05,
      "sampling/sampling_logp_difference/max": 9.257530212402344,
      "sampling/sampling_logp_difference/mean": 0.09742757678031921,
      "step": 1333
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.22133174538612366,
      "epoch": 2.1378205128205128,
      "grad_norm": 0.012168043293058872,
      "learning_rate": 1e-06,
      "loss": -0.0017,
      "step": 1334
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.22107722610235214,
      "epoch": 2.139423076923077,
      "grad_norm": 0.010279431007802486,
      "learning_rate": 1e-06,
      "loss": -0.0065,
      "step": 1335
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.21761613339185715,
      "epoch": 2.141025641025641,
      "grad_norm": 0.008288303390145302,
      "learning_rate": 1e-06,
      "loss": 0.0063,
      "step": 1336
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2819.0,
      "completions/mean_length": 1633.904296875,
      "completions/mean_terminated_length": 1605.0391845703125,
      "completions/min_length": 408.0,
      "completions/min_terminated_length": 408.0,
      "entropy": 0.20767099410295486,
      "epoch": 2.1426282051282053,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 0.021908177062869072,
      "learning_rate": 1e-06,
      "loss": 0.005,
      "num_tokens": 930332929.0,
      "reward": 0.27803462743759155,
      "reward_std": 0.053543590009212494,
      "rewards/progression_diversity/mean": -0.0002496470115147531,
      "rewards/progression_diversity/std": 0.0035664818715304136,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.661816418170929,
      "rewards/symbolic_reward_partial_score/std": 0.1937265247106552,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0444786548614502,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.3441740572452545,
      "step": 1337
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.20627369731664658,
      "epoch": 2.144230769230769,
      "grad_norm": 0.026881849393248558,
      "learning_rate": 1e-06,
      "loss": 0.0346,
      "step": 1338
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21029096841812134,
      "epoch": 2.1458333333333335,
      "grad_norm": 0.015624100342392921,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "step": 1339
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.21201331168413162,
      "epoch": 2.1474358974358974,
      "grad_norm": 0.023685337975621223,
      "learning_rate": 1e-06,
      "loss": -0.0102,
      "step": 1340
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2974.0,
      "completions/max_terminated_length": 2974.0,
      "completions/mean_length": 1587.693359375,
      "completions/mean_terminated_length": 1587.693359375,
      "completions/min_length": 461.0,
      "completions/min_terminated_length": 461.0,
      "entropy": 0.20493042469024658,
      "epoch": 2.1490384615384617,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.022984595969319344,
      "learning_rate": 1e-06,
      "loss": -0.0027,
      "num_tokens": 931986948.0,
      "reward": 0.4041173458099365,
      "reward_std": 0.039462070912122726,
      "rewards/progression_diversity/mean": -0.0008660133462399244,
      "rewards/progression_diversity/std": 0.006552197504788637,
      "rewards/symbolic_reward_accuracy/mean": 0.296875,
      "rewards/symbolic_reward_accuracy/std": 0.45732781291007996,
      "rewards/symbolic_reward_partial_score/mean": 0.7533365488052368,
      "rewards/symbolic_reward_partial_score/std": 0.21289843320846558,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0439233779907227,
      "sampling/importance_sampling_ratio/min": 0.0010176009964197874,
      "sampling/sampling_logp_difference/max": 6.890307426452637,
      "sampling/sampling_logp_difference/mean": 0.0925874412059784,
      "step": 1341
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.20170185714960098,
      "epoch": 2.1506410256410255,
      "grad_norm": 0.022239655256271362,
      "learning_rate": 1e-06,
      "loss": -0.0016,
      "step": 1342
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2027420625090599,
      "epoch": 2.15224358974359,
      "grad_norm": 0.01715942844748497,
      "learning_rate": 1e-06,
      "loss": 0.0027,
      "step": 1343
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.20686108618974686,
      "epoch": 2.1538461538461537,
      "grad_norm": 0.010147986933588982,
      "learning_rate": 1e-06,
      "loss": 0.0076,
      "step": 1344
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3213.0,
      "completions/max_terminated_length": 3213.0,
      "completions/mean_length": 1698.078125,
      "completions/mean_terminated_length": 1698.078125,
      "completions/min_length": 559.0,
      "completions/min_terminated_length": 559.0,
      "entropy": 0.2128048539161682,
      "epoch": 2.155448717948718,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.016743527725338936,
      "learning_rate": 1e-06,
      "loss": -0.0082,
      "num_tokens": 933690460.0,
      "reward": 0.3048563003540039,
      "reward_std": 0.04834011197090149,
      "rewards/progression_diversity/mean": -0.0007014497532509267,
      "rewards/progression_diversity/std": 0.00812255684286356,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.7115234136581421,
      "rewards/symbolic_reward_partial_score/std": 0.18424317240715027,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0451080799102783,
      "sampling/importance_sampling_ratio/min": 6.933895201655105e-05,
      "sampling/sampling_logp_difference/max": 9.57650375366211,
      "sampling/sampling_logp_difference/mean": 0.09451892971992493,
      "step": 1345
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.21316298097372055,
      "epoch": 2.157051282051282,
      "grad_norm": 0.013251153752207756,
      "learning_rate": 1e-06,
      "loss": 0.0005,
      "step": 1346
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.20967798680067062,
      "epoch": 2.1586538461538463,
      "grad_norm": 0.015155055560171604,
      "learning_rate": 1e-06,
      "loss": 0.0159,
      "step": 1347
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.2121565341949463,
      "epoch": 2.16025641025641,
      "grad_norm": 0.024950312450528145,
      "learning_rate": 1e-06,
      "loss": -0.0028,
      "step": 1348
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3127.0,
      "completions/mean_length": 1755.244140625,
      "completions/mean_terminated_length": 1697.8765869140625,
      "completions/min_length": 394.0,
      "completions/min_terminated_length": 394.0,
      "entropy": 0.20723742246627808,
      "epoch": 2.1618589743589745,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.02118254266679287,
      "learning_rate": 1e-06,
      "loss": 0.0059,
      "num_tokens": 935479065.0,
      "reward": 0.4168354868888855,
      "reward_std": 0.05675492808222771,
      "rewards/progression_diversity/mean": -0.000535505183506757,
      "rewards/progression_diversity/std": 0.0042114765383303165,
      "rewards/symbolic_reward_accuracy/mean": 0.32421875,
      "rewards/symbolic_reward_accuracy/std": 0.4685399830341339,
      "rewards/symbolic_reward_partial_score/mean": 0.7423340082168579,
      "rewards/symbolic_reward_partial_score/std": 0.23071099817752838,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0427265167236328,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.5817426443099976,
      "step": 1349
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.20122719556093216,
      "epoch": 2.1634615384615383,
      "grad_norm": 0.01896592602133751,
      "learning_rate": 1e-06,
      "loss": 0.0248,
      "step": 1350
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.20486315339803696,
      "epoch": 2.1650641025641026,
      "grad_norm": 0.01839040406048298,
      "learning_rate": 1e-06,
      "loss": 0.0417,
      "step": 1351
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.20524092018604279,
      "epoch": 2.1666666666666665,
      "grad_norm": 0.013213125988841057,
      "learning_rate": 1e-06,
      "loss": -0.0134,
      "step": 1352
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3402.0,
      "completions/mean_length": 1791.140625,
      "completions/mean_terminated_length": 1705.1317138671875,
      "completions/min_length": 372.0,
      "completions/min_terminated_length": 372.0,
      "entropy": 0.1993129849433899,
      "epoch": 2.168269230769231,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 0.018554260954260826,
      "learning_rate": 1e-06,
      "loss": 0.0079,
      "num_tokens": 937293713.0,
      "reward": 0.3082357347011566,
      "reward_std": 0.05308583378791809,
      "rewards/progression_diversity/mean": -0.0016233250498771667,
      "rewards/progression_diversity/std": 0.010440926998853683,
      "rewards/symbolic_reward_accuracy/mean": 0.1640625,
      "rewards/symbolic_reward_accuracy/std": 0.37069445848464966,
      "rewards/symbolic_reward_partial_score/mean": 0.70068359375,
      "rewards/symbolic_reward_partial_score/std": 0.1991189867258072,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0414732694625854,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.7726681232452393,
      "step": 1353
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2003878429532051,
      "epoch": 2.1698717948717947,
      "grad_norm": 0.024082843214273453,
      "learning_rate": 1e-06,
      "loss": 0.0203,
      "step": 1354
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.20279623568058014,
      "epoch": 2.171474358974359,
      "grad_norm": 0.02481573261320591,
      "learning_rate": 1e-06,
      "loss": 0.0156,
      "step": 1355
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.20432738959789276,
      "epoch": 2.173076923076923,
      "grad_norm": 0.023944402113556862,
      "learning_rate": 1e-06,
      "loss": 0.0234,
      "step": 1356
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4818.0,
      "completions/mean_length": 1645.9921875,
      "completions/mean_terminated_length": 1617.150634765625,
      "completions/min_length": 515.0,
      "completions/min_terminated_length": 515.0,
      "entropy": 0.20042243599891663,
      "epoch": 2.1746794871794872,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.025058995932340622,
      "learning_rate": 1e-06,
      "loss": -0.0065,
      "num_tokens": 939053501.0,
      "reward": 0.3049207925796509,
      "reward_std": 0.029460538178682327,
      "rewards/progression_diversity/mean": -0.0020646725315600634,
      "rewards/progression_diversity/std": 0.010834318585693836,
      "rewards/symbolic_reward_accuracy/mean": 0.154296875,
      "rewards/symbolic_reward_accuracy/std": 0.36158639192581177,
      "rewards/symbolic_reward_partial_score/mean": 0.7085286378860474,
      "rewards/symbolic_reward_partial_score/std": 0.19462068378925323,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0425053834915161,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.37059780955314636,
      "step": 1357
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.19977743178606033,
      "epoch": 2.176282051282051,
      "grad_norm": 0.01753915846347809,
      "learning_rate": 1e-06,
      "loss": -0.0028,
      "step": 1358
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.19901012629270554,
      "epoch": 2.1778846153846154,
      "grad_norm": 0.020039677619934082,
      "learning_rate": 1e-06,
      "loss": 0.0348,
      "step": 1359
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.1974909007549286,
      "epoch": 2.1794871794871793,
      "grad_norm": 0.016383672133088112,
      "learning_rate": 1e-06,
      "loss": 0.0079,
      "step": 1360
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3294.0,
      "completions/mean_length": 1614.376953125,
      "completions/mean_terminated_length": 1556.4569091796875,
      "completions/min_length": 407.0,
      "completions/min_terminated_length": 407.0,
      "entropy": 0.2029455080628395,
      "epoch": 2.1810897435897436,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 80.58993530273438,
      "learning_rate": 1e-06,
      "loss": 0.0221,
      "num_tokens": 940647406.0,
      "reward": 0.41535690426826477,
      "reward_std": 0.07679130136966705,
      "rewards/progression_diversity/mean": -0.0019082010257989168,
      "rewards/progression_diversity/std": 0.01008455641567707,
      "rewards/symbolic_reward_accuracy/mean": 0.310546875,
      "rewards/symbolic_reward_accuracy/std": 0.46317005157470703,
      "rewards/symbolic_reward_partial_score/mean": 0.7647949457168579,
      "rewards/symbolic_reward_partial_score/std": 0.22581757605075836,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.042673110961914,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.7193973660469055,
      "step": 1361
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.203001469373703,
      "epoch": 2.1826923076923075,
      "grad_norm": 0.021764498203992844,
      "learning_rate": 1e-06,
      "loss": 0.0228,
      "step": 1362
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2056587189435959,
      "epoch": 2.184294871794872,
      "grad_norm": 0.020050769671797752,
      "learning_rate": 1e-06,
      "loss": -0.0023,
      "step": 1363
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.2047322392463684,
      "epoch": 2.185897435897436,
      "grad_norm": 0.02470891922712326,
      "learning_rate": 1e-06,
      "loss": 0.0074,
      "step": 1364
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3347.0,
      "completions/mean_length": 1624.029296875,
      "completions/mean_terminated_length": 1537.035400390625,
      "completions/min_length": 432.0,
      "completions/min_terminated_length": 432.0,
      "entropy": 0.19499259442090988,
      "epoch": 2.1875,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 0.025656770914793015,
      "learning_rate": 1e-06,
      "loss": -0.0055,
      "num_tokens": 942358845.0,
      "reward": 0.33442234992980957,
      "reward_std": 0.039018601179122925,
      "rewards/progression_diversity/mean": -0.0016129279974848032,
      "rewards/progression_diversity/std": 0.011494509875774384,
      "rewards/symbolic_reward_accuracy/mean": 0.1953125,
      "rewards/symbolic_reward_accuracy/std": 0.3968288004398346,
      "rewards/symbolic_reward_partial_score/mean": 0.7261229753494263,
      "rewards/symbolic_reward_partial_score/std": 0.2054896205663681,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0397260189056396,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 0.8769243955612183,
      "step": 1365
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.19767170399427414,
      "epoch": 2.189102564102564,
      "grad_norm": 0.015550960786640644,
      "learning_rate": 1e-06,
      "loss": 0.0246,
      "step": 1366
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.19075892865657806,
      "epoch": 2.190705128205128,
      "grad_norm": 0.024761386215686798,
      "learning_rate": 1e-06,
      "loss": 0.019,
      "step": 1367
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.1886235773563385,
      "epoch": 2.1923076923076925,
      "grad_norm": 0.015679223462939262,
      "learning_rate": 1e-06,
      "loss": 0.0347,
      "step": 1368
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3384.0,
      "completions/mean_length": 1550.146484375,
      "completions/mean_terminated_length": 1521.117431640625,
      "completions/min_length": 358.0,
      "completions/min_terminated_length": 358.0,
      "entropy": 0.1886545717716217,
      "epoch": 2.1939102564102564,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.02213294804096222,
      "learning_rate": 1e-06,
      "loss": 0.025,
      "num_tokens": 944050936.0,
      "reward": 0.3354012668132782,
      "reward_std": 0.0484355203807354,
      "rewards/progression_diversity/mean": -0.0004021337954327464,
      "rewards/progression_diversity/std": 0.0033471970818936825,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.6967936158180237,
      "rewards/symbolic_reward_partial_score/std": 0.24216388165950775,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0405962467193604,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 0.31618547439575195,
      "step": 1369
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.19220811873674393,
      "epoch": 2.1955128205128207,
      "grad_norm": 0.017105696722865105,
      "learning_rate": 1e-06,
      "loss": 0.0032,
      "step": 1370
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.18837008625268936,
      "epoch": 2.1971153846153846,
      "grad_norm": 0.020541273057460785,
      "learning_rate": 1e-06,
      "loss": -0.0059,
      "step": 1371
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.19123337417840958,
      "epoch": 2.198717948717949,
      "grad_norm": 0.016050660982728004,
      "learning_rate": 1e-06,
      "loss": 0.0019,
      "step": 1372
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3055.0,
      "completions/mean_length": 1501.89453125,
      "completions/mean_terminated_length": 1443.533447265625,
      "completions/min_length": 529.0,
      "completions/min_terminated_length": 529.0,
      "entropy": 0.1963193044066429,
      "epoch": 2.2003205128205128,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 839.10888671875,
      "learning_rate": 1e-06,
      "loss": 0.0385,
      "num_tokens": 945680274.0,
      "reward": 0.3143922984600067,
      "reward_std": 0.04647096246480942,
      "rewards/progression_diversity/mean": -0.0016889558173716068,
      "rewards/progression_diversity/std": 0.009619291871786118,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.6782389283180237,
      "rewards/symbolic_reward_partial_score/std": 0.20476947724819183,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0405044555664062,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 0.8936862349510193,
      "step": 1373
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.19663988798856735,
      "epoch": 2.201923076923077,
      "grad_norm": 0.023362474516034126,
      "learning_rate": 1e-06,
      "loss": 0.0048,
      "step": 1374
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.19550593197345734,
      "epoch": 2.203525641025641,
      "grad_norm": 0.01925286464393139,
      "learning_rate": 1e-06,
      "loss": 0.0164,
      "step": 1375
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.198850616812706,
      "epoch": 2.2051282051282053,
      "grad_norm": 0.012360334396362305,
      "learning_rate": 1e-06,
      "loss": 0.0038,
      "step": 1376
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 10325.0,
      "completions/mean_length": 1557.294921875,
      "completions/mean_terminated_length": 1499.151123046875,
      "completions/min_length": 446.0,
      "completions/min_terminated_length": 446.0,
      "entropy": 0.1966118887066841,
      "epoch": 2.206730769230769,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.030109280720353127,
      "learning_rate": 1e-06,
      "loss": -0.0075,
      "num_tokens": 947367161.0,
      "reward": 0.2886817753314972,
      "reward_std": 0.03847378119826317,
      "rewards/progression_diversity/mean": -0.0009640345815569162,
      "rewards/progression_diversity/std": 0.006367819383740425,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.6979818344116211,
      "rewards/symbolic_reward_partial_score/std": 0.1839800328016281,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0392873287200928,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.0182666778564453,
      "step": 1377
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.1928621083498001,
      "epoch": 2.2083333333333335,
      "grad_norm": 0.01684403605759144,
      "learning_rate": 1e-06,
      "loss": 0.0053,
      "step": 1378
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.19299018383026123,
      "epoch": 2.2099358974358974,
      "grad_norm": 0.013669331558048725,
      "learning_rate": 1e-06,
      "loss": 0.0245,
      "step": 1379
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.19235268235206604,
      "epoch": 2.2115384615384617,
      "grad_norm": 0.013778350315988064,
      "learning_rate": 1e-06,
      "loss": 0.0121,
      "step": 1380
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3313.0,
      "completions/mean_length": 1556.845703125,
      "completions/mean_terminated_length": 1527.8297119140625,
      "completions/min_length": 408.0,
      "completions/min_terminated_length": 408.0,
      "entropy": 0.19688106328248978,
      "epoch": 2.2131410256410255,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 0.03251493349671364,
      "learning_rate": 1e-06,
      "loss": -0.0023,
      "num_tokens": 949020410.0,
      "reward": 0.26278871297836304,
      "reward_std": 0.05675513297319412,
      "rewards/progression_diversity/mean": -0.00042738509364426136,
      "rewards/progression_diversity/std": 0.004007916897535324,
      "rewards/symbolic_reward_accuracy/mean": 0.111328125,
      "rewards/symbolic_reward_accuracy/std": 0.31484565138816833,
      "rewards/symbolic_reward_partial_score/mean": 0.6539713144302368,
      "rewards/symbolic_reward_partial_score/std": 0.2113257795572281,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0415072441101074,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.9642093777656555,
      "step": 1381
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2019263282418251,
      "epoch": 2.21474358974359,
      "grad_norm": 0.013695978559553623,
      "learning_rate": 1e-06,
      "loss": 0.0128,
      "step": 1382
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.20185761898756027,
      "epoch": 2.2163461538461537,
      "grad_norm": 0.016754373908042908,
      "learning_rate": 1e-06,
      "loss": -0.0021,
      "step": 1383
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.20480705797672272,
      "epoch": 2.217948717948718,
      "grad_norm": 0.013411487452685833,
      "learning_rate": 1e-06,
      "loss": 0.0017,
      "step": 1384
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3598.0,
      "completions/mean_length": 1521.16015625,
      "completions/mean_terminated_length": 1433.5599365234375,
      "completions/min_length": 459.0,
      "completions/min_terminated_length": 459.0,
      "entropy": 0.21776793152093887,
      "epoch": 2.219551282051282,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 0.022128723561763763,
      "learning_rate": 1e-06,
      "loss": -0.0037,
      "num_tokens": 950676124.0,
      "reward": 0.3249192237854004,
      "reward_std": 0.05313926190137863,
      "rewards/progression_diversity/mean": -0.0017315060831606388,
      "rewards/progression_diversity/std": 0.009827325120568275,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.7178873419761658,
      "rewards/symbolic_reward_partial_score/std": 0.20977161824703217,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0407921075820923,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 2.0792016983032227,
      "step": 1385
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.20703934133052826,
      "epoch": 2.2211538461538463,
      "grad_norm": 0.028047246858477592,
      "learning_rate": 1e-06,
      "loss": 0.0584,
      "step": 1386
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.21588321775197983,
      "epoch": 2.22275641025641,
      "grad_norm": 0.017534522339701653,
      "learning_rate": 1e-06,
      "loss": -0.0009,
      "step": 1387
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.2086542472243309,
      "epoch": 2.2243589743589745,
      "grad_norm": 0.015552827157080173,
      "learning_rate": 1e-06,
      "loss": 0.0135,
      "step": 1388
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3854.0,
      "completions/mean_length": 1460.521484375,
      "completions/mean_terminated_length": 1401.9981689453125,
      "completions/min_length": 463.0,
      "completions/min_terminated_length": 463.0,
      "entropy": 0.2114865928888321,
      "epoch": 2.2259615384615383,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 393.90008544921875,
      "learning_rate": 1e-06,
      "loss": 0.0184,
      "num_tokens": 952301287.0,
      "reward": 0.3778611421585083,
      "reward_std": 0.04598058760166168,
      "rewards/progression_diversity/mean": -0.000995249254629016,
      "rewards/progression_diversity/std": 0.006696970667690039,
      "rewards/symbolic_reward_accuracy/mean": 0.24609375,
      "rewards/symbolic_reward_accuracy/std": 0.4311550557613373,
      "rewards/symbolic_reward_partial_score/mean": 0.7686848640441895,
      "rewards/symbolic_reward_partial_score/std": 0.1856938898563385,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0438477993011475,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.968491792678833,
      "step": 1389
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.22220055013895035,
      "epoch": 2.2275641025641026,
      "grad_norm": 0.007562046870589256,
      "learning_rate": 1e-06,
      "loss": -0.0002,
      "step": 1390
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.21346572041511536,
      "epoch": 2.2291666666666665,
      "grad_norm": 0.02975591830909252,
      "learning_rate": 1e-06,
      "loss": 0.0023,
      "step": 1391
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.20686539262533188,
      "epoch": 2.230769230769231,
      "grad_norm": 0.023594049736857414,
      "learning_rate": 1e-06,
      "loss": 0.0378,
      "step": 1392
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2763.0,
      "completions/mean_length": 1543.576171875,
      "completions/mean_terminated_length": 1456.108154296875,
      "completions/min_length": 386.0,
      "completions/min_terminated_length": 386.0,
      "entropy": 0.21457184851169586,
      "epoch": 2.2323717948717947,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.024636728689074516,
      "learning_rate": 1e-06,
      "loss": -0.0019,
      "num_tokens": 953897950.0,
      "reward": 0.3451923131942749,
      "reward_std": 0.03799974173307419,
      "rewards/progression_diversity/mean": -0.0012781170662492514,
      "rewards/progression_diversity/std": 0.014912668615579605,
      "rewards/symbolic_reward_accuracy/mean": 0.205078125,
      "rewards/symbolic_reward_accuracy/std": 0.4041535556316376,
      "rewards/symbolic_reward_partial_score/mean": 0.7424805164337158,
      "rewards/symbolic_reward_partial_score/std": 0.19341571629047394,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0439865589141846,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.7786634564399719,
      "step": 1393
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.20991724729537964,
      "epoch": 2.233974358974359,
      "grad_norm": 0.018171893432736397,
      "learning_rate": 1e-06,
      "loss": 0.0823,
      "step": 1394
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.21980993449687958,
      "epoch": 2.235576923076923,
      "grad_norm": 0.018304111436009407,
      "learning_rate": 1e-06,
      "loss": -0.0022,
      "step": 1395
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21467426419258118,
      "epoch": 2.2371794871794872,
      "grad_norm": 0.011014263145625591,
      "learning_rate": 1e-06,
      "loss": -0.0085,
      "step": 1396
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 11943.0,
      "completions/mean_length": 1486.80859375,
      "completions/mean_terminated_length": 1428.3883056640625,
      "completions/min_length": 396.0,
      "completions/min_terminated_length": 396.0,
      "entropy": 0.20776310563087463,
      "epoch": 2.238782051282051,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.04726417362689972,
      "learning_rate": 1e-06,
      "loss": 0.0351,
      "num_tokens": 955540620.0,
      "reward": 0.279166579246521,
      "reward_std": 0.017867092043161392,
      "rewards/progression_diversity/mean": -0.001313370536081493,
      "rewards/progression_diversity/std": 0.011600782163441181,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.6819010972976685,
      "rewards/symbolic_reward_partial_score/std": 0.2057180553674698,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0435209274291992,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.23488587141036987,
      "step": 1397
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.20805874466896057,
      "epoch": 2.2403846153846154,
      "grad_norm": 0.020725863054394722,
      "learning_rate": 1e-06,
      "loss": -0.0034,
      "step": 1398
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2029411420226097,
      "epoch": 2.2419871794871793,
      "grad_norm": 0.008639580570161343,
      "learning_rate": 1e-06,
      "loss": 0.0162,
      "step": 1399
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2050759717822075,
      "epoch": 2.2435897435897436,
      "grad_norm": 0.011195927858352661,
      "learning_rate": 1e-06,
      "loss": 0.0212,
      "step": 1400
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3068.0,
      "completions/mean_length": 1376.677734375,
      "completions/mean_terminated_length": 1347.3092041015625,
      "completions/min_length": 396.0,
      "completions/min_terminated_length": 396.0,
      "entropy": 0.21754567325115204,
      "epoch": 2.2451923076923075,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.01782279461622238,
      "learning_rate": 1e-06,
      "loss": -0.0023,
      "num_tokens": 957101943.0,
      "reward": 0.3711217939853668,
      "reward_std": 0.0564521960914135,
      "rewards/progression_diversity/mean": -0.0006168890977278352,
      "rewards/progression_diversity/std": 0.005630532745271921,
      "rewards/symbolic_reward_accuracy/mean": 0.271484375,
      "rewards/symbolic_reward_accuracy/std": 0.44516023993492126,
      "rewards/symbolic_reward_partial_score/mean": 0.6947753429412842,
      "rewards/symbolic_reward_partial_score/std": 0.22683373093605042,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0451213121414185,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 0.8511496782302856,
      "step": 1401
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.21188554912805557,
      "epoch": 2.246794871794872,
      "grad_norm": 0.01593591831624508,
      "learning_rate": 1e-06,
      "loss": 0.0121,
      "step": 1402
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21926379948854446,
      "epoch": 2.248397435897436,
      "grad_norm": 0.013704286888241768,
      "learning_rate": 1e-06,
      "loss": 0.0011,
      "step": 1403
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.21662156283855438,
      "epoch": 2.25,
      "grad_norm": 0.009653382934629917,
      "learning_rate": 1e-06,
      "loss": 0.0036,
      "step": 1404
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3479.0,
      "completions/mean_length": 1432.54296875,
      "completions/mean_terminated_length": 1403.28369140625,
      "completions/min_length": 429.0,
      "completions/min_terminated_length": 429.0,
      "entropy": 0.21846728771924973,
      "epoch": 2.251602564102564,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.02795334719121456,
      "learning_rate": 1e-06,
      "loss": -0.0013,
      "num_tokens": 958673405.0,
      "reward": 0.28600820899009705,
      "reward_std": 0.04315223544836044,
      "rewards/progression_diversity/mean": -0.0007437419844791293,
      "rewards/progression_diversity/std": 0.008263318799436092,
      "rewards/symbolic_reward_accuracy/mean": 0.130859375,
      "rewards/symbolic_reward_accuracy/std": 0.33757632970809937,
      "rewards/symbolic_reward_partial_score/mean": 0.6923177242279053,
      "rewards/symbolic_reward_partial_score/std": 0.18848715722560883,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0477138757705688,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 716.0,
      "sampling/sampling_logp_difference/mean": 0.16357702016830444,
      "step": 1405
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.22100605070590973,
      "epoch": 2.253205128205128,
      "grad_norm": 0.02042258158326149,
      "learning_rate": 1e-06,
      "loss": 0.0307,
      "step": 1406
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.2175752967596054,
      "epoch": 2.2548076923076925,
      "grad_norm": 0.013299060985445976,
      "learning_rate": 1e-06,
      "loss": 0.0054,
      "step": 1407
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2246379554271698,
      "epoch": 2.2564102564102564,
      "grad_norm": 0.01301884651184082,
      "learning_rate": 1e-06,
      "loss": -0.0056,
      "step": 1408
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 6158.0,
      "completions/mean_length": 1540.216796875,
      "completions/mean_terminated_length": 1482.0059814453125,
      "completions/min_length": 498.0,
      "completions/min_terminated_length": 498.0,
      "entropy": 0.22940433025360107,
      "epoch": 2.2580128205128207,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.01926892064511776,
      "learning_rate": 1e-06,
      "loss": -0.0045,
      "num_tokens": 960228028.0,
      "reward": 0.3234546184539795,
      "reward_std": 0.027436789125204086,
      "rewards/progression_diversity/mean": -0.0012210736749693751,
      "rewards/progression_diversity/std": 0.010683650150895119,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.7084310054779053,
      "rewards/symbolic_reward_partial_score/std": 0.20990176498889923,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0487642288208008,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.48923975229263306,
      "step": 1409
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.23073109239339828,
      "epoch": 2.2596153846153846,
      "grad_norm": 0.005647487938404083,
      "learning_rate": 1e-06,
      "loss": -0.0,
      "step": 1410
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2332421839237213,
      "epoch": 2.261217948717949,
      "grad_norm": 0.02655133046209812,
      "learning_rate": 1e-06,
      "loss": 0.0222,
      "step": 1411
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.22827783972024918,
      "epoch": 2.2628205128205128,
      "grad_norm": 0.013446721248328686,
      "learning_rate": 1e-06,
      "loss": 0.0165,
      "step": 1412
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4043.0,
      "completions/mean_length": 1543.96484375,
      "completions/mean_terminated_length": 1456.4990234375,
      "completions/min_length": 332.0,
      "completions/min_terminated_length": 332.0,
      "entropy": 0.22754886746406555,
      "epoch": 2.264423076923077,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 23.43297576904297,
      "learning_rate": 1e-06,
      "loss": 0.028,
      "num_tokens": 961827418.0,
      "reward": 0.3246934413909912,
      "reward_std": 0.04636027663946152,
      "rewards/progression_diversity/mean": -0.0023379838094115257,
      "rewards/progression_diversity/std": 0.023444252088665962,
      "rewards/symbolic_reward_accuracy/mean": 0.17578125,
      "rewards/symbolic_reward_accuracy/std": 0.3810062110424042,
      "rewards/symbolic_reward_partial_score/mean": 0.7321288585662842,
      "rewards/symbolic_reward_partial_score/std": 0.18431484699249268,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0483700037002563,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 1.1377493143081665,
      "step": 1413
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.23179607093334198,
      "epoch": 2.266025641025641,
      "grad_norm": 0.014759178273379803,
      "learning_rate": 1e-06,
      "loss": -0.0051,
      "step": 1414
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.22570616006851196,
      "epoch": 2.2676282051282053,
      "grad_norm": 0.022691620513796806,
      "learning_rate": 1e-06,
      "loss": 0.0214,
      "step": 1415
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.23608656972646713,
      "epoch": 2.269230769230769,
      "grad_norm": 0.010918180458247662,
      "learning_rate": 1e-06,
      "loss": 0.0075,
      "step": 1416
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 4234.0,
      "completions/max_terminated_length": 4234.0,
      "completions/mean_length": 1428.330078125,
      "completions/mean_terminated_length": 1428.330078125,
      "completions/min_length": 361.0,
      "completions/min_terminated_length": 361.0,
      "entropy": 0.2318318784236908,
      "epoch": 2.2708333333333335,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.02078605629503727,
      "learning_rate": 1e-06,
      "loss": -0.0004,
      "num_tokens": 963460099.0,
      "reward": 0.3764505386352539,
      "reward_std": 0.07807396352291107,
      "rewards/progression_diversity/mean": -0.0009436353575438261,
      "rewards/progression_diversity/std": 0.011008831672370434,
      "rewards/symbolic_reward_accuracy/mean": 0.27734375,
      "rewards/symbolic_reward_accuracy/std": 0.4481254518032074,
      "rewards/symbolic_reward_partial_score/mean": 0.7001790404319763,
      "rewards/symbolic_reward_partial_score/std": 0.22067667543888092,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0520446300506592,
      "sampling/importance_sampling_ratio/min": 0.0008081789710558951,
      "sampling/sampling_logp_difference/max": 7.120727062225342,
      "sampling/sampling_logp_difference/mean": 0.10327097773551941,
      "step": 1417
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.23364517837762833,
      "epoch": 2.2724358974358974,
      "grad_norm": 0.018402917310595512,
      "learning_rate": 1e-06,
      "loss": 0.0045,
      "step": 1418
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2324867844581604,
      "epoch": 2.2740384615384617,
      "grad_norm": 0.015934674069285393,
      "learning_rate": 1e-06,
      "loss": 0.0109,
      "step": 1419
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.23535658419132233,
      "epoch": 2.2756410256410255,
      "grad_norm": 0.021948518231511116,
      "learning_rate": 1e-06,
      "loss": -0.0146,
      "step": 1420
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4152.0,
      "completions/mean_length": 1650.921875,
      "completions/mean_terminated_length": 1593.1451416015625,
      "completions/min_length": 475.0,
      "completions/min_terminated_length": 475.0,
      "entropy": 0.23526672273874283,
      "epoch": 2.27724358974359,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 0.023312116041779518,
      "learning_rate": 1e-06,
      "loss": -0.0075,
      "num_tokens": 965069179.0,
      "reward": 0.34999796748161316,
      "reward_std": 0.03697862848639488,
      "rewards/progression_diversity/mean": -0.0016677755629643798,
      "rewards/progression_diversity/std": 0.013735439628362656,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.7461426258087158,
      "rewards/symbolic_reward_partial_score/std": 0.19722376763820648,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.04874849319458,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 1.2209948301315308,
      "step": 1421
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.23266298323869705,
      "epoch": 2.2788461538461537,
      "grad_norm": 0.028419379144906998,
      "learning_rate": 1e-06,
      "loss": -0.0001,
      "step": 1422
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2373318374156952,
      "epoch": 2.280448717948718,
      "grad_norm": 684.584228515625,
      "learning_rate": 1e-06,
      "loss": 0.0291,
      "step": 1423
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.23147784173488617,
      "epoch": 2.282051282051282,
      "grad_norm": 0.015300055034458637,
      "learning_rate": 1e-06,
      "loss": 0.0354,
      "step": 1424
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 7387.0,
      "completions/mean_length": 1647.84765625,
      "completions/mean_terminated_length": 1590.0589599609375,
      "completions/min_length": 459.0,
      "completions/min_terminated_length": 459.0,
      "entropy": 0.22117284685373306,
      "epoch": 2.2836538461538463,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.02695753425359726,
      "learning_rate": 1e-06,
      "loss": 0.0424,
      "num_tokens": 966809885.0,
      "reward": 0.3317036032676697,
      "reward_std": 0.04540080577135086,
      "rewards/progression_diversity/mean": -0.0010283860610798001,
      "rewards/progression_diversity/std": 0.014310967177152634,
      "rewards/symbolic_reward_accuracy/mean": 0.197265625,
      "rewards/symbolic_reward_accuracy/std": 0.3983237147331238,
      "rewards/symbolic_reward_partial_score/mean": 0.7124837040901184,
      "rewards/symbolic_reward_partial_score/std": 0.21933910250663757,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0476967096328735,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 1.099198818206787,
      "step": 1425
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2237677425146103,
      "epoch": 2.28525641025641,
      "grad_norm": 0.013611048460006714,
      "learning_rate": 1e-06,
      "loss": 0.0155,
      "step": 1426
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.22483280301094055,
      "epoch": 2.2868589743589745,
      "grad_norm": 0.02124941535294056,
      "learning_rate": 1e-06,
      "loss": 0.004,
      "step": 1427
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2248641401529312,
      "epoch": 2.2884615384615383,
      "grad_norm": 0.024975134059786797,
      "learning_rate": 1e-06,
      "loss": -0.01,
      "step": 1428
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2827.0,
      "completions/mean_length": 1540.505859375,
      "completions/mean_terminated_length": 1482.296142578125,
      "completions/min_length": 386.0,
      "completions/min_terminated_length": 386.0,
      "entropy": 0.23250573128461838,
      "epoch": 2.2900641025641026,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 823.3074340820312,
      "learning_rate": 1e-06,
      "loss": 0.0196,
      "num_tokens": 968396016.0,
      "reward": 0.2985820770263672,
      "reward_std": 0.015081477351486683,
      "rewards/progression_diversity/mean": -0.0006824900628998876,
      "rewards/progression_diversity/std": 0.005089160054922104,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.7465982437133789,
      "rewards/symbolic_reward_partial_score/std": 0.16918420791625977,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0513556003570557,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.42589616775512695,
      "step": 1429
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.2318640798330307,
      "epoch": 2.2916666666666665,
      "grad_norm": 0.01384742185473442,
      "learning_rate": 1e-06,
      "loss": -0.0038,
      "step": 1430
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.2302088588476181,
      "epoch": 2.293269230769231,
      "grad_norm": 0.015527794137597084,
      "learning_rate": 1e-06,
      "loss": -0.0001,
      "step": 1431
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.23823175579309464,
      "epoch": 2.2948717948717947,
      "grad_norm": 40.3269157409668,
      "learning_rate": 1e-06,
      "loss": 0.0411,
      "step": 1432
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4556.0,
      "completions/mean_length": 1663.361328125,
      "completions/mean_terminated_length": 1576.5992431640625,
      "completions/min_length": 281.0,
      "completions/min_terminated_length": 281.0,
      "entropy": 0.21347655355930328,
      "epoch": 2.296474358974359,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 78.3970947265625,
      "learning_rate": 1e-06,
      "loss": 0.0194,
      "num_tokens": 970189689.0,
      "reward": 0.3578779399394989,
      "reward_std": 0.055181439965963364,
      "rewards/progression_diversity/mean": -0.0007812740514054894,
      "rewards/progression_diversity/std": 0.008964339271187782,
      "rewards/symbolic_reward_accuracy/mean": 0.25390625,
      "rewards/symbolic_reward_accuracy/std": 0.43567025661468506,
      "rewards/symbolic_reward_partial_score/mean": 0.6870931386947632,
      "rewards/symbolic_reward_partial_score/std": 0.22139233350753784,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0451353788375854,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 1.4013971090316772,
      "step": 1433
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.21957285702228546,
      "epoch": 2.298076923076923,
      "grad_norm": 0.013484461233019829,
      "learning_rate": 1e-06,
      "loss": -0.0143,
      "step": 1434
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.21703750640153885,
      "epoch": 2.2996794871794872,
      "grad_norm": 0.030732089653611183,
      "learning_rate": 1e-06,
      "loss": 0.0609,
      "step": 1435
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.22398217022418976,
      "epoch": 2.301282051282051,
      "grad_norm": 0.021893825381994247,
      "learning_rate": 1e-06,
      "loss": 0.0065,
      "step": 1436
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3516.0,
      "completions/mean_length": 1687.796875,
      "completions/mean_terminated_length": 1601.1788330078125,
      "completions/min_length": 342.0,
      "completions/min_terminated_length": 342.0,
      "entropy": 0.21840395033359528,
      "epoch": 2.3028846153846154,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 47.56247329711914,
      "learning_rate": 1e-06,
      "loss": 0.0497,
      "num_tokens": 971978193.0,
      "reward": 0.3468187153339386,
      "reward_std": 0.043784305453300476,
      "rewards/progression_diversity/mean": -0.0017242016037926078,
      "rewards/progression_diversity/std": 0.017134975641965866,
      "rewards/symbolic_reward_accuracy/mean": 0.21875,
      "rewards/symbolic_reward_accuracy/std": 0.41380295157432556,
      "rewards/symbolic_reward_partial_score/mean": 0.7205728888511658,
      "rewards/symbolic_reward_partial_score/std": 0.2209366112947464,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0465869903564453,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 1.2852548360824585,
      "step": 1437
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.22310731559991837,
      "epoch": 2.3044871794871793,
      "grad_norm": 0.029605181887745857,
      "learning_rate": 1e-06,
      "loss": -0.0095,
      "step": 1438
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.21875733882188797,
      "epoch": 2.3060897435897436,
      "grad_norm": 0.022749239578843117,
      "learning_rate": 1e-06,
      "loss": 0.0269,
      "step": 1439
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.22598165273666382,
      "epoch": 2.3076923076923075,
      "grad_norm": 0.014465555548667908,
      "learning_rate": 1e-06,
      "loss": 0.0126,
      "step": 1440
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3771.0,
      "completions/mean_length": 1645.763671875,
      "completions/mean_terminated_length": 1616.9217529296875,
      "completions/min_length": 373.0,
      "completions/min_terminated_length": 373.0,
      "entropy": 0.22972248494625092,
      "epoch": 2.309294871794872,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.02680191770195961,
      "learning_rate": 1e-06,
      "loss": -0.0029,
      "num_tokens": 973626632.0,
      "reward": 0.3762263059616089,
      "reward_std": 0.05338224023580551,
      "rewards/progression_diversity/mean": -0.0018829565960913897,
      "rewards/progression_diversity/std": 0.017473464831709862,
      "rewards/symbolic_reward_accuracy/mean": 0.25,
      "rewards/symbolic_reward_accuracy/std": 0.43343618512153625,
      "rewards/symbolic_reward_partial_score/mean": 0.754150390625,
      "rewards/symbolic_reward_partial_score/std": 0.19963204860687256,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0482197999954224,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 1.3025373220443726,
      "step": 1441
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.22302179783582687,
      "epoch": 2.310897435897436,
      "grad_norm": 0.016930239275097847,
      "learning_rate": 1e-06,
      "loss": 0.01,
      "step": 1442
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.22931306064128876,
      "epoch": 2.3125,
      "grad_norm": 0.017596367746591568,
      "learning_rate": 1e-06,
      "loss": 0.0018,
      "step": 1443
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2299589440226555,
      "epoch": 2.314102564102564,
      "grad_norm": 0.01738305389881134,
      "learning_rate": 1e-06,
      "loss": -0.0058,
      "step": 1444
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 5412.0,
      "completions/max_terminated_length": 5412.0,
      "completions/mean_length": 1501.36328125,
      "completions/mean_terminated_length": 1501.36328125,
      "completions/min_length": 367.0,
      "completions/min_terminated_length": 367.0,
      "entropy": 0.23561769723892212,
      "epoch": 2.315705128205128,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.024762306362390518,
      "learning_rate": 1e-06,
      "loss": 0.0078,
      "num_tokens": 975200130.0,
      "reward": 0.39472496509552,
      "reward_std": 0.027176737785339355,
      "rewards/progression_diversity/mean": -0.0006506302743218839,
      "rewards/progression_diversity/std": 0.007446780800819397,
      "rewards/symbolic_reward_accuracy/mean": 0.27734375,
      "rewards/symbolic_reward_accuracy/std": 0.4481254518032074,
      "rewards/symbolic_reward_partial_score/mean": 0.7610839605331421,
      "rewards/symbolic_reward_partial_score/std": 0.19756866991519928,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0539369583129883,
      "sampling/importance_sampling_ratio/min": 4.62967858494423e-12,
      "sampling/sampling_logp_difference/max": 26.098533630371094,
      "sampling/sampling_logp_difference/mean": 0.10675007849931717,
      "step": 1445
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.24369582533836365,
      "epoch": 2.3173076923076925,
      "grad_norm": 0.02469770982861519,
      "learning_rate": 1e-06,
      "loss": -0.002,
      "step": 1446
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.24448219686746597,
      "epoch": 2.3189102564102564,
      "grad_norm": 0.018661778420209885,
      "learning_rate": 1e-06,
      "loss": -0.0009,
      "step": 1447
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.24316756427288055,
      "epoch": 2.3205128205128207,
      "grad_norm": 0.010404370725154877,
      "learning_rate": 1e-06,
      "loss": 0.0039,
      "step": 1448
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3668.0,
      "completions/mean_length": 1508.080078125,
      "completions/mean_terminated_length": 1449.7431640625,
      "completions/min_length": 391.0,
      "completions/min_terminated_length": 391.0,
      "entropy": 0.24711936712265015,
      "epoch": 2.3221153846153846,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.029730960726737976,
      "learning_rate": 1e-06,
      "loss": 0.0011,
      "num_tokens": 976758619.0,
      "reward": 0.36418893933296204,
      "reward_std": 0.02938855066895485,
      "rewards/progression_diversity/mean": -0.0005412165191955864,
      "rewards/progression_diversity/std": 0.006034459453076124,
      "rewards/symbolic_reward_accuracy/mean": 0.236328125,
      "rewards/symbolic_reward_accuracy/std": 0.42524150013923645,
      "rewards/symbolic_reward_partial_score/mean": 0.742626965045929,
      "rewards/symbolic_reward_partial_score/std": 0.20223668217658997,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0534663200378418,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.6994922161102295,
      "step": 1449
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.24198869615793228,
      "epoch": 2.323717948717949,
      "grad_norm": 0.018425533547997475,
      "learning_rate": 1e-06,
      "loss": 0.0282,
      "step": 1450
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.24656683206558228,
      "epoch": 2.3253205128205128,
      "grad_norm": 0.013853712007403374,
      "learning_rate": 1e-06,
      "loss": -0.0072,
      "step": 1451
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.24908402562141418,
      "epoch": 2.326923076923077,
      "grad_norm": 0.01092120073735714,
      "learning_rate": 1e-06,
      "loss": 0.0053,
      "step": 1452
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4055.0,
      "completions/mean_length": 1472.052734375,
      "completions/mean_terminated_length": 1442.870849609375,
      "completions/min_length": 435.0,
      "completions/min_terminated_length": 435.0,
      "entropy": 0.2532293498516083,
      "epoch": 2.328525641025641,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.021764414384961128,
      "learning_rate": 1e-06,
      "loss": -0.0082,
      "num_tokens": 978359622.0,
      "reward": 0.3215084671974182,
      "reward_std": 0.039750613272190094,
      "rewards/progression_diversity/mean": -0.0005217455327510834,
      "rewards/progression_diversity/std": 0.008216914720833302,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.7129882574081421,
      "rewards/symbolic_reward_partial_score/std": 0.21137480437755585,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0538359880447388,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.4409805238246918,
      "step": 1453
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.24427498131990433,
      "epoch": 2.3301282051282053,
      "grad_norm": 0.015310117974877357,
      "learning_rate": 1e-06,
      "loss": 0.004,
      "step": 1454
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2432166486978531,
      "epoch": 2.331730769230769,
      "grad_norm": 895.1507568359375,
      "learning_rate": 1e-06,
      "loss": 0.046,
      "step": 1455
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.24313075095415115,
      "epoch": 2.3333333333333335,
      "grad_norm": 0.013614215888082981,
      "learning_rate": 1e-06,
      "loss": -0.002,
      "step": 1456
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3000.0,
      "completions/mean_length": 1463.748046875,
      "completions/mean_terminated_length": 1434.5499267578125,
      "completions/min_length": 353.0,
      "completions/min_terminated_length": 353.0,
      "entropy": 0.2332357093691826,
      "epoch": 2.3349358974358974,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.02649446576833725,
      "learning_rate": 1e-06,
      "loss": -0.0052,
      "num_tokens": 979968421.0,
      "reward": 0.33722105622291565,
      "reward_std": 0.0333821102976799,
      "rewards/progression_diversity/mean": -0.0005536978715099394,
      "rewards/progression_diversity/std": 0.006957771256566048,
      "rewards/symbolic_reward_accuracy/mean": 0.208984375,
      "rewards/symbolic_reward_accuracy/std": 0.40698084235191345,
      "rewards/symbolic_reward_partial_score/mean": 0.7067708373069763,
      "rewards/symbolic_reward_partial_score/std": 0.19518350064754486,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.05262291431427,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.4595038890838623,
      "step": 1457
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.240996353328228,
      "epoch": 2.3365384615384617,
      "grad_norm": 0.00819353200495243,
      "learning_rate": 1e-06,
      "loss": 0.0182,
      "step": 1458
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.23650028556585312,
      "epoch": 2.3381410256410255,
      "grad_norm": 0.021920593455433846,
      "learning_rate": 1e-06,
      "loss": 0.017,
      "step": 1459
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1171875,
      "entropy": 0.2370261400938034,
      "epoch": 2.33974358974359,
      "grad_norm": 0.011823596432805061,
      "learning_rate": 1e-06,
      "loss": 0.001,
      "step": 1460
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5619.0,
      "completions/mean_length": 1463.85546875,
      "completions/mean_terminated_length": 1375.91748046875,
      "completions/min_length": 301.0,
      "completions/min_terminated_length": 301.0,
      "entropy": 0.22917187213897705,
      "epoch": 2.3413461538461537,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 196.0175018310547,
      "learning_rate": 1e-06,
      "loss": 0.0506,
      "num_tokens": 981567035.0,
      "reward": 0.39167657494544983,
      "reward_std": 0.050342485308647156,
      "rewards/progression_diversity/mean": -0.0008015389903448522,
      "rewards/progression_diversity/std": 0.0074392156675457954,
      "rewards/symbolic_reward_accuracy/mean": 0.2734375,
      "rewards/symbolic_reward_accuracy/std": 0.4461594223976135,
      "rewards/symbolic_reward_partial_score/mean": 0.7600423097610474,
      "rewards/symbolic_reward_partial_score/std": 0.2186768352985382,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0496225357055664,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 740.0,
      "sampling/sampling_logp_difference/mean": 0.7763689756393433,
      "step": 1461
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.23547165840864182,
      "epoch": 2.342948717948718,
      "grad_norm": 0.014117385260760784,
      "learning_rate": 1e-06,
      "loss": 0.0003,
      "step": 1462
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2310246080160141,
      "epoch": 2.344551282051282,
      "grad_norm": 0.038323234766721725,
      "learning_rate": 1e-06,
      "loss": 0.0181,
      "step": 1463
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2359386458992958,
      "epoch": 2.3461538461538463,
      "grad_norm": 0.018555141985416412,
      "learning_rate": 1e-06,
      "loss": 0.0034,
      "step": 1464
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3024.0,
      "completions/mean_length": 1395.6953125,
      "completions/mean_terminated_length": 1336.917724609375,
      "completions/min_length": 313.0,
      "completions/min_terminated_length": 313.0,
      "entropy": 0.23501623421907425,
      "epoch": 2.34775641025641,
      "frac_reward_zero_std": 0.625,
      "grad_norm": 378.42333984375,
      "learning_rate": 1e-06,
      "loss": 0.0315,
      "num_tokens": 983101327.0,
      "reward": 0.3867397904396057,
      "reward_std": 0.011792978271842003,
      "rewards/progression_diversity/mean": -0.00033900359994731843,
      "rewards/progression_diversity/std": 0.004276837222278118,
      "rewards/symbolic_reward_accuracy/mean": 0.248046875,
      "rewards/symbolic_reward_accuracy/std": 0.4323015511035919,
      "rewards/symbolic_reward_partial_score/mean": 0.7943521738052368,
      "rewards/symbolic_reward_partial_score/std": 0.1727304607629776,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0502277612686157,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 0.6608302593231201,
      "step": 1465
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.22964345663785934,
      "epoch": 2.3493589743589745,
      "grad_norm": 0.007282482460141182,
      "learning_rate": 1e-06,
      "loss": 0.0247,
      "step": 1466
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.23587816208600998,
      "epoch": 2.3509615384615383,
      "grad_norm": 0.02002773992717266,
      "learning_rate": 1e-06,
      "loss": 0.0049,
      "step": 1467
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.23493095487356186,
      "epoch": 2.3525641025641026,
      "grad_norm": 0.007156542036682367,
      "learning_rate": 1e-06,
      "loss": -0.0054,
      "step": 1468
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2754.0,
      "completions/mean_length": 1412.869140625,
      "completions/mean_terminated_length": 1354.158935546875,
      "completions/min_length": 294.0,
      "completions/min_terminated_length": 294.0,
      "entropy": 0.23347257822752,
      "epoch": 2.3541666666666665,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.020152533426880836,
      "learning_rate": 1e-06,
      "loss": 0.0307,
      "num_tokens": 984703036.0,
      "reward": 0.3319286108016968,
      "reward_std": 0.04241711646318436,
      "rewards/progression_diversity/mean": -0.0009903897298499942,
      "rewards/progression_diversity/std": 0.011630616150796413,
      "rewards/symbolic_reward_accuracy/mean": 0.189453125,
      "rewards/symbolic_reward_accuracy/std": 0.3922513723373413,
      "rewards/symbolic_reward_partial_score/mean": 0.7282063961029053,
      "rewards/symbolic_reward_partial_score/std": 0.18158793449401855,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0504162311553955,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 736.0,
      "sampling/sampling_logp_difference/mean": 0.44282588362693787,
      "step": 1469
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.23258361965417862,
      "epoch": 2.355769230769231,
      "grad_norm": 0.016104018315672874,
      "learning_rate": 1e-06,
      "loss": -0.0002,
      "step": 1470
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.22412847727537155,
      "epoch": 2.3573717948717947,
      "grad_norm": 0.009414435364305973,
      "learning_rate": 1e-06,
      "loss": 0.0275,
      "step": 1471
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.2289385423064232,
      "epoch": 2.358974358974359,
      "grad_norm": 0.014195187948644161,
      "learning_rate": 1e-06,
      "loss": -0.0008,
      "step": 1472
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2557.0,
      "completions/max_terminated_length": 2557.0,
      "completions/mean_length": 1360.466796875,
      "completions/mean_terminated_length": 1360.466796875,
      "completions/min_length": 358.0,
      "completions/min_terminated_length": 358.0,
      "entropy": 0.22527696937322617,
      "epoch": 2.360576923076923,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.013988327234983444,
      "learning_rate": 1e-06,
      "loss": -0.0092,
      "num_tokens": 986369547.0,
      "reward": 0.31377702951431274,
      "reward_std": 0.011640192940831184,
      "rewards/progression_diversity/mean": -0.00022870188695378602,
      "rewards/progression_diversity/std": 0.0025177807547152042,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.6709309816360474,
      "rewards/symbolic_reward_partial_score/std": 0.20638255774974823,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.050323247909546,
      "sampling/importance_sampling_ratio/min": 0.003235085867345333,
      "sampling/sampling_logp_difference/max": 5.733699798583984,
      "sampling/sampling_logp_difference/mean": 0.10156988352537155,
      "step": 1473
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.2217569351196289,
      "epoch": 2.3621794871794872,
      "grad_norm": 0.02185184508562088,
      "learning_rate": 1e-06,
      "loss": 0.0026,
      "step": 1474
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.22397807240486145,
      "epoch": 2.363782051282051,
      "grad_norm": 0.018456028774380684,
      "learning_rate": 1e-06,
      "loss": 0.0056,
      "step": 1475
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.22414569556713104,
      "epoch": 2.3653846153846154,
      "grad_norm": 0.012643019668757915,
      "learning_rate": 1e-06,
      "loss": -0.0055,
      "step": 1476
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2527.0,
      "completions/mean_length": 1278.501953125,
      "completions/mean_terminated_length": 1189.4715576171875,
      "completions/min_length": 310.0,
      "completions/min_terminated_length": 310.0,
      "entropy": 0.2367832437157631,
      "epoch": 2.3669871794871793,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.01879701018333435,
      "learning_rate": 1e-06,
      "loss": 0.0067,
      "num_tokens": 987864604.0,
      "reward": 0.43188178539276123,
      "reward_std": 0.03625435382127762,
      "rewards/progression_diversity/mean": -0.00029982085106894374,
      "rewards/progression_diversity/std": 0.004407213069498539,
      "rewards/symbolic_reward_accuracy/mean": 0.330078125,
      "rewards/symbolic_reward_accuracy/std": 0.47070086002349854,
      "rewards/symbolic_reward_partial_score/mean": 0.7814127206802368,
      "rewards/symbolic_reward_partial_score/std": 0.21501034498214722,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0473359823226929,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 744.0,
      "sampling/sampling_logp_difference/mean": 2.7259652614593506,
      "step": 1477
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.2279703989624977,
      "epoch": 2.3685897435897436,
      "grad_norm": 30.450448989868164,
      "learning_rate": 1e-06,
      "loss": 0.0318,
      "step": 1478
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.23828475177288055,
      "epoch": 2.3701923076923075,
      "grad_norm": 0.024164684116840363,
      "learning_rate": 1e-06,
      "loss": -0.0028,
      "step": 1479
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.23449359089136124,
      "epoch": 2.371794871794872,
      "grad_norm": 0.01611475460231304,
      "learning_rate": 1e-06,
      "loss": 0.0146,
      "step": 1480
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3828.0,
      "completions/mean_length": 1289.505859375,
      "completions/mean_terminated_length": 1230.3118896484375,
      "completions/min_length": 347.0,
      "completions/min_terminated_length": 347.0,
      "entropy": 0.23153682053089142,
      "epoch": 2.373397435897436,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.02150103822350502,
      "learning_rate": 1e-06,
      "loss": -0.0002,
      "num_tokens": 989388479.0,
      "reward": 0.41440147161483765,
      "reward_std": 0.038612596690654755,
      "rewards/progression_diversity/mean": -0.0002827388816513121,
      "rewards/progression_diversity/std": 0.003904320765286684,
      "rewards/symbolic_reward_accuracy/mean": 0.306640625,
      "rewards/symbolic_reward_accuracy/std": 0.4615498185157776,
      "rewards/symbolic_reward_partial_score/mean": 0.7693685293197632,
      "rewards/symbolic_reward_partial_score/std": 0.21836112439632416,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0490467548370361,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 748.0,
      "sampling/sampling_logp_difference/mean": 1.6480077505111694,
      "step": 1481
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2258523479104042,
      "epoch": 2.375,
      "grad_norm": 0.009424294345080853,
      "learning_rate": 1e-06,
      "loss": 0.1124,
      "step": 1482
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.23436233401298523,
      "epoch": 2.376602564102564,
      "grad_norm": 0.010978851467370987,
      "learning_rate": 1e-06,
      "loss": 0.0041,
      "step": 1483
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2368711531162262,
      "epoch": 2.378205128205128,
      "grad_norm": 0.015040039084851742,
      "learning_rate": 1e-06,
      "loss": -0.0046,
      "step": 1484
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3070.0,
      "completions/mean_length": 1375.23046875,
      "completions/mean_terminated_length": 1286.7701416015625,
      "completions/min_length": 292.0,
      "completions/min_terminated_length": 292.0,
      "entropy": 0.22265422344207764,
      "epoch": 2.3798076923076925,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 365.71807861328125,
      "learning_rate": 1e-06,
      "loss": 0.0797,
      "num_tokens": 990941957.0,
      "reward": 0.2840445041656494,
      "reward_std": 0.026247035712003708,
      "rewards/progression_diversity/mean": -0.00033585538039915264,
      "rewards/progression_diversity/std": 0.0034242472611367702,
      "rewards/symbolic_reward_accuracy/mean": 0.12109375,
      "rewards/symbolic_reward_accuracy/std": 0.3265552520751953,
      "rewards/symbolic_reward_partial_score/mean": 0.706591784954071,
      "rewards/symbolic_reward_partial_score/std": 0.19762486219406128,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0476056337356567,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 748.0,
      "sampling/sampling_logp_difference/mean": 2.300495147705078,
      "step": 1485
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2373371571302414,
      "epoch": 2.3814102564102564,
      "grad_norm": 0.01105690747499466,
      "learning_rate": 1e-06,
      "loss": 0.0004,
      "step": 1486
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.23985521495342255,
      "epoch": 2.3830128205128207,
      "grad_norm": 0.015546144917607307,
      "learning_rate": 1e-06,
      "loss": -0.0024,
      "step": 1487
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.23665790259838104,
      "epoch": 2.3846153846153846,
      "grad_norm": 0.01926671527326107,
      "learning_rate": 1e-06,
      "loss": -0.0068,
      "step": 1488
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2707.0,
      "completions/mean_length": 1406.267578125,
      "completions/mean_terminated_length": 1288.3326416015625,
      "completions/min_length": 349.0,
      "completions/min_terminated_length": 349.0,
      "entropy": 0.23878996819257736,
      "epoch": 2.386217948717949,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.019514024257659912,
      "learning_rate": 1e-06,
      "loss": 0.0072,
      "num_tokens": 992456558.0,
      "reward": 0.2826894521713257,
      "reward_std": 0.03426457941532135,
      "rewards/progression_diversity/mean": -0.0005875998758710921,
      "rewards/progression_diversity/std": 0.00664177630096674,
      "rewards/symbolic_reward_accuracy/mean": 0.12109375,
      "rewards/symbolic_reward_accuracy/std": 0.3265552520751953,
      "rewards/symbolic_reward_partial_score/mean": 0.7027344107627869,
      "rewards/symbolic_reward_partial_score/std": 0.1787167340517044,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0489890575408936,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 2.2175588607788086,
      "step": 1489
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.24255738407373428,
      "epoch": 2.3878205128205128,
      "grad_norm": 2492.634033203125,
      "learning_rate": 1e-06,
      "loss": 0.0505,
      "step": 1490
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.24588491022586823,
      "epoch": 2.389423076923077,
      "grad_norm": 0.017873678356409073,
      "learning_rate": 1e-06,
      "loss": -0.0036,
      "step": 1491
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.24287723004817963,
      "epoch": 2.391025641025641,
      "grad_norm": 0.013864979147911072,
      "learning_rate": 1e-06,
      "loss": 0.0298,
      "step": 1492
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3078.0,
      "completions/mean_length": 1381.619140625,
      "completions/mean_terminated_length": 1322.786376953125,
      "completions/min_length": 383.0,
      "completions/min_terminated_length": 383.0,
      "entropy": 0.23624292761087418,
      "epoch": 2.3926282051282053,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.01766468957066536,
      "learning_rate": 1e-06,
      "loss": -0.0018,
      "num_tokens": 994038987.0,
      "reward": 0.3370998203754425,
      "reward_std": 0.02162216417491436,
      "rewards/progression_diversity/mean": -0.000469559890916571,
      "rewards/progression_diversity/std": 0.005131890531629324,
      "rewards/symbolic_reward_accuracy/mean": 0.17578125,
      "rewards/symbolic_reward_accuracy/std": 0.3810062110424042,
      "rewards/symbolic_reward_partial_score/mean": 0.7734212279319763,
      "rewards/symbolic_reward_partial_score/std": 0.15896020829677582,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0494273900985718,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 1.574073076248169,
      "step": 1493
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.23203369975090027,
      "epoch": 2.394230769230769,
      "grad_norm": 0.007126044947654009,
      "learning_rate": 1e-06,
      "loss": 0.0539,
      "step": 1494
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.23312392085790634,
      "epoch": 2.3958333333333335,
      "grad_norm": 0.007795875892043114,
      "learning_rate": 1e-06,
      "loss": -0.0035,
      "step": 1495
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.23666821420192719,
      "epoch": 2.3974358974358974,
      "grad_norm": 0.007526164874434471,
      "learning_rate": 1e-06,
      "loss": 0.0024,
      "step": 1496
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2845.0,
      "completions/mean_length": 1428.548828125,
      "completions/mean_terminated_length": 1281.0592041015625,
      "completions/min_length": 359.0,
      "completions/min_terminated_length": 359.0,
      "entropy": 0.23371727764606476,
      "epoch": 2.3990384615384617,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 837.6632080078125,
      "learning_rate": 1e-06,
      "loss": 0.0729,
      "num_tokens": 995634948.0,
      "reward": 0.3479752242565155,
      "reward_std": 0.019150175154209137,
      "rewards/progression_diversity/mean": -0.00033024564618244767,
      "rewards/progression_diversity/std": 0.004327813629060984,
      "rewards/symbolic_reward_accuracy/mean": 0.216796875,
      "rewards/symbolic_reward_accuracy/std": 0.4124660789966583,
      "rewards/symbolic_reward_partial_score/mean": 0.7295898199081421,
      "rewards/symbolic_reward_partial_score/std": 0.2091808170080185,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.046621322631836,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 3.1054999828338623,
      "step": 1497
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.23887716233730316,
      "epoch": 2.4006410256410255,
      "grad_norm": 7.230476379394531,
      "learning_rate": 1e-06,
      "loss": 0.0044,
      "step": 1498
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.24147003889083862,
      "epoch": 2.40224358974359,
      "grad_norm": 0.013911988586187363,
      "learning_rate": 1e-06,
      "loss": -0.0055,
      "step": 1499
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.24635042250156403,
      "epoch": 2.4038461538461537,
      "grad_norm": 0.013229484669864178,
      "learning_rate": 1e-06,
      "loss": 0.023,
      "step": 1500
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5214.0,
      "completions/mean_length": 1461.9140625,
      "completions/mean_terminated_length": 1314.75341796875,
      "completions/min_length": 430.0,
      "completions/min_terminated_length": 430.0,
      "entropy": 0.2452658787369728,
      "epoch": 2.405448717948718,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 125.90882873535156,
      "learning_rate": 1e-06,
      "loss": 0.0306,
      "num_tokens": 997203848.0,
      "reward": 0.2770632207393646,
      "reward_std": 0.023135796189308167,
      "rewards/progression_diversity/mean": -0.0007097757770679891,
      "rewards/progression_diversity/std": 0.01300110761076212,
      "rewards/symbolic_reward_accuracy/mean": 0.12109375,
      "rewards/symbolic_reward_accuracy/std": 0.3265552520751953,
      "rewards/symbolic_reward_partial_score/mean": 0.6839843988418579,
      "rewards/symbolic_reward_partial_score/std": 0.1883450597524643,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0471558570861816,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 2.527804374694824,
      "step": 1501
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.23859861493110657,
      "epoch": 2.407051282051282,
      "grad_norm": 577.0975952148438,
      "learning_rate": 1e-06,
      "loss": 0.0477,
      "step": 1502
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.2435135617852211,
      "epoch": 2.4086538461538463,
      "grad_norm": 0.014631603844463825,
      "learning_rate": 1e-06,
      "loss": 0.0252,
      "step": 1503
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2454121932387352,
      "epoch": 2.41025641025641,
      "grad_norm": 0.008022090420126915,
      "learning_rate": 1e-06,
      "loss": 0.0225,
      "step": 1504
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4812.0,
      "completions/mean_length": 1448.013671875,
      "completions/mean_terminated_length": 1330.407470703125,
      "completions/min_length": 309.0,
      "completions/min_terminated_length": 309.0,
      "entropy": 0.24072068184614182,
      "epoch": 2.4118589743589745,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.024386148899793625,
      "learning_rate": 1e-06,
      "loss": -0.0065,
      "num_tokens": 998845663.0,
      "reward": 0.31266647577285767,
      "reward_std": 0.029008327051997185,
      "rewards/progression_diversity/mean": -0.00044497710769064724,
      "rewards/progression_diversity/std": 0.007153489161282778,
      "rewards/symbolic_reward_accuracy/mean": 0.166015625,
      "rewards/symbolic_reward_accuracy/std": 0.3724585771560669,
      "rewards/symbolic_reward_partial_score/mean": 0.7128092646598816,
      "rewards/symbolic_reward_partial_score/std": 0.1801445633172989,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0494557619094849,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 1.1010427474975586,
      "step": 1505
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.22912582755088806,
      "epoch": 2.4134615384615383,
      "grad_norm": 0.013357684947550297,
      "learning_rate": 1e-06,
      "loss": 0.0729,
      "step": 1506
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.238428495824337,
      "epoch": 2.4150641025641026,
      "grad_norm": 0.02128361538052559,
      "learning_rate": 1e-06,
      "loss": 0.0199,
      "step": 1507
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.23703129589557648,
      "epoch": 2.4166666666666665,
      "grad_norm": 0.011308502405881882,
      "learning_rate": 1e-06,
      "loss": 0.0045,
      "step": 1508
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2632.0,
      "completions/mean_length": 1454.595703125,
      "completions/mean_terminated_length": 1366.6031494140625,
      "completions/min_length": 373.0,
      "completions/min_terminated_length": 373.0,
      "entropy": 0.23324482142925262,
      "epoch": 2.418269230769231,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.020565425977110863,
      "learning_rate": 1e-06,
      "loss": 0.0137,
      "num_tokens": 1000476448.0,
      "reward": 0.35293227434158325,
      "reward_std": 0.02206794172525406,
      "rewards/progression_diversity/mean": -0.00022909138351678848,
      "rewards/progression_diversity/std": 0.0050324564799666405,
      "rewards/symbolic_reward_accuracy/mean": 0.21875,
      "rewards/symbolic_reward_accuracy/std": 0.41380295157432556,
      "rewards/symbolic_reward_partial_score/mean": 0.7409017086029053,
      "rewards/symbolic_reward_partial_score/std": 0.1797100305557251,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.047702431678772,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 756.0,
      "sampling/sampling_logp_difference/mean": 2.1723339557647705,
      "step": 1509
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2380816489458084,
      "epoch": 2.4198717948717947,
      "grad_norm": 0.009358495473861694,
      "learning_rate": 1e-06,
      "loss": 0.0104,
      "step": 1510
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.23932605236768723,
      "epoch": 2.421474358974359,
      "grad_norm": 0.012663176283240318,
      "learning_rate": 1e-06,
      "loss": 0.033,
      "step": 1511
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.23995714634656906,
      "epoch": 2.423076923076923,
      "grad_norm": 0.008335144259035587,
      "learning_rate": 1e-06,
      "loss": -0.0006,
      "step": 1512
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2968.0,
      "completions/mean_length": 1581.033203125,
      "completions/mean_terminated_length": 1375.8436279296875,
      "completions/min_length": 429.0,
      "completions/min_terminated_length": 429.0,
      "entropy": 0.23756451904773712,
      "epoch": 2.4246794871794872,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 581.8189697265625,
      "learning_rate": 1e-06,
      "loss": 0.0544,
      "num_tokens": 1002237585.0,
      "reward": 0.3079501986503601,
      "reward_std": 0.04506230354309082,
      "rewards/progression_diversity/mean": -0.0013673059875145555,
      "rewards/progression_diversity/std": 0.019772524014115334,
      "rewards/symbolic_reward_accuracy/mean": 0.166015625,
      "rewards/symbolic_reward_accuracy/std": 0.3724585771560669,
      "rewards/symbolic_reward_partial_score/mean": 0.6990722417831421,
      "rewards/symbolic_reward_partial_score/std": 0.19391867518424988,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0470722913742065,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 2.45670747756958,
      "step": 1513
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2470928058028221,
      "epoch": 2.426282051282051,
      "grad_norm": 0.01648869179189205,
      "learning_rate": 1e-06,
      "loss": -0.0059,
      "step": 1514
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.23297542333602905,
      "epoch": 2.4278846153846154,
      "grad_norm": 384.6247253417969,
      "learning_rate": 1e-06,
      "loss": 0.0991,
      "step": 1515
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.23933256417512894,
      "epoch": 2.4294871794871793,
      "grad_norm": 0.019361699000000954,
      "learning_rate": 1e-06,
      "loss": 0.0483,
      "step": 1516
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 8324.0,
      "completions/mean_length": 1477.720703125,
      "completions/mean_terminated_length": 1360.348388671875,
      "completions/min_length": 362.0,
      "completions/min_terminated_length": 362.0,
      "entropy": 0.2378390207886696,
      "epoch": 2.4310897435897436,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.024526890367269516,
      "learning_rate": 1e-06,
      "loss": 0.0322,
      "num_tokens": 1003883634.0,
      "reward": 0.3165303170681,
      "reward_std": 0.02786339819431305,
      "rewards/progression_diversity/mean": -0.000289862509816885,
      "rewards/progression_diversity/std": 0.004467545077204704,
      "rewards/symbolic_reward_accuracy/mean": 0.146484375,
      "rewards/symbolic_reward_accuracy/std": 0.35393697023391724,
      "rewards/symbolic_reward_partial_score/mean": 0.7647460699081421,
      "rewards/symbolic_reward_partial_score/std": 0.18048354983329773,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0483635663986206,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.4491806030273438,
      "step": 1517
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.2445373684167862,
      "epoch": 2.4326923076923075,
      "grad_norm": 0.01748676225543022,
      "learning_rate": 1e-06,
      "loss": 0.0166,
      "step": 1518
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.23935458809137344,
      "epoch": 2.434294871794872,
      "grad_norm": 0.04788840189576149,
      "learning_rate": 1e-06,
      "loss": 0.0312,
      "step": 1519
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.24610072374343872,
      "epoch": 2.435897435897436,
      "grad_norm": 0.014646312221884727,
      "learning_rate": 1e-06,
      "loss": 0.0119,
      "step": 1520
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3350.0,
      "completions/mean_length": 1556.814453125,
      "completions/mean_terminated_length": 1469.4244384765625,
      "completions/min_length": 377.0,
      "completions/min_terminated_length": 377.0,
      "entropy": 0.24429644644260406,
      "epoch": 2.4375,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 461.6620788574219,
      "learning_rate": 1e-06,
      "loss": 0.0507,
      "num_tokens": 1005568947.0,
      "reward": 0.3079223334789276,
      "reward_std": 0.038352809846401215,
      "rewards/progression_diversity/mean": -0.00024961589951999485,
      "rewards/progression_diversity/std": 0.004486089572310448,
      "rewards/symbolic_reward_accuracy/mean": 0.14453125,
      "rewards/symbolic_reward_accuracy/std": 0.35197147727012634,
      "rewards/symbolic_reward_partial_score/mean": 0.739306628704071,
      "rewards/symbolic_reward_partial_score/std": 0.18258456885814667,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0502259731292725,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.174125909805298,
      "step": 1521
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.24371539801359177,
      "epoch": 2.439102564102564,
      "grad_norm": 0.015633653849363327,
      "learning_rate": 1e-06,
      "loss": 0.0287,
      "step": 1522
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.2456585019826889,
      "epoch": 2.440705128205128,
      "grad_norm": 0.021190311759710312,
      "learning_rate": 1e-06,
      "loss": 0.0031,
      "step": 1523
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.24361753463745117,
      "epoch": 2.4423076923076925,
      "grad_norm": 0.011759229004383087,
      "learning_rate": 1e-06,
      "loss": -0.0044,
      "step": 1524
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5319.0,
      "completions/mean_length": 1642.1484375,
      "completions/mean_terminated_length": 1467.3438720703125,
      "completions/min_length": 350.0,
      "completions/min_terminated_length": 350.0,
      "entropy": 0.24563409388065338,
      "epoch": 2.4439102564102564,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 1274.161376953125,
      "learning_rate": 1e-06,
      "loss": 0.0363,
      "num_tokens": 1007251439.0,
      "reward": 0.3221524655818939,
      "reward_std": 0.05205688625574112,
      "rewards/progression_diversity/mean": -0.0015531220706179738,
      "rewards/progression_diversity/std": 0.016780469566583633,
      "rewards/symbolic_reward_accuracy/mean": 0.189453125,
      "rewards/symbolic_reward_accuracy/std": 0.3922513723373413,
      "rewards/symbolic_reward_partial_score/mean": 0.6982421875,
      "rewards/symbolic_reward_partial_score/std": 0.20122045278549194,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0466976165771484,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.579723834991455,
      "step": 1525
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.23505118489265442,
      "epoch": 2.4455128205128207,
      "grad_norm": 5386.1142578125,
      "learning_rate": 1e-06,
      "loss": 0.7638,
      "step": 1526
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23239125311374664,
      "epoch": 2.4471153846153846,
      "grad_norm": 0.012965874746441841,
      "learning_rate": 1e-06,
      "loss": 0.0457,
      "step": 1527
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.23721230030059814,
      "epoch": 2.448717948717949,
      "grad_norm": 0.01935112476348877,
      "learning_rate": 1e-06,
      "loss": 0.0124,
      "step": 1528
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5069.0,
      "completions/mean_length": 1434.068359375,
      "completions/mean_terminated_length": 1345.954833984375,
      "completions/min_length": 500.0,
      "completions/min_terminated_length": 500.0,
      "entropy": 0.2473963499069214,
      "epoch": 2.4503205128205128,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.016699276864528656,
      "learning_rate": 1e-06,
      "loss": -0.0043,
      "num_tokens": 1008849506.0,
      "reward": 0.3746347427368164,
      "reward_std": 0.0481521412730217,
      "rewards/progression_diversity/mean": -0.0008823598036542535,
      "rewards/progression_diversity/std": 0.01371445506811142,
      "rewards/symbolic_reward_accuracy/mean": 0.25390625,
      "rewards/symbolic_reward_accuracy/std": 0.43567025661468506,
      "rewards/symbolic_reward_partial_score/mean": 0.7429524660110474,
      "rewards/symbolic_reward_partial_score/std": 0.21133190393447876,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0495123863220215,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 2.0027153491973877,
      "step": 1529
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.24604248255491257,
      "epoch": 2.451923076923077,
      "grad_norm": 0.012123506516218185,
      "learning_rate": 1e-06,
      "loss": 0.013,
      "step": 1530
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.24536657333374023,
      "epoch": 2.453525641025641,
      "grad_norm": 0.010596076026558876,
      "learning_rate": 1e-06,
      "loss": -0.007,
      "step": 1531
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.23595212399959564,
      "epoch": 2.4551282051282053,
      "grad_norm": 0.009966249577701092,
      "learning_rate": 1e-06,
      "loss": 0.1189,
      "step": 1532
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5517.0,
      "completions/mean_length": 1541.44921875,
      "completions/mean_terminated_length": 1424.5787353515625,
      "completions/min_length": 372.0,
      "completions/min_terminated_length": 372.0,
      "entropy": 0.23892727494239807,
      "epoch": 2.456730769230769,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.019874002784490585,
      "learning_rate": 1e-06,
      "loss": -0.0075,
      "num_tokens": 1010481960.0,
      "reward": 0.4098053574562073,
      "reward_std": 0.033480141311883926,
      "rewards/progression_diversity/mean": -0.001889547100290656,
      "rewards/progression_diversity/std": 0.03145535662770271,
      "rewards/symbolic_reward_accuracy/mean": 0.30859375,
      "rewards/symbolic_reward_accuracy/std": 0.4623647928237915,
      "rewards/symbolic_reward_partial_score/mean": 0.7514973878860474,
      "rewards/symbolic_reward_partial_score/std": 0.20948295295238495,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0455158948898315,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 3.3153014183044434,
      "step": 1533
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.23280858993530273,
      "epoch": 2.4583333333333335,
      "grad_norm": 0.01161316316574812,
      "learning_rate": 1e-06,
      "loss": 0.0201,
      "step": 1534
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2302737832069397,
      "epoch": 2.4599358974358974,
      "grad_norm": 0.030310511589050293,
      "learning_rate": 1e-06,
      "loss": 0.0335,
      "step": 1535
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.23068059235811234,
      "epoch": 2.4615384615384617,
      "grad_norm": 0.010563918389379978,
      "learning_rate": 1e-06,
      "loss": 0.054,
      "step": 1536
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4376.0,
      "completions/mean_length": 1490.501953125,
      "completions/mean_terminated_length": 1373.2303466796875,
      "completions/min_length": 461.0,
      "completions/min_terminated_length": 461.0,
      "entropy": 0.2301376685500145,
      "epoch": 2.4631410256410255,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.016749106347560883,
      "learning_rate": 1e-06,
      "loss": -0.0055,
      "num_tokens": 1012094073.0,
      "reward": 0.36312228441238403,
      "reward_std": 0.039926785975694656,
      "rewards/progression_diversity/mean": -0.0017360053025186062,
      "rewards/progression_diversity/std": 0.01971437782049179,
      "rewards/symbolic_reward_accuracy/mean": 0.236328125,
      "rewards/symbolic_reward_accuracy/std": 0.42524150013923645,
      "rewards/symbolic_reward_partial_score/mean": 0.7404134273529053,
      "rewards/symbolic_reward_partial_score/std": 0.20340432226657867,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045566439628601,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 2.109564781188965,
      "step": 1537
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22669820487499237,
      "epoch": 2.46474358974359,
      "grad_norm": 0.013202717527747154,
      "learning_rate": 1e-06,
      "loss": 0.0247,
      "step": 1538
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.23074190318584442,
      "epoch": 2.4663461538461537,
      "grad_norm": 0.028765065595507622,
      "learning_rate": 1e-06,
      "loss": 0.0033,
      "step": 1539
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.224589504301548,
      "epoch": 2.467948717948718,
      "grad_norm": 0.021383749321103096,
      "learning_rate": 1e-06,
      "loss": 0.0617,
      "step": 1540
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4875.0,
      "completions/mean_length": 1534.798828125,
      "completions/mean_terminated_length": 1388.3570556640625,
      "completions/min_length": 402.0,
      "completions/min_terminated_length": 402.0,
      "entropy": 0.2195545881986618,
      "epoch": 2.469551282051282,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.039864543825387955,
      "learning_rate": 1e-06,
      "loss": 0.0975,
      "num_tokens": 1013848578.0,
      "reward": 0.2403162568807602,
      "reward_std": 0.0261523500084877,
      "rewards/progression_diversity/mean": -0.0010903782676905394,
      "rewards/progression_diversity/std": 0.020245160907506943,
      "rewards/symbolic_reward_accuracy/mean": 0.060546875,
      "rewards/symbolic_reward_accuracy/std": 0.2387305200099945,
      "rewards/symbolic_reward_partial_score/mean": 0.6832519769668579,
      "rewards/symbolic_reward_partial_score/std": 0.18514306843280792,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0433101654052734,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 3.134794235229492,
      "step": 1541
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.22596587240695953,
      "epoch": 2.4711538461538463,
      "grad_norm": 0.01767021417617798,
      "learning_rate": 1e-06,
      "loss": -0.0098,
      "step": 1542
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.22497133910655975,
      "epoch": 2.47275641025641,
      "grad_norm": 0.013781007379293442,
      "learning_rate": 1e-06,
      "loss": 0.0239,
      "step": 1543
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.2270878478884697,
      "epoch": 2.4743589743589745,
      "grad_norm": 0.011908365413546562,
      "learning_rate": 1e-06,
      "loss": 0.019,
      "step": 1544
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3915.0,
      "completions/mean_length": 1276.716796875,
      "completions/mean_terminated_length": 1247.152587890625,
      "completions/min_length": 390.0,
      "completions/min_terminated_length": 390.0,
      "entropy": 0.23582519590854645,
      "epoch": 2.4759615384615383,
      "frac_reward_zero_std": 0.65625,
      "grad_norm": 581.7129516601562,
      "learning_rate": 1e-06,
      "loss": 0.0289,
      "num_tokens": 1015333441.0,
      "reward": 0.40914207696914673,
      "reward_std": 0.014557499438524246,
      "rewards/progression_diversity/mean": -0.0003453929675742984,
      "rewards/progression_diversity/std": 0.00676377210766077,
      "rewards/symbolic_reward_accuracy/mean": 0.310546875,
      "rewards/symbolic_reward_accuracy/std": 0.46317005157470703,
      "rewards/symbolic_reward_partial_score/mean": 0.7433756589889526,
      "rewards/symbolic_reward_partial_score/std": 0.229752317070961,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0511703491210938,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 0.7563881278038025,
      "step": 1545
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.23144873976707458,
      "epoch": 2.4775641025641026,
      "grad_norm": 0.010454997420310974,
      "learning_rate": 1e-06,
      "loss": 0.0032,
      "step": 1546
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2326817438006401,
      "epoch": 2.4791666666666665,
      "grad_norm": 0.01226350199431181,
      "learning_rate": 1e-06,
      "loss": -0.0076,
      "step": 1547
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.23457026481628418,
      "epoch": 2.480769230769231,
      "grad_norm": 0.009266077540814877,
      "learning_rate": 1e-06,
      "loss": -0.0032,
      "step": 1548
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4929.0,
      "completions/mean_length": 1437.140625,
      "completions/mean_terminated_length": 1289.7357177734375,
      "completions/min_length": 422.0,
      "completions/min_terminated_length": 422.0,
      "entropy": 0.21618512272834778,
      "epoch": 2.4823717948717947,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 437.86102294921875,
      "learning_rate": 1e-06,
      "loss": 0.0556,
      "num_tokens": 1016930265.0,
      "reward": 0.3245599865913391,
      "reward_std": 0.04626595973968506,
      "rewards/progression_diversity/mean": -0.0015235436148941517,
      "rewards/progression_diversity/std": 0.01856166310608387,
      "rewards/symbolic_reward_accuracy/mean": 0.193359375,
      "rewards/symbolic_reward_accuracy/std": 0.39531853795051575,
      "rewards/symbolic_reward_partial_score/mean": 0.6984537839889526,
      "rewards/symbolic_reward_partial_score/std": 0.2078658640384674,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.044947862625122,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 756.0,
      "sampling/sampling_logp_difference/mean": 2.158604145050049,
      "step": 1549
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.23416189849376678,
      "epoch": 2.483974358974359,
      "grad_norm": 0.011794805526733398,
      "learning_rate": 1e-06,
      "loss": -0.0026,
      "step": 1550
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2325107902288437,
      "epoch": 2.485576923076923,
      "grad_norm": 0.018296558409929276,
      "learning_rate": 1e-06,
      "loss": 0.0022,
      "step": 1551
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.21691740304231644,
      "epoch": 2.4871794871794872,
      "grad_norm": 0.03286243975162506,
      "learning_rate": 1e-06,
      "loss": 0.0341,
      "step": 1552
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4565.0,
      "completions/mean_length": 1326.94921875,
      "completions/mean_terminated_length": 1238.204345703125,
      "completions/min_length": 377.0,
      "completions/min_terminated_length": 377.0,
      "entropy": 0.21416771411895752,
      "epoch": 2.488782051282051,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 242.9971160888672,
      "learning_rate": 1e-06,
      "loss": 0.0253,
      "num_tokens": 1018587343.0,
      "reward": 0.24987833201885223,
      "reward_std": 0.025197582319378853,
      "rewards/progression_diversity/mean": -0.0004485528916120529,
      "rewards/progression_diversity/std": 0.0056631616316735744,
      "rewards/symbolic_reward_accuracy/mean": 0.08984375,
      "rewards/symbolic_reward_accuracy/std": 0.2862374484539032,
      "rewards/symbolic_reward_partial_score/mean": 0.6552083492279053,
      "rewards/symbolic_reward_partial_score/std": 0.1901572048664093,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0427167415618896,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.669922351837158,
      "step": 1553
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.21156911551952362,
      "epoch": 2.4903846153846154,
      "grad_norm": 0.014419353567063808,
      "learning_rate": 1e-06,
      "loss": 0.0138,
      "step": 1554
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.21717411279678345,
      "epoch": 2.4919871794871793,
      "grad_norm": 0.011920792050659657,
      "learning_rate": 1e-06,
      "loss": 0.0193,
      "step": 1555
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.21699684858322144,
      "epoch": 2.4935897435897436,
      "grad_norm": 0.011259951628744602,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "step": 1556
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3890.0,
      "completions/mean_length": 1333.09765625,
      "completions/mean_terminated_length": 1244.3890380859375,
      "completions/min_length": 372.0,
      "completions/min_terminated_length": 372.0,
      "entropy": 0.2051466777920723,
      "epoch": 2.4951923076923075,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 179.58285522460938,
      "learning_rate": 1e-06,
      "loss": 0.0567,
      "num_tokens": 1020162753.0,
      "reward": 0.4125392436981201,
      "reward_std": 0.02173442952334881,
      "rewards/progression_diversity/mean": -0.0004710496577899903,
      "rewards/progression_diversity/std": 0.004797840025275946,
      "rewards/symbolic_reward_accuracy/mean": 0.3046875,
      "rewards/symbolic_reward_accuracy/std": 0.4607250988483429,
      "rewards/symbolic_reward_partial_score/mean": 0.7670735716819763,
      "rewards/symbolic_reward_partial_score/std": 0.21079780161380768,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0423927307128906,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 1.7513947486877441,
      "step": 1557
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.21501071751117706,
      "epoch": 2.496794871794872,
      "grad_norm": 0.00948250014334917,
      "learning_rate": 1e-06,
      "loss": -0.0021,
      "step": 1558
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2102562114596367,
      "epoch": 2.498397435897436,
      "grad_norm": 0.007870234549045563,
      "learning_rate": 1e-06,
      "loss": 0.0207,
      "step": 1559
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.21096129715442657,
      "epoch": 2.5,
      "grad_norm": 0.007708584889769554,
      "learning_rate": 1e-06,
      "loss": -0.0001,
      "step": 1560
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4414.0,
      "completions/mean_length": 1309.21484375,
      "completions/mean_terminated_length": 1220.365478515625,
      "completions/min_length": 393.0,
      "completions/min_terminated_length": 393.0,
      "entropy": 0.20147471129894257,
      "epoch": 2.501602564102564,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.02653343416750431,
      "learning_rate": 1e-06,
      "loss": 0.0258,
      "num_tokens": 1021758815.0,
      "reward": 0.30301350355148315,
      "reward_std": 0.051949888467788696,
      "rewards/progression_diversity/mean": -0.003340594470500946,
      "rewards/progression_diversity/std": 0.030672553926706314,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.7074218988418579,
      "rewards/symbolic_reward_partial_score/std": 0.1824207305908203,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0406644344329834,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.7471396923065186,
      "step": 1561
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.20691817998886108,
      "epoch": 2.503205128205128,
      "grad_norm": 0.018390124663710594,
      "learning_rate": 1e-06,
      "loss": 0.0263,
      "step": 1562
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.20330524444580078,
      "epoch": 2.5048076923076925,
      "grad_norm": 0.01390912476927042,
      "learning_rate": 1e-06,
      "loss": 0.0307,
      "step": 1563
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.20695596933364868,
      "epoch": 2.5064102564102564,
      "grad_norm": 0.01517625991255045,
      "learning_rate": 1e-06,
      "loss": -0.0015,
      "step": 1564
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4008.0,
      "completions/mean_length": 1362.990234375,
      "completions/mean_terminated_length": 1214.85400390625,
      "completions/min_length": 461.0,
      "completions/min_terminated_length": 461.0,
      "entropy": 0.19462278485298157,
      "epoch": 2.5080128205128203,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 506.2625732421875,
      "learning_rate": 1e-06,
      "loss": 0.0486,
      "num_tokens": 1023359562.0,
      "reward": 0.34047943353652954,
      "reward_std": 0.035891205072402954,
      "rewards/progression_diversity/mean": -0.0008854115731082857,
      "rewards/progression_diversity/std": 0.008040647022426128,
      "rewards/symbolic_reward_accuracy/mean": 0.205078125,
      "rewards/symbolic_reward_accuracy/std": 0.4041535556316376,
      "rewards/symbolic_reward_partial_score/mean": 0.7280598878860474,
      "rewards/symbolic_reward_partial_score/std": 0.1988910287618637,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.037065029144287,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 3.887390375137329,
      "step": 1565
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.20349303632974625,
      "epoch": 2.5096153846153846,
      "grad_norm": 0.02085845172405243,
      "learning_rate": 1e-06,
      "loss": 0.0164,
      "step": 1566
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.1964193433523178,
      "epoch": 2.511217948717949,
      "grad_norm": 0.010187533684074879,
      "learning_rate": 1e-06,
      "loss": 0.0258,
      "step": 1567
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.20827952772378922,
      "epoch": 2.5128205128205128,
      "grad_norm": 0.00918852724134922,
      "learning_rate": 1e-06,
      "loss": 0.0049,
      "step": 1568
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4599.0,
      "completions/mean_length": 1323.048828125,
      "completions/mean_terminated_length": 1234.281005859375,
      "completions/min_length": 421.0,
      "completions/min_terminated_length": 421.0,
      "entropy": 0.21191302686929703,
      "epoch": 2.5144230769230766,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.020111212506890297,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "num_tokens": 1024883955.0,
      "reward": 0.3286048173904419,
      "reward_std": 0.0436086431145668,
      "rewards/progression_diversity/mean": -0.0013349888613447547,
      "rewards/progression_diversity/std": 0.01073912438005209,
      "rewards/symbolic_reward_accuracy/mean": 0.189453125,
      "rewards/symbolic_reward_accuracy/std": 0.3922513723373413,
      "rewards/symbolic_reward_partial_score/mean": 0.7184407711029053,
      "rewards/symbolic_reward_partial_score/std": 0.20771624147891998,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0429773330688477,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 756.0,
      "sampling/sampling_logp_difference/mean": 1.301384687423706,
      "step": 1569
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2052868902683258,
      "epoch": 2.516025641025641,
      "grad_norm": 0.018205396831035614,
      "learning_rate": 1e-06,
      "loss": 0.0922,
      "step": 1570
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.20686771720647812,
      "epoch": 2.5176282051282053,
      "grad_norm": 0.010167265310883522,
      "learning_rate": 1e-06,
      "loss": 0.0094,
      "step": 1571
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.20857567340135574,
      "epoch": 2.519230769230769,
      "grad_norm": 0.012847342528402805,
      "learning_rate": 1e-06,
      "loss": 0.026,
      "step": 1572
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2660.0,
      "completions/mean_length": 1316.029296875,
      "completions/mean_terminated_length": 1197.3839111328125,
      "completions/min_length": 310.0,
      "completions/min_terminated_length": 310.0,
      "entropy": 0.2158561423420906,
      "epoch": 2.5208333333333335,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.013560623861849308,
      "learning_rate": 1e-06,
      "loss": -0.0024,
      "num_tokens": 1026358946.0,
      "reward": 0.29157090187072754,
      "reward_std": 0.04181563854217529,
      "rewards/progression_diversity/mean": -0.0011149711208418012,
      "rewards/progression_diversity/std": 0.010894270613789558,
      "rewards/symbolic_reward_accuracy/mean": 0.130859375,
      "rewards/symbolic_reward_accuracy/std": 0.33757632970809937,
      "rewards/symbolic_reward_partial_score/mean": 0.7115234136581421,
      "rewards/symbolic_reward_partial_score/std": 0.1730296015739441,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0387065410614014,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 3.532683849334717,
      "step": 1573
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.20614393055438995,
      "epoch": 2.5224358974358974,
      "grad_norm": 0.023240137845277786,
      "learning_rate": 1e-06,
      "loss": 0.0632,
      "step": 1574
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.21372488886117935,
      "epoch": 2.5240384615384617,
      "grad_norm": 0.016561856493353844,
      "learning_rate": 1e-06,
      "loss": -0.0029,
      "step": 1575
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.209242083132267,
      "epoch": 2.5256410256410255,
      "grad_norm": 0.011606593616306782,
      "learning_rate": 1e-06,
      "loss": 0.0508,
      "step": 1576
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4388.0,
      "completions/mean_length": 1337.361328125,
      "completions/mean_terminated_length": 1218.8839111328125,
      "completions/min_length": 453.0,
      "completions/min_terminated_length": 453.0,
      "entropy": 0.21050991117954254,
      "epoch": 2.52724358974359,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 682.1195068359375,
      "learning_rate": 1e-06,
      "loss": 0.0348,
      "num_tokens": 1027800187.0,
      "reward": 0.3529677093029022,
      "reward_std": 0.03494004160165787,
      "rewards/progression_diversity/mean": -0.002058391459286213,
      "rewards/progression_diversity/std": 0.028598375618457794,
      "rewards/symbolic_reward_accuracy/mean": 0.21875,
      "rewards/symbolic_reward_accuracy/std": 0.41380295157432556,
      "rewards/symbolic_reward_partial_score/mean": 0.7410807609558105,
      "rewards/symbolic_reward_partial_score/std": 0.19470229744911194,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0408451557159424,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.35636568069458,
      "step": 1577
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.20368197560310364,
      "epoch": 2.5288461538461537,
      "grad_norm": 0.014921136200428009,
      "learning_rate": 1e-06,
      "loss": 0.0237,
      "step": 1578
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.21754014492034912,
      "epoch": 2.530448717948718,
      "grad_norm": 0.018416838720440865,
      "learning_rate": 1e-06,
      "loss": -0.0029,
      "step": 1579
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.21085964888334274,
      "epoch": 2.532051282051282,
      "grad_norm": 0.022706199437379837,
      "learning_rate": 1e-06,
      "loss": 0.0528,
      "step": 1580
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4464.0,
      "completions/mean_length": 1427.904296875,
      "completions/mean_terminated_length": 1250.559326171875,
      "completions/min_length": 433.0,
      "completions/min_terminated_length": 433.0,
      "entropy": 0.21092313528060913,
      "epoch": 2.5336538461538463,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 0.020028864964842796,
      "learning_rate": 1e-06,
      "loss": -0.0007,
      "num_tokens": 1029409994.0,
      "reward": 0.32646113634109497,
      "reward_std": 0.04266618192195892,
      "rewards/progression_diversity/mean": -0.0013511213473975658,
      "rewards/progression_diversity/std": 0.012499148957431316,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.7243163585662842,
      "rewards/symbolic_reward_partial_score/std": 0.20394515991210938,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0392813682556152,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.5135602951049805,
      "step": 1581
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.20285531133413315,
      "epoch": 2.53525641025641,
      "grad_norm": 0.013795009814202785,
      "learning_rate": 1e-06,
      "loss": 0.035,
      "step": 1582
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.2033623531460762,
      "epoch": 2.5368589743589745,
      "grad_norm": 0.010789959691464901,
      "learning_rate": 1e-06,
      "loss": 0.0538,
      "step": 1583
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.2029578685760498,
      "epoch": 2.5384615384615383,
      "grad_norm": 0.009514546021819115,
      "learning_rate": 1e-06,
      "loss": 0.0284,
      "step": 1584
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3833.0,
      "completions/mean_length": 1337.1015625,
      "completions/mean_terminated_length": 1218.6220703125,
      "completions/min_length": 364.0,
      "completions/min_terminated_length": 364.0,
      "entropy": 0.19549165666103363,
      "epoch": 2.5400641025641026,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 663.7005004882812,
      "learning_rate": 1e-06,
      "loss": 0.0671,
      "num_tokens": 1031023614.0,
      "reward": 0.3425098657608032,
      "reward_std": 0.050741590559482574,
      "rewards/progression_diversity/mean": -0.0014545343583449721,
      "rewards/progression_diversity/std": 0.01686347834765911,
      "rewards/symbolic_reward_accuracy/mean": 0.220703125,
      "rewards/symbolic_reward_accuracy/std": 0.4151262938976288,
      "rewards/symbolic_reward_partial_score/mean": 0.7029459476470947,
      "rewards/symbolic_reward_partial_score/std": 0.21629534661769867,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0390223264694214,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.6212563514709473,
      "step": 1585
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2048923447728157,
      "epoch": 2.5416666666666665,
      "grad_norm": 0.015606733970344067,
      "learning_rate": 1e-06,
      "loss": -0.0031,
      "step": 1586
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.19919227808713913,
      "epoch": 2.543269230769231,
      "grad_norm": 0.02132081612944603,
      "learning_rate": 1e-06,
      "loss": 0.0211,
      "step": 1587
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2042047083377838,
      "epoch": 2.5448717948717947,
      "grad_norm": 0.01672261208295822,
      "learning_rate": 1e-06,
      "loss": -0.0112,
      "step": 1588
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3849.0,
      "completions/mean_length": 1405.666015625,
      "completions/mean_terminated_length": 1287.726318359375,
      "completions/min_length": 417.0,
      "completions/min_terminated_length": 417.0,
      "entropy": 0.20447014272212982,
      "epoch": 2.546474358974359,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 262.0090637207031,
      "learning_rate": 1e-06,
      "loss": 0.0289,
      "num_tokens": 1032674947.0,
      "reward": 0.3156275749206543,
      "reward_std": 0.04013008996844292,
      "rewards/progression_diversity/mean": -0.0007197089726105332,
      "rewards/progression_diversity/std": 0.008052974939346313,
      "rewards/symbolic_reward_accuracy/mean": 0.181640625,
      "rewards/symbolic_reward_accuracy/std": 0.38592514395713806,
      "rewards/symbolic_reward_partial_score/mean": 0.6914387941360474,
      "rewards/symbolic_reward_partial_score/std": 0.2133348137140274,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0389673709869385,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.4872024059295654,
      "step": 1589
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.1969490796327591,
      "epoch": 2.5480769230769234,
      "grad_norm": 0.020587697625160217,
      "learning_rate": 1e-06,
      "loss": 0.0394,
      "step": 1590
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2012443020939827,
      "epoch": 2.5496794871794872,
      "grad_norm": 0.012877855449914932,
      "learning_rate": 1e-06,
      "loss": -0.004,
      "step": 1591
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2060464546084404,
      "epoch": 2.551282051282051,
      "grad_norm": 0.017594829201698303,
      "learning_rate": 1e-06,
      "loss": 0.0178,
      "step": 1592
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4988.0,
      "completions/mean_length": 1498.439453125,
      "completions/mean_terminated_length": 1351.6390380859375,
      "completions/min_length": 495.0,
      "completions/min_terminated_length": 495.0,
      "entropy": 0.20640329271554947,
      "epoch": 2.5528846153846154,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 406.1407470703125,
      "learning_rate": 1e-06,
      "loss": 0.0347,
      "num_tokens": 1034189668.0,
      "reward": 0.30137068033218384,
      "reward_std": 0.058662112802267075,
      "rewards/progression_diversity/mean": -0.00013947187107987702,
      "rewards/progression_diversity/std": 0.0018558743176981807,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.6946777105331421,
      "rewards/symbolic_reward_partial_score/std": 0.20915855467319489,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0420677661895752,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 756.0,
      "sampling/sampling_logp_difference/mean": 1.9947669506072998,
      "step": 1593
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.21118755638599396,
      "epoch": 2.5544871794871797,
      "grad_norm": 1052.406494140625,
      "learning_rate": 1e-06,
      "loss": 0.043,
      "step": 1594
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2124936804175377,
      "epoch": 2.5560897435897436,
      "grad_norm": 0.026562146842479706,
      "learning_rate": 1e-06,
      "loss": 0.0204,
      "step": 1595
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.21463356912136078,
      "epoch": 2.5576923076923075,
      "grad_norm": 0.01585652120411396,
      "learning_rate": 1e-06,
      "loss": 0.0246,
      "step": 1596
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4207.0,
      "completions/mean_length": 1389.62890625,
      "completions/mean_terminated_length": 1330.8275146484375,
      "completions/min_length": 422.0,
      "completions/min_terminated_length": 422.0,
      "entropy": 0.2117195725440979,
      "epoch": 2.559294871794872,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.02877124771475792,
      "learning_rate": 1e-06,
      "loss": -0.0013,
      "num_tokens": 1035825670.0,
      "reward": 0.3299679160118103,
      "reward_std": 0.03433844819664955,
      "rewards/progression_diversity/mean": -0.00028184783877804875,
      "rewards/progression_diversity/std": 0.003830707399174571,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.7262043952941895,
      "rewards/symbolic_reward_partial_score/std": 0.18292047083377838,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0448970794677734,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 0.6958411931991577,
      "step": 1597
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.20743755996227264,
      "epoch": 2.560897435897436,
      "grad_norm": 0.028087658807635307,
      "learning_rate": 1e-06,
      "loss": 0.0076,
      "step": 1598
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.20653068274259567,
      "epoch": 2.5625,
      "grad_norm": 0.014355894178152084,
      "learning_rate": 1e-06,
      "loss": 0.0279,
      "step": 1599
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.2106596827507019,
      "epoch": 2.564102564102564,
      "grad_norm": 0.01578899845480919,
      "learning_rate": 1e-06,
      "loss": 0.0246,
      "step": 1600
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3138.0,
      "completions/mean_length": 1489.35546875,
      "completions/mean_terminated_length": 1282.8951416015625,
      "completions/min_length": 453.0,
      "completions/min_terminated_length": 453.0,
      "entropy": 0.21462351083755493,
      "epoch": 2.565705128205128,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 820.033935546875,
      "learning_rate": 1e-06,
      "loss": 0.0379,
      "num_tokens": 1037399484.0,
      "reward": 0.3129954934120178,
      "reward_std": 0.025571607053279877,
      "rewards/progression_diversity/mean": -0.0007452977006323636,
      "rewards/progression_diversity/std": 0.01137583889067173,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.7347493171691895,
      "rewards/symbolic_reward_partial_score/std": 0.177503302693367,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0383716821670532,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 3.7996788024902344,
      "step": 1601
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21920231729745865,
      "epoch": 2.5673076923076925,
      "grad_norm": 0.01972138322889805,
      "learning_rate": 1e-06,
      "loss": 0.036,
      "step": 1602
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.21449632197618484,
      "epoch": 2.5689102564102564,
      "grad_norm": 0.015361227095127106,
      "learning_rate": 1e-06,
      "loss": 0.0557,
      "step": 1603
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.21294260025024414,
      "epoch": 2.5705128205128203,
      "grad_norm": 0.009509088471531868,
      "learning_rate": 1e-06,
      "loss": 0.0432,
      "step": 1604
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5541.0,
      "completions/mean_length": 1318.08984375,
      "completions/mean_terminated_length": 1199.4605712890625,
      "completions/min_length": 375.0,
      "completions/min_terminated_length": 375.0,
      "entropy": 0.21971651911735535,
      "epoch": 2.5721153846153846,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.024938074871897697,
      "learning_rate": 1e-06,
      "loss": 0.0235,
      "num_tokens": 1039039034.0,
      "reward": 0.29758530855178833,
      "reward_std": 0.030321484431624413,
      "rewards/progression_diversity/mean": -0.0017264732159674168,
      "rewards/progression_diversity/std": 0.027802200987935066,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.6899251341819763,
      "rewards/symbolic_reward_partial_score/std": 0.21471048891544342,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0427124500274658,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 752.0,
      "sampling/sampling_logp_difference/mean": 3.036716938018799,
      "step": 1605
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2219068706035614,
      "epoch": 2.573717948717949,
      "grad_norm": 0.012791654095053673,
      "learning_rate": 1e-06,
      "loss": 0.0082,
      "step": 1606
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.20912151038646698,
      "epoch": 2.5753205128205128,
      "grad_norm": 0.008532814681529999,
      "learning_rate": 1e-06,
      "loss": 0.033,
      "step": 1607
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.22284353524446487,
      "epoch": 2.5769230769230766,
      "grad_norm": 0.019355937838554382,
      "learning_rate": 1e-06,
      "loss": 0.0272,
      "step": 1608
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4659.0,
      "completions/mean_length": 1479.640625,
      "completions/mean_terminated_length": 1302.9091796875,
      "completions/min_length": 426.0,
      "completions/min_terminated_length": 426.0,
      "entropy": 0.22081206738948822,
      "epoch": 2.578525641025641,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.018469780683517456,
      "learning_rate": 1e-06,
      "loss": 0.0248,
      "num_tokens": 1040598594.0,
      "reward": 0.41613101959228516,
      "reward_std": 0.07369263470172882,
      "rewards/progression_diversity/mean": -0.0035969598684459925,
      "rewards/progression_diversity/std": 0.033173561096191406,
      "rewards/symbolic_reward_accuracy/mean": 0.306640625,
      "rewards/symbolic_reward_accuracy/std": 0.4615498185157776,
      "rewards/symbolic_reward_partial_score/mean": 0.7778483033180237,
      "rewards/symbolic_reward_partial_score/std": 0.20071126520633698,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0395357608795166,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 756.0,
      "sampling/sampling_logp_difference/mean": 4.858417510986328,
      "step": 1609
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.2339448481798172,
      "epoch": 2.5801282051282053,
      "grad_norm": 0.021145859733223915,
      "learning_rate": 1e-06,
      "loss": -0.0069,
      "step": 1610
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.22270727157592773,
      "epoch": 2.581730769230769,
      "grad_norm": 0.014868995174765587,
      "learning_rate": 1e-06,
      "loss": 0.0422,
      "step": 1611
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.21710411459207535,
      "epoch": 2.5833333333333335,
      "grad_norm": 1147.5833740234375,
      "learning_rate": 1e-06,
      "loss": 0.2004,
      "step": 1612
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5854.0,
      "completions/mean_length": 1488.01953125,
      "completions/mean_terminated_length": 1311.387451171875,
      "completions/min_length": 384.0,
      "completions/min_terminated_length": 384.0,
      "entropy": 0.2146845906972885,
      "epoch": 2.5849358974358974,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 0.021604053676128387,
      "learning_rate": 1e-06,
      "loss": 0.0206,
      "num_tokens": 1042269212.0,
      "reward": 0.3228908181190491,
      "reward_std": 0.06467078626155853,
      "rewards/progression_diversity/mean": -0.0019355263793841004,
      "rewards/progression_diversity/std": 0.025548765435814857,
      "rewards/symbolic_reward_accuracy/mean": 0.16015625,
      "rewards/symbolic_reward_accuracy/std": 0.3671095669269562,
      "rewards/symbolic_reward_partial_score/mean": 0.7599608898162842,
      "rewards/symbolic_reward_partial_score/std": 0.1819555014371872,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0413014888763428,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.9018664360046387,
      "step": 1613
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.20722515881061554,
      "epoch": 2.5865384615384617,
      "grad_norm": 440.9079895019531,
      "learning_rate": 1e-06,
      "loss": 0.1145,
      "step": 1614
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.20744751393795013,
      "epoch": 2.5881410256410255,
      "grad_norm": 2084.2294921875,
      "learning_rate": 1e-06,
      "loss": 0.2864,
      "step": 1615
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.22113841027021408,
      "epoch": 2.58974358974359,
      "grad_norm": 0.01803417317569256,
      "learning_rate": 1e-06,
      "loss": -0.0001,
      "step": 1616
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3889.0,
      "completions/mean_length": 1386.021484375,
      "completions/mean_terminated_length": 1297.624755859375,
      "completions/min_length": 322.0,
      "completions/min_terminated_length": 322.0,
      "entropy": 0.21602119505405426,
      "epoch": 2.5913461538461537,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.0642000138759613,
      "learning_rate": 1e-06,
      "loss": 0.0316,
      "num_tokens": 1043874039.0,
      "reward": 0.2649605870246887,
      "reward_std": 0.02608659118413925,
      "rewards/progression_diversity/mean": -0.0019893196877092123,
      "rewards/progression_diversity/std": 0.021042335778474808,
      "rewards/symbolic_reward_accuracy/mean": 0.095703125,
      "rewards/symbolic_reward_accuracy/std": 0.2944713830947876,
      "rewards/symbolic_reward_partial_score/mean": 0.693164050579071,
      "rewards/symbolic_reward_partial_score/std": 0.17582449316978455,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0435926914215088,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 3.047320604324341,
      "step": 1617
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2237110733985901,
      "epoch": 2.592948717948718,
      "grad_norm": 0.022002186626195908,
      "learning_rate": 1e-06,
      "loss": -0.0024,
      "step": 1618
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.21986420452594757,
      "epoch": 2.594551282051282,
      "grad_norm": 0.01582951843738556,
      "learning_rate": 1e-06,
      "loss": 0.0343,
      "step": 1619
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.22045395523309708,
      "epoch": 2.5961538461538463,
      "grad_norm": 0.021329455077648163,
      "learning_rate": 1e-06,
      "loss": 0.0004,
      "step": 1620
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2783.0,
      "completions/mean_length": 1341.115234375,
      "completions/mean_terminated_length": 1311.6771240234375,
      "completions/min_length": 443.0,
      "completions/min_terminated_length": 443.0,
      "entropy": 0.22125251591205597,
      "epoch": 2.59775641025641,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.018027139827609062,
      "learning_rate": 1e-06,
      "loss": 0.0207,
      "num_tokens": 1045449474.0,
      "reward": 0.36266857385635376,
      "reward_std": 0.047155484557151794,
      "rewards/progression_diversity/mean": -0.0007219260442070663,
      "rewards/progression_diversity/std": 0.00793998222798109,
      "rewards/symbolic_reward_accuracy/mean": 0.25390625,
      "rewards/symbolic_reward_accuracy/std": 0.43567025661468506,
      "rewards/symbolic_reward_partial_score/mean": 0.7017577886581421,
      "rewards/symbolic_reward_partial_score/std": 0.2232026308774948,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0492632389068604,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 0.9296596050262451,
      "step": 1621
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.22497374564409256,
      "epoch": 2.5993589743589745,
      "grad_norm": 0.01935565285384655,
      "learning_rate": 1e-06,
      "loss": 0.0026,
      "step": 1622
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.22052931785583496,
      "epoch": 2.6009615384615383,
      "grad_norm": 0.007344152312725782,
      "learning_rate": 1e-06,
      "loss": 0.0077,
      "step": 1623
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.21955804526805878,
      "epoch": 2.6025641025641026,
      "grad_norm": 0.013848908245563507,
      "learning_rate": 1e-06,
      "loss": 0.0017,
      "step": 1624
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3666.0,
      "completions/mean_length": 1220.310546875,
      "completions/mean_terminated_length": 1190.635986328125,
      "completions/min_length": 414.0,
      "completions/min_terminated_length": 414.0,
      "entropy": 0.23253807425498962,
      "epoch": 2.6041666666666665,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.01905795931816101,
      "learning_rate": 1e-06,
      "loss": 0.0033,
      "num_tokens": 1046961025.0,
      "reward": 0.31768307089805603,
      "reward_std": 0.016084838658571243,
      "rewards/progression_diversity/mean": -0.0002500782429706305,
      "rewards/progression_diversity/std": 0.003307226812466979,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.6885091066360474,
      "rewards/symbolic_reward_partial_score/std": 0.20202499628067017,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0517390966415405,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 756.0,
      "sampling/sampling_logp_difference/mean": 0.775344729423523,
      "step": 1625
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2306702509522438,
      "epoch": 2.605769230769231,
      "grad_norm": 0.009856280870735645,
      "learning_rate": 1e-06,
      "loss": 0.0303,
      "step": 1626
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.22651749849319458,
      "epoch": 2.6073717948717947,
      "grad_norm": 0.03335234895348549,
      "learning_rate": 1e-06,
      "loss": -0.003,
      "step": 1627
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.23627790808677673,
      "epoch": 2.608974358974359,
      "grad_norm": 0.01697065308690071,
      "learning_rate": 1e-06,
      "loss": -0.0023,
      "step": 1628
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 10288.0,
      "completions/mean_length": 1333.587890625,
      "completions/mean_terminated_length": 1274.5667724609375,
      "completions/min_length": 498.0,
      "completions/min_terminated_length": 498.0,
      "entropy": 0.2252894639968872,
      "epoch": 2.6105769230769234,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.10012649744749069,
      "learning_rate": 1e-06,
      "loss": 0.0249,
      "num_tokens": 1048435198.0,
      "reward": 0.3364197015762329,
      "reward_std": 0.03154347091913223,
      "rewards/progression_diversity/mean": -0.0025622413959354162,
      "rewards/progression_diversity/std": 0.030175259336829185,
      "rewards/symbolic_reward_accuracy/mean": 0.208984375,
      "rewards/symbolic_reward_accuracy/std": 0.40698084235191345,
      "rewards/symbolic_reward_partial_score/mean": 0.7035156488418579,
      "rewards/symbolic_reward_partial_score/std": 0.19566431641578674,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0484051704406738,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.8174188137054443,
      "step": 1629
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.23386212438344955,
      "epoch": 2.6121794871794872,
      "grad_norm": 0.015600372105836868,
      "learning_rate": 1e-06,
      "loss": 0.0257,
      "step": 1630
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.23984410613775253,
      "epoch": 2.613782051282051,
      "grad_norm": 0.0189208984375,
      "learning_rate": 1e-06,
      "loss": -0.0074,
      "step": 1631
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.22813158482313156,
      "epoch": 2.6153846153846154,
      "grad_norm": 0.01574343629181385,
      "learning_rate": 1e-06,
      "loss": 0.0725,
      "step": 1632
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3291.0,
      "completions/max_terminated_length": 3291.0,
      "completions/mean_length": 1187.220703125,
      "completions/mean_terminated_length": 1187.220703125,
      "completions/min_length": 344.0,
      "completions/min_terminated_length": 344.0,
      "entropy": 0.23443499207496643,
      "epoch": 2.6169871794871797,
      "frac_reward_zero_std": 0.71875,
      "grad_norm": 0.015873905271291733,
      "learning_rate": 1e-06,
      "loss": 0.0039,
      "num_tokens": 1049949615.0,
      "reward": 0.33012884855270386,
      "reward_std": 0.014471746981143951,
      "rewards/progression_diversity/mean": -0.0003017832641489804,
      "rewards/progression_diversity/std": 0.0035524494014680386,
      "rewards/symbolic_reward_accuracy/mean": 0.181640625,
      "rewards/symbolic_reward_accuracy/std": 0.38592514395713806,
      "rewards/symbolic_reward_partial_score/mean": 0.7371581792831421,
      "rewards/symbolic_reward_partial_score/std": 0.1770060658454895,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0539133548736572,
      "sampling/importance_sampling_ratio/min": 0.0032137008383870125,
      "sampling/sampling_logp_difference/max": 5.740332126617432,
      "sampling/sampling_logp_difference/mean": 0.1061110571026802,
      "step": 1633
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.2335355207324028,
      "epoch": 2.6185897435897436,
      "grad_norm": 0.008864963427186012,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "step": 1634
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.109375,
      "entropy": 0.23033030331134796,
      "epoch": 2.6201923076923075,
      "grad_norm": 0.015018898993730545,
      "learning_rate": 1e-06,
      "loss": -0.0089,
      "step": 1635
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1015625,
      "entropy": 0.23148179799318314,
      "epoch": 2.621794871794872,
      "grad_norm": 0.008959776721894741,
      "learning_rate": 1e-06,
      "loss": 0.0062,
      "step": 1636
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2425.0,
      "completions/mean_length": 1215.390625,
      "completions/mean_terminated_length": 1155.906005859375,
      "completions/min_length": 375.0,
      "completions/min_terminated_length": 375.0,
      "entropy": 0.2370157241821289,
      "epoch": 2.623397435897436,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.01420600526034832,
      "learning_rate": 1e-06,
      "loss": 0.0079,
      "num_tokens": 1051439463.0,
      "reward": 0.3070603907108307,
      "reward_std": 0.03020370379090309,
      "rewards/progression_diversity/mean": -0.0019726285245269537,
      "rewards/progression_diversity/std": 0.030601099133491516,
      "rewards/symbolic_reward_accuracy/mean": 0.16796875,
      "rewards/symbolic_reward_accuracy/std": 0.374204158782959,
      "rewards/symbolic_reward_partial_score/mean": 0.68896484375,
      "rewards/symbolic_reward_partial_score/std": 0.1941937357187271,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0498253107070923,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 1.9474784135818481,
      "step": 1637
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.23885345458984375,
      "epoch": 2.625,
      "grad_norm": 0.011274533346295357,
      "learning_rate": 1e-06,
      "loss": 0.024,
      "step": 1638
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.23263810575008392,
      "epoch": 2.626602564102564,
      "grad_norm": 0.008748682215809822,
      "learning_rate": 1e-06,
      "loss": 0.0306,
      "step": 1639
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.23605605959892273,
      "epoch": 2.628205128205128,
      "grad_norm": 0.020408501848578453,
      "learning_rate": 1e-06,
      "loss": -0.0049,
      "step": 1640
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2093.0,
      "completions/mean_length": 1151.375,
      "completions/mean_terminated_length": 1121.5655517578125,
      "completions/min_length": 338.0,
      "completions/min_terminated_length": 338.0,
      "entropy": 0.24552424252033234,
      "epoch": 2.6298076923076925,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.022998787462711334,
      "learning_rate": 1e-06,
      "loss": 0.0009,
      "num_tokens": 1052812775.0,
      "reward": 0.3075063228607178,
      "reward_std": 0.009556922130286694,
      "rewards/progression_diversity/mean": -0.0008343125809915364,
      "rewards/progression_diversity/std": 0.018091805279254913,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.7125488519668579,
      "rewards/symbolic_reward_partial_score/std": 0.18109886348247528,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0533249378204346,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 1.3622500896453857,
      "step": 1641
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.2431017979979515,
      "epoch": 2.6314102564102564,
      "grad_norm": 0.019999193027615547,
      "learning_rate": 1e-06,
      "loss": 0.0205,
      "step": 1642
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.24063821882009506,
      "epoch": 2.6330128205128203,
      "grad_norm": 0.01059049554169178,
      "learning_rate": 1e-06,
      "loss": -0.0094,
      "step": 1643
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.24199344217777252,
      "epoch": 2.6346153846153846,
      "grad_norm": 0.015171530656516552,
      "learning_rate": 1e-06,
      "loss": 0.0055,
      "step": 1644
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2033.0,
      "completions/mean_length": 1190.326171875,
      "completions/mean_terminated_length": 1160.5928955078125,
      "completions/min_length": 392.0,
      "completions/min_terminated_length": 392.0,
      "entropy": 0.24322400987148285,
      "epoch": 2.636217948717949,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.021913990378379822,
      "learning_rate": 1e-06,
      "loss": 0.0014,
      "num_tokens": 1054322382.0,
      "reward": 0.30744144320487976,
      "reward_std": 0.010562664829194546,
      "rewards/progression_diversity/mean": -0.0009744351846165955,
      "rewards/progression_diversity/std": 0.020204655826091766,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.7129882574081421,
      "rewards/symbolic_reward_partial_score/std": 0.17659372091293335,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.053013801574707,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 1.5290266275405884,
      "step": 1645
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2419656366109848,
      "epoch": 2.6378205128205128,
      "grad_norm": 0.02146642841398716,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "step": 1646
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.24547536671161652,
      "epoch": 2.6394230769230766,
      "grad_norm": 0.012392179109156132,
      "learning_rate": 1e-06,
      "loss": 0.0009,
      "step": 1647
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0390625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.23846372961997986,
      "epoch": 2.641025641025641,
      "grad_norm": 1201.439208984375,
      "learning_rate": 1e-06,
      "loss": 0.034,
      "step": 1648
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2423.0,
      "completions/mean_length": 1209.61328125,
      "completions/mean_terminated_length": 1179.9178466796875,
      "completions/min_length": 458.0,
      "completions/min_terminated_length": 458.0,
      "entropy": 0.24202610552310944,
      "epoch": 2.6426282051282053,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.014862586744129658,
      "learning_rate": 1e-06,
      "loss": 0.0009,
      "num_tokens": 1055858648.0,
      "reward": 0.3161856532096863,
      "reward_std": 0.019514337182044983,
      "rewards/progression_diversity/mean": -0.00106208142824471,
      "rewards/progression_diversity/std": 0.018535632640123367,
      "rewards/symbolic_reward_accuracy/mean": 0.19140625,
      "rewards/symbolic_reward_accuracy/std": 0.3937928080558777,
      "rewards/symbolic_reward_partial_score/mean": 0.6711751222610474,
      "rewards/symbolic_reward_partial_score/std": 0.2153313010931015,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0529862642288208,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 1.4675102233886719,
      "step": 1649
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2419712170958519,
      "epoch": 2.644230769230769,
      "grad_norm": 0.009646804071962833,
      "learning_rate": 1e-06,
      "loss": 0.0225,
      "step": 1650
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.24016830325126648,
      "epoch": 2.6458333333333335,
      "grad_norm": 0.02187509275972843,
      "learning_rate": 1e-06,
      "loss": -0.0017,
      "step": 1651
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.2460290938615799,
      "epoch": 2.6474358974358974,
      "grad_norm": 0.018824264407157898,
      "learning_rate": 1e-06,
      "loss": -0.0001,
      "step": 1652
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3980.0,
      "completions/max_terminated_length": 3980.0,
      "completions/mean_length": 1227.869140625,
      "completions/mean_terminated_length": 1227.869140625,
      "completions/min_length": 431.0,
      "completions/min_terminated_length": 431.0,
      "entropy": 0.24531883001327515,
      "epoch": 2.6490384615384617,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.15125811100006104,
      "learning_rate": 1e-06,
      "loss": 0.0129,
      "num_tokens": 1057351989.0,
      "reward": 0.260877788066864,
      "reward_std": 0.027386680245399475,
      "rewards/progression_diversity/mean": -0.00011138351692352444,
      "rewards/progression_diversity/std": 0.0017274393467232585,
      "rewards/symbolic_reward_accuracy/mean": 0.09765625,
      "rewards/symbolic_reward_accuracy/std": 0.29713961482048035,
      "rewards/symbolic_reward_partial_score/mean": 0.6742838621139526,
      "rewards/symbolic_reward_partial_score/std": 0.16698527336120605,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0578720569610596,
      "sampling/importance_sampling_ratio/min": 0.0001738879072945565,
      "sampling/sampling_logp_difference/max": 8.657099723815918,
      "sampling/sampling_logp_difference/mean": 0.11238566040992737,
      "step": 1653
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2520115375518799,
      "epoch": 2.6506410256410255,
      "grad_norm": 0.016538454219698906,
      "learning_rate": 1e-06,
      "loss": -0.0015,
      "step": 1654
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.25293679535388947,
      "epoch": 2.65224358974359,
      "grad_norm": 0.020968111231923103,
      "learning_rate": 1e-06,
      "loss": -0.0085,
      "step": 1655
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0390625,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.24919523298740387,
      "epoch": 2.6538461538461537,
      "grad_norm": 0.009686674922704697,
      "learning_rate": 1e-06,
      "loss": 0.0023,
      "step": 1656
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2423.0,
      "completions/mean_length": 1314.712890625,
      "completions/mean_terminated_length": 1225.8958740234375,
      "completions/min_length": 391.0,
      "completions/min_terminated_length": 391.0,
      "entropy": 0.25128719210624695,
      "epoch": 2.655448717948718,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 296.33270263671875,
      "learning_rate": 1e-06,
      "loss": 0.0172,
      "num_tokens": 1058869346.0,
      "reward": 0.3539027273654938,
      "reward_std": 0.02425507828593254,
      "rewards/progression_diversity/mean": -0.00230794376693666,
      "rewards/progression_diversity/std": 0.02924177795648575,
      "rewards/symbolic_reward_accuracy/mean": 0.236328125,
      "rewards/symbolic_reward_accuracy/std": 0.42524150013923645,
      "rewards/symbolic_reward_partial_score/mean": 0.7077473998069763,
      "rewards/symbolic_reward_partial_score/std": 0.22301146388053894,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0505740642547607,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 3.750328540802002,
      "step": 1657
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.24680255353450775,
      "epoch": 2.657051282051282,
      "grad_norm": 0.022292504087090492,
      "learning_rate": 1e-06,
      "loss": 0.0275,
      "step": 1658
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2595788687467575,
      "epoch": 2.6586538461538463,
      "grad_norm": 0.007560514844954014,
      "learning_rate": 1e-06,
      "loss": -0.0003,
      "step": 1659
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.25351230800151825,
      "epoch": 2.66025641025641,
      "grad_norm": 0.011664212681353092,
      "learning_rate": 1e-06,
      "loss": 0.0056,
      "step": 1660
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2635.0,
      "completions/mean_length": 1181.86328125,
      "completions/mean_terminated_length": 1152.113525390625,
      "completions/min_length": 343.0,
      "completions/min_terminated_length": 343.0,
      "entropy": 0.2603015750646591,
      "epoch": 2.6618589743589745,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.01761663518846035,
      "learning_rate": 1e-06,
      "loss": 0.0039,
      "num_tokens": 1060424988.0,
      "reward": 0.3372478485107422,
      "reward_std": 0.02446237951517105,
      "rewards/progression_diversity/mean": -0.0007998401415534317,
      "rewards/progression_diversity/std": 0.01764107309281826,
      "rewards/symbolic_reward_accuracy/mean": 0.181640625,
      "rewards/symbolic_reward_accuracy/std": 0.38592514395713806,
      "rewards/symbolic_reward_partial_score/mean": 0.7609050273895264,
      "rewards/symbolic_reward_partial_score/std": 0.18756793439388275,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0561124086380005,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 1.5122106075286865,
      "step": 1661
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2552846670150757,
      "epoch": 2.6634615384615383,
      "grad_norm": 0.009117074310779572,
      "learning_rate": 1e-06,
      "loss": 0.0035,
      "step": 1662
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.24909861385822296,
      "epoch": 2.6650641025641026,
      "grad_norm": 0.019102146849036217,
      "learning_rate": 1e-06,
      "loss": 0.0028,
      "step": 1663
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.25790928304195404,
      "epoch": 2.6666666666666665,
      "grad_norm": 0.010081185959279537,
      "learning_rate": 1e-06,
      "loss": -0.0069,
      "step": 1664
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2593.0,
      "completions/max_terminated_length": 2593.0,
      "completions/mean_length": 1168.1640625,
      "completions/mean_terminated_length": 1168.1640625,
      "completions/min_length": 350.0,
      "completions/min_terminated_length": 350.0,
      "entropy": 0.26866577565670013,
      "epoch": 2.668269230769231,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 0.02700965479016304,
      "learning_rate": 1e-06,
      "loss": 0.0046,
      "num_tokens": 1061838736.0,
      "reward": 0.32009169459342957,
      "reward_std": 0.021595628932118416,
      "rewards/progression_diversity/mean": -0.00010825111530721188,
      "rewards/progression_diversity/std": 0.0018366762669757009,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.6919758915901184,
      "rewards/symbolic_reward_partial_score/std": 0.20659130811691284,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0602461099624634,
      "sampling/importance_sampling_ratio/min": 7.974762411322445e-05,
      "sampling/sampling_logp_difference/max": 9.436643600463867,
      "sampling/sampling_logp_difference/mean": 0.11774517595767975,
      "step": 1665
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.125,
      "entropy": 0.2683710306882858,
      "epoch": 2.6698717948717947,
      "grad_norm": 0.02114093489944935,
      "learning_rate": 1e-06,
      "loss": -0.0047,
      "step": 1666
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.268236443400383,
      "epoch": 2.671474358974359,
      "grad_norm": 0.009307913482189178,
      "learning_rate": 1e-06,
      "loss": 0.0008,
      "step": 1667
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.2630407512187958,
      "epoch": 2.6730769230769234,
      "grad_norm": 0.01072956994175911,
      "learning_rate": 1e-06,
      "loss": -0.0018,
      "step": 1668
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2049.0,
      "completions/max_terminated_length": 2049.0,
      "completions/mean_length": 1157.5234375,
      "completions/mean_terminated_length": 1157.5234375,
      "completions/min_length": 318.0,
      "completions/min_terminated_length": 318.0,
      "entropy": 0.26273058354854584,
      "epoch": 2.6746794871794872,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.022477174177765846,
      "learning_rate": 1e-06,
      "loss": 0.0039,
      "num_tokens": 1063326540.0,
      "reward": 0.3207314610481262,
      "reward_std": 0.03389011323451996,
      "rewards/progression_diversity/mean": -9.633351874072105e-05,
      "rewards/progression_diversity/std": 0.0015180562622845173,
      "rewards/symbolic_reward_accuracy/mean": 0.17578125,
      "rewards/symbolic_reward_accuracy/std": 0.3810062110424042,
      "rewards/symbolic_reward_partial_score/mean": 0.7175455689430237,
      "rewards/symbolic_reward_partial_score/std": 0.2033698707818985,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0595388412475586,
      "sampling/importance_sampling_ratio/min": 0.0003233875031583011,
      "sampling/sampling_logp_difference/max": 8.036659240722656,
      "sampling/sampling_logp_difference/mean": 0.11578333377838135,
      "step": 1669
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.26235876977443695,
      "epoch": 2.676282051282051,
      "grad_norm": 0.01767921820282936,
      "learning_rate": 1e-06,
      "loss": 0.0022,
      "step": 1670
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.2589954137802124,
      "epoch": 2.6778846153846154,
      "grad_norm": 0.01393952313810587,
      "learning_rate": 1e-06,
      "loss": 0.0022,
      "step": 1671
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2695692628622055,
      "epoch": 2.6794871794871797,
      "grad_norm": 0.02042005956172943,
      "learning_rate": 1e-06,
      "loss": -0.0056,
      "step": 1672
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2282.0,
      "completions/max_terminated_length": 2282.0,
      "completions/mean_length": 1089.64453125,
      "completions/mean_terminated_length": 1089.64453125,
      "completions/min_length": 395.0,
      "completions/min_terminated_length": 395.0,
      "entropy": 0.2677207440137863,
      "epoch": 2.6810897435897436,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 0.010551492683589458,
      "learning_rate": 1e-06,
      "loss": -0.004,
      "num_tokens": 1064821670.0,
      "reward": 0.2943932116031647,
      "reward_std": 0.024308741092681885,
      "rewards/progression_diversity/mean": -0.00013180242967791855,
      "rewards/progression_diversity/std": 0.0023829475976526737,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.7156900763511658,
      "rewards/symbolic_reward_partial_score/std": 0.18010446429252625,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.060936450958252,
      "sampling/importance_sampling_ratio/min": 0.002229129895567894,
      "sampling/sampling_logp_difference/max": 6.106143951416016,
      "sampling/sampling_logp_difference/mean": 0.11975084245204926,
      "step": 1673
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.2699047178030014,
      "epoch": 2.6826923076923075,
      "grad_norm": 0.017179284244775772,
      "learning_rate": 1e-06,
      "loss": 0.0027,
      "step": 1674
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.26817895472049713,
      "epoch": 2.684294871794872,
      "grad_norm": 0.022273728623986244,
      "learning_rate": 1e-06,
      "loss": -0.0012,
      "step": 1675
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.27202919125556946,
      "epoch": 2.685897435897436,
      "grad_norm": 0.007613576482981443,
      "learning_rate": 1e-06,
      "loss": 0.0027,
      "step": 1676
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2164.0,
      "completions/max_terminated_length": 2164.0,
      "completions/mean_length": 1106.2890625,
      "completions/mean_terminated_length": 1106.2890625,
      "completions/min_length": 440.0,
      "completions/min_terminated_length": 440.0,
      "entropy": 0.27295637130737305,
      "epoch": 2.6875,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.018397819250822067,
      "learning_rate": 1e-06,
      "loss": 0.0038,
      "num_tokens": 1066328586.0,
      "reward": 0.21672746539115906,
      "reward_std": 0.02070911042392254,
      "rewards/progression_diversity/mean": -0.00010542837844695896,
      "rewards/progression_diversity/std": 0.0020016725175082684,
      "rewards/symbolic_reward_accuracy/mean": 0.02734375,
      "rewards/symbolic_reward_accuracy/std": 0.16324250400066376,
      "rewards/symbolic_reward_partial_score/mean": 0.6677408814430237,
      "rewards/symbolic_reward_partial_score/std": 0.1652555614709854,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.062442421913147,
      "sampling/importance_sampling_ratio/min": 0.002278130268678069,
      "sampling/sampling_logp_difference/max": 6.084400177001953,
      "sampling/sampling_logp_difference/mean": 0.12101897597312927,
      "step": 1677
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2777542769908905,
      "epoch": 2.689102564102564,
      "grad_norm": 0.009125451557338238,
      "learning_rate": 1e-06,
      "loss": 0.0035,
      "step": 1678
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.273501455783844,
      "epoch": 2.690705128205128,
      "grad_norm": 0.010436906479299068,
      "learning_rate": 1e-06,
      "loss": 0.0021,
      "step": 1679
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2719372361898422,
      "epoch": 2.6923076923076925,
      "grad_norm": 0.012276146560907364,
      "learning_rate": 1e-06,
      "loss": -0.0069,
      "step": 1680
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2715.0,
      "completions/max_terminated_length": 2715.0,
      "completions/mean_length": 1127.42578125,
      "completions/mean_terminated_length": 1127.42578125,
      "completions/min_length": 412.0,
      "completions/min_terminated_length": 412.0,
      "entropy": 0.27352291345596313,
      "epoch": 2.6939102564102564,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 0.020503515377640724,
      "learning_rate": 1e-06,
      "loss": -0.0016,
      "num_tokens": 1067652580.0,
      "reward": 0.39523687958717346,
      "reward_std": 0.03253398835659027,
      "rewards/progression_diversity/mean": -0.0002392895403318107,
      "rewards/progression_diversity/std": 0.004227474331855774,
      "rewards/symbolic_reward_accuracy/mean": 0.2578125,
      "rewards/symbolic_reward_accuracy/std": 0.43785804510116577,
      "rewards/symbolic_reward_partial_score/mean": 0.8018391728401184,
      "rewards/symbolic_reward_partial_score/std": 0.16659002006053925,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0630683898925781,
      "sampling/importance_sampling_ratio/min": 0.0012960727326571941,
      "sampling/sampling_logp_difference/max": 6.648416519165039,
      "sampling/sampling_logp_difference/mean": 0.12232454121112823,
      "step": 1681
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.2841361165046692,
      "epoch": 2.6955128205128203,
      "grad_norm": 0.013175196945667267,
      "learning_rate": 1e-06,
      "loss": 0.0021,
      "step": 1682
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.2792322635650635,
      "epoch": 2.6971153846153846,
      "grad_norm": 0.008740157820284367,
      "learning_rate": 1e-06,
      "loss": -0.0037,
      "step": 1683
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0390625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.2852344512939453,
      "epoch": 2.698717948717949,
      "grad_norm": 0.007328695617616177,
      "learning_rate": 1e-06,
      "loss": 0.0062,
      "step": 1684
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2764.0,
      "completions/max_terminated_length": 2764.0,
      "completions/mean_length": 1182.7734375,
      "completions/mean_terminated_length": 1182.7734375,
      "completions/min_length": 393.0,
      "completions/min_terminated_length": 393.0,
      "entropy": 0.27371834218502045,
      "epoch": 2.7003205128205128,
      "frac_reward_zero_std": 0.625,
      "grad_norm": 0.016033172607421875,
      "learning_rate": 1e-06,
      "loss": -0.0049,
      "num_tokens": 1069083168.0,
      "reward": 0.3871532678604126,
      "reward_std": 0.036548394709825516,
      "rewards/progression_diversity/mean": -8.401693776249886e-06,
      "rewards/progression_diversity/std": 0.00019010863616131246,
      "rewards/symbolic_reward_accuracy/mean": 0.265625,
      "rewards/symbolic_reward_accuracy/std": 0.44209739565849304,
      "rewards/symbolic_reward_partial_score/mean": 0.7592610716819763,
      "rewards/symbolic_reward_partial_score/std": 0.18804891407489777,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0619139671325684,
      "sampling/importance_sampling_ratio/min": 6.210852006915957e-05,
      "sampling/sampling_logp_difference/max": 9.686627388000488,
      "sampling/sampling_logp_difference/mean": 0.1205776035785675,
      "step": 1685
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.27223040163517,
      "epoch": 2.7019230769230766,
      "grad_norm": 0.009487297385931015,
      "learning_rate": 1e-06,
      "loss": 0.0021,
      "step": 1686
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.2768988460302353,
      "epoch": 2.703525641025641,
      "grad_norm": 0.010462704114615917,
      "learning_rate": 1e-06,
      "loss": 0.0031,
      "step": 1687
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.2717527002096176,
      "epoch": 2.7051282051282053,
      "grad_norm": 0.016962876543402672,
      "learning_rate": 1e-06,
      "loss": 0.003,
      "step": 1688
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2341.0,
      "completions/max_terminated_length": 2341.0,
      "completions/mean_length": 1154.96875,
      "completions/mean_terminated_length": 1154.96875,
      "completions/min_length": 379.0,
      "completions/min_terminated_length": 379.0,
      "entropy": 0.2763988673686981,
      "epoch": 2.706730769230769,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 0.011929732747375965,
      "learning_rate": 1e-06,
      "loss": 0.0029,
      "num_tokens": 1070536032.0,
      "reward": 0.3324165344238281,
      "reward_std": 0.020372124388813972,
      "rewards/progression_diversity/mean": -4.5338445488596335e-05,
      "rewards/progression_diversity/std": 0.0006139642791822553,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.7408691644668579,
      "rewards/symbolic_reward_partial_score/std": 0.18128949403762817,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0632294416427612,
      "sampling/importance_sampling_ratio/min": 0.0005881256074644625,
      "sampling/sampling_logp_difference/max": 7.438570022583008,
      "sampling/sampling_logp_difference/mean": 0.12137885391712189,
      "step": 1689
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.27255162596702576,
      "epoch": 2.7083333333333335,
      "grad_norm": 0.02167939394712448,
      "learning_rate": 1e-06,
      "loss": 0.0019,
      "step": 1690
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.28303536772727966,
      "epoch": 2.7099358974358974,
      "grad_norm": 0.008452474139630795,
      "learning_rate": 1e-06,
      "loss": -0.0025,
      "step": 1691
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.2690129578113556,
      "epoch": 2.7115384615384617,
      "grad_norm": 0.010433075949549675,
      "learning_rate": 1e-06,
      "loss": 0.0047,
      "step": 1692
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2530.0,
      "completions/mean_length": 1084.271484375,
      "completions/mean_terminated_length": 1054.3306884765625,
      "completions/min_length": 290.0,
      "completions/min_terminated_length": 290.0,
      "entropy": 0.2840660959482193,
      "epoch": 2.7131410256410255,
      "frac_reward_zero_std": 0.65625,
      "grad_norm": 0.007845147512853146,
      "learning_rate": 1e-06,
      "loss": -0.0006,
      "num_tokens": 1071857355.0,
      "reward": 0.41788747906684875,
      "reward_std": 0.011897813528776169,
      "rewards/progression_diversity/mean": -0.0008051774930208921,
      "rewards/progression_diversity/std": 0.01665707863867283,
      "rewards/symbolic_reward_accuracy/mean": 0.310546875,
      "rewards/symbolic_reward_accuracy/std": 0.46317005157470703,
      "rewards/symbolic_reward_partial_score/mean": 0.7718912363052368,
      "rewards/symbolic_reward_partial_score/std": 0.1966981589794159,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0603790283203125,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 1.6770976781845093,
      "step": 1693
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.27424998581409454,
      "epoch": 2.71474358974359,
      "grad_norm": 0.011158179491758347,
      "learning_rate": 1e-06,
      "loss": 0.0195,
      "step": 1694
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.279265433549881,
      "epoch": 2.7163461538461537,
      "grad_norm": 0.012044363655149937,
      "learning_rate": 1e-06,
      "loss": 0.0047,
      "step": 1695
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.27725885808467865,
      "epoch": 2.717948717948718,
      "grad_norm": 0.004128993488848209,
      "learning_rate": 1e-06,
      "loss": -0.0004,
      "step": 1696
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2449.0,
      "completions/mean_length": 1174.107421875,
      "completions/mean_terminated_length": 1084.461669921875,
      "completions/min_length": 370.0,
      "completions/min_terminated_length": 370.0,
      "entropy": 0.2545585185289383,
      "epoch": 2.719551282051282,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.01764233596622944,
      "learning_rate": 1e-06,
      "loss": -0.0001,
      "num_tokens": 1073390098.0,
      "reward": 0.3306075632572174,
      "reward_std": 0.017754849046468735,
      "rewards/progression_diversity/mean": -0.002232456346973777,
      "rewards/progression_diversity/std": 0.02855554409325123,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.731005847454071,
      "rewards/symbolic_reward_partial_score/std": 0.17919674515724182,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0483648777008057,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 5.032663345336914,
      "step": 1697
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.2522525489330292,
      "epoch": 2.7211538461538463,
      "grad_norm": 0.014643147587776184,
      "learning_rate": 1e-06,
      "loss": -0.0005,
      "step": 1698
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.23972028493881226,
      "epoch": 2.72275641025641,
      "grad_norm": 0.014284703880548477,
      "learning_rate": 1e-06,
      "loss": 0.0231,
      "step": 1699
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.2528829574584961,
      "epoch": 2.7243589743589745,
      "grad_norm": 0.007763241417706013,
      "learning_rate": 1e-06,
      "loss": 0.0057,
      "step": 1700
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2125.0,
      "completions/mean_length": 1085.763671875,
      "completions/mean_terminated_length": 1055.8258056640625,
      "completions/min_length": 360.0,
      "completions/min_terminated_length": 360.0,
      "entropy": 0.25211136788129807,
      "epoch": 2.7259615384615383,
      "frac_reward_zero_std": 0.625,
      "grad_norm": 0.01731860637664795,
      "learning_rate": 1e-06,
      "loss": 0.0223,
      "num_tokens": 1074838889.0,
      "reward": 0.37423521280288696,
      "reward_std": 0.012984178960323334,
      "rewards/progression_diversity/mean": -0.0007978131761774421,
      "rewards/progression_diversity/std": 0.017717914655804634,
      "rewards/symbolic_reward_accuracy/mean": 0.24609375,
      "rewards/symbolic_reward_accuracy/std": 0.4311550557613373,
      "rewards/symbolic_reward_partial_score/mean": 0.7552896738052368,
      "rewards/symbolic_reward_partial_score/std": 0.20005594193935394,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0550222396850586,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 1.6980559825897217,
      "step": 1701
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.109375,
      "entropy": 0.24782519042491913,
      "epoch": 2.7275641025641026,
      "grad_norm": 0.01143115945160389,
      "learning_rate": 1e-06,
      "loss": 0.001,
      "step": 1702
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.25386030972003937,
      "epoch": 2.7291666666666665,
      "grad_norm": 0.010584347881376743,
      "learning_rate": 1e-06,
      "loss": -0.001,
      "step": 1703
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.25097061693668365,
      "epoch": 2.730769230769231,
      "grad_norm": 0.014512878842651844,
      "learning_rate": 1e-06,
      "loss": -0.0028,
      "step": 1704
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 1766.0,
      "completions/mean_length": 1077.181640625,
      "completions/mean_terminated_length": 1017.1549682617188,
      "completions/min_length": 387.0,
      "completions/min_terminated_length": 387.0,
      "entropy": 0.2463693991303444,
      "epoch": 2.7323717948717947,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.022319326177239418,
      "learning_rate": 1e-06,
      "loss": 0.0029,
      "num_tokens": 1076451366.0,
      "reward": 0.3169002830982208,
      "reward_std": 0.018300164490938187,
      "rewards/progression_diversity/mean": -0.0013787832576781511,
      "rewards/progression_diversity/std": 0.02173839509487152,
      "rewards/symbolic_reward_accuracy/mean": 0.16796875,
      "rewards/symbolic_reward_accuracy/std": 0.374204158782959,
      "rewards/symbolic_reward_partial_score/mean": 0.7204427123069763,
      "rewards/symbolic_reward_partial_score/std": 0.20151053369045258,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0520621538162231,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 3.0996155738830566,
      "step": 1705
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.24724167585372925,
      "epoch": 2.733974358974359,
      "grad_norm": 465.19256591796875,
      "learning_rate": 1e-06,
      "loss": 0.0165,
      "step": 1706
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.25137076526880264,
      "epoch": 2.7355769230769234,
      "grad_norm": 0.011684092693030834,
      "learning_rate": 1e-06,
      "loss": 0.0032,
      "step": 1707
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.24099694192409515,
      "epoch": 2.7371794871794872,
      "grad_norm": 0.01180903147906065,
      "learning_rate": 1e-06,
      "loss": -0.0007,
      "step": 1708
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2634.0,
      "completions/mean_length": 1115.263671875,
      "completions/mean_terminated_length": 1025.2711181640625,
      "completions/min_length": 357.0,
      "completions/min_terminated_length": 357.0,
      "entropy": 0.24923396110534668,
      "epoch": 2.738782051282051,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 0.022832239046692848,
      "learning_rate": 1e-06,
      "loss": -0.0048,
      "num_tokens": 1077919117.0,
      "reward": 0.4506452679634094,
      "reward_std": 0.006873677484691143,
      "rewards/progression_diversity/mean": -0.002368855057284236,
      "rewards/progression_diversity/std": 0.02794456295669079,
      "rewards/symbolic_reward_accuracy/mean": 0.34375,
      "rewards/symbolic_reward_accuracy/std": 0.4754233956336975,
      "rewards/symbolic_reward_partial_score/mean": 0.8147298097610474,
      "rewards/symbolic_reward_partial_score/std": 0.17687095701694489,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0485377311706543,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 4.422557353973389,
      "step": 1709
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.24275388568639755,
      "epoch": 2.7403846153846154,
      "grad_norm": 0.020597660914063454,
      "learning_rate": 1e-06,
      "loss": 0.0452,
      "step": 1710
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.25289829075336456,
      "epoch": 2.7419871794871797,
      "grad_norm": 0.005713851656764746,
      "learning_rate": 1e-06,
      "loss": -0.0012,
      "step": 1711
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1015625,
      "entropy": 0.2434333637356758,
      "epoch": 2.7435897435897436,
      "grad_norm": 0.010084899142384529,
      "learning_rate": 1e-06,
      "loss": 0.0226,
      "step": 1712
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2047.0,
      "completions/mean_length": 1117.982421875,
      "completions/mean_terminated_length": 1088.107666015625,
      "completions/min_length": 346.0,
      "completions/min_terminated_length": 346.0,
      "entropy": 0.24176087975502014,
      "epoch": 2.7451923076923075,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.014847025275230408,
      "learning_rate": 1e-06,
      "loss": -0.023,
      "num_tokens": 1079362852.0,
      "reward": 0.4161057472229004,
      "reward_std": 0.020435160025954247,
      "rewards/progression_diversity/mean": -0.0012444315943866968,
      "rewards/progression_diversity/std": 0.017949236556887627,
      "rewards/symbolic_reward_accuracy/mean": 0.302734375,
      "rewards/symbolic_reward_accuracy/std": 0.45989060401916504,
      "rewards/symbolic_reward_partial_score/mean": 0.7815917730331421,
      "rewards/symbolic_reward_partial_score/std": 0.19830654561519623,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.052306890487671,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 1.7721519470214844,
      "step": 1713
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.2418888807296753,
      "epoch": 2.746794871794872,
      "grad_norm": 0.005131000652909279,
      "learning_rate": 1e-06,
      "loss": -0.0005,
      "step": 1714
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.24066882580518723,
      "epoch": 2.748397435897436,
      "grad_norm": 0.012241684831678867,
      "learning_rate": 1e-06,
      "loss": 0.0029,
      "step": 1715
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.2462095245718956,
      "epoch": 2.75,
      "grad_norm": 0.011896181851625443,
      "learning_rate": 1e-06,
      "loss": -0.0003,
      "step": 1716
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 1879.0,
      "completions/mean_length": 1122.0,
      "completions/mean_terminated_length": 1062.1490478515625,
      "completions/min_length": 296.0,
      "completions/min_terminated_length": 296.0,
      "entropy": 0.24476610869169235,
      "epoch": 2.751602564102564,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.011216786690056324,
      "learning_rate": 1e-06,
      "loss": -0.0053,
      "num_tokens": 1080775556.0,
      "reward": 0.31833428144454956,
      "reward_std": 0.009690433740615845,
      "rewards/progression_diversity/mean": -0.0015309633454307914,
      "rewards/progression_diversity/std": 0.02212413214147091,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.7486653923988342,
      "rewards/symbolic_reward_partial_score/std": 0.18320314586162567,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.050942063331604,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.9385156631469727,
      "step": 1717
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.23835279792547226,
      "epoch": 2.753205128205128,
      "grad_norm": 0.012194461189210415,
      "learning_rate": 1e-06,
      "loss": 0.232,
      "step": 1718
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.24551533162593842,
      "epoch": 2.7548076923076925,
      "grad_norm": 0.015713347122073174,
      "learning_rate": 1e-06,
      "loss": 0.0023,
      "step": 1719
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.25484059751033783,
      "epoch": 2.7564102564102564,
      "grad_norm": 0.01567044109106064,
      "learning_rate": 1e-06,
      "loss": -0.001,
      "step": 1720
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2034.0,
      "completions/max_terminated_length": 2034.0,
      "completions/mean_length": 970.169921875,
      "completions/mean_terminated_length": 970.169921875,
      "completions/min_length": 289.0,
      "completions/min_terminated_length": 289.0,
      "entropy": 0.249983012676239,
      "epoch": 2.7580128205128203,
      "frac_reward_zero_std": 0.78125,
      "grad_norm": 0.010865814052522182,
      "learning_rate": 1e-06,
      "loss": 0.0003,
      "num_tokens": 1082279163.0,
      "reward": 0.2930664122104645,
      "reward_std": 0.005064056254923344,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.7268880009651184,
      "rewards/symbolic_reward_partial_score/std": 0.17721883952617645,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0566366910934448,
      "sampling/importance_sampling_ratio/min": 0.0014420977095142007,
      "sampling/sampling_logp_difference/max": 6.541656494140625,
      "sampling/sampling_logp_difference/mean": 0.11134675145149231,
      "step": 1721
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1171875,
      "entropy": 0.25236815214157104,
      "epoch": 2.7596153846153846,
      "grad_norm": 0.0043575153686106205,
      "learning_rate": 1e-06,
      "loss": -0.0021,
      "step": 1722
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0859375,
      "entropy": 0.25090380012989044,
      "epoch": 2.761217948717949,
      "grad_norm": 0.00688148895278573,
      "learning_rate": 1e-06,
      "loss": -0.0008,
      "step": 1723
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.015625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0625,
      "entropy": 0.2414015531539917,
      "epoch": 2.7628205128205128,
      "grad_norm": 0.0044527724385261536,
      "learning_rate": 1e-06,
      "loss": 0.002,
      "step": 1724
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 1888.0,
      "completions/mean_length": 1074.3984375,
      "completions/mean_terminated_length": 1044.4383544921875,
      "completions/min_length": 373.0,
      "completions/min_terminated_length": 373.0,
      "entropy": 0.26541852951049805,
      "epoch": 2.7644230769230766,
      "frac_reward_zero_std": 0.65625,
      "grad_norm": 0.007640301249921322,
      "learning_rate": 1e-06,
      "loss": -0.0031,
      "num_tokens": 1083616935.0,
      "reward": 0.39704516530036926,
      "reward_std": 0.01641342230141163,
      "rewards/progression_diversity/mean": -0.0005633344408124685,
      "rewards/progression_diversity/std": 0.01171040628105402,
      "rewards/symbolic_reward_accuracy/mean": 0.2890625,
      "rewards/symbolic_reward_accuracy/std": 0.45377036929130554,
      "rewards/symbolic_reward_partial_score/mean": 0.7453775405883789,
      "rewards/symbolic_reward_partial_score/std": 0.20832401514053345,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0565519332885742,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 1.417606234550476,
      "step": 1725
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.26599733531475067,
      "epoch": 2.766025641025641,
      "grad_norm": 0.016951909288764,
      "learning_rate": 1e-06,
      "loss": -0.002,
      "step": 1726
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.25999362766742706,
      "epoch": 2.7676282051282053,
      "grad_norm": 1.5784918069839478,
      "learning_rate": 1e-06,
      "loss": 0.0046,
      "step": 1727
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2638109028339386,
      "epoch": 2.769230769230769,
      "grad_norm": 0.009875231422483921,
      "learning_rate": 1e-06,
      "loss": -0.0014,
      "step": 1728
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2051.0,
      "completions/mean_length": 1121.234375,
      "completions/mean_terminated_length": 1061.3804931640625,
      "completions/min_length": 384.0,
      "completions/min_terminated_length": 384.0,
      "entropy": 0.2469266802072525,
      "epoch": 2.7708333333333335,
      "frac_reward_zero_std": 0.625,
      "grad_norm": 0.012807636521756649,
      "learning_rate": 1e-06,
      "loss": 0.0239,
      "num_tokens": 1085117135.0,
      "reward": 0.3859965205192566,
      "reward_std": 0.007197174243628979,
      "rewards/progression_diversity/mean": -0.0014215593691915274,
      "rewards/progression_diversity/std": 0.021102240309119225,
      "rewards/symbolic_reward_accuracy/mean": 0.28125,
      "rewards/symbolic_reward_accuracy/std": 0.45004892349243164,
      "rewards/symbolic_reward_partial_score/mean": 0.7242025136947632,
      "rewards/symbolic_reward_partial_score/std": 0.20905202627182007,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0523045063018799,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 2.6319713592529297,
      "step": 1729
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.2536063939332962,
      "epoch": 2.7724358974358974,
      "grad_norm": 0.009693522937595844,
      "learning_rate": 1e-06,
      "loss": -0.0024,
      "step": 1730
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.24922660738229752,
      "epoch": 2.7740384615384617,
      "grad_norm": 0.0067869494669139385,
      "learning_rate": 1e-06,
      "loss": 0.0263,
      "step": 1731
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.25212302803993225,
      "epoch": 2.7756410256410255,
      "grad_norm": 0.003368781413882971,
      "learning_rate": 1e-06,
      "loss": -0.0037,
      "step": 1732
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 1915.0,
      "completions/mean_length": 1387.89453125,
      "completions/mean_terminated_length": 1149.8612060546875,
      "completions/min_length": 354.0,
      "completions/min_terminated_length": 354.0,
      "entropy": 0.2443920373916626,
      "epoch": 2.77724358974359,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 510.38726806640625,
      "learning_rate": 1e-06,
      "loss": 0.0371,
      "num_tokens": 1086770873.0,
      "reward": 0.28071457147598267,
      "reward_std": 0.026247015222907066,
      "rewards/progression_diversity/mean": -0.0052025578916072845,
      "rewards/progression_diversity/std": 0.04079189524054527,
      "rewards/symbolic_reward_accuracy/mean": 0.115234375,
      "rewards/symbolic_reward_accuracy/std": 0.3196168541908264,
      "rewards/symbolic_reward_partial_score/mean": 0.7054198980331421,
      "rewards/symbolic_reward_partial_score/std": 0.18301890790462494,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0395216941833496,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 8.083383560180664,
      "step": 1733
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.23890351504087448,
      "epoch": 2.7788461538461537,
      "grad_norm": 0.005275333765894175,
      "learning_rate": 1e-06,
      "loss": 0.0321,
      "step": 1734
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2496597245335579,
      "epoch": 2.780448717948718,
      "grad_norm": 0.018677234649658203,
      "learning_rate": 1e-06,
      "loss": -0.0038,
      "step": 1735
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.23257097601890564,
      "epoch": 2.782051282051282,
      "grad_norm": 0.006251712329685688,
      "learning_rate": 1e-06,
      "loss": 0.0448,
      "step": 1736
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2008.0,
      "completions/mean_length": 1218.24609375,
      "completions/mean_terminated_length": 1128.860595703125,
      "completions/min_length": 369.0,
      "completions/min_terminated_length": 369.0,
      "entropy": 0.25235116481781006,
      "epoch": 2.7836538461538463,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 480.960693359375,
      "learning_rate": 1e-06,
      "loss": 0.0245,
      "num_tokens": 1088287719.0,
      "reward": 0.30648308992385864,
      "reward_std": 0.03177080303430557,
      "rewards/progression_diversity/mean": -0.0020833953749388456,
      "rewards/progression_diversity/std": 0.02671230398118496,
      "rewards/symbolic_reward_accuracy/mean": 0.150390625,
      "rewards/symbolic_reward_accuracy/std": 0.35780346393585205,
      "rewards/symbolic_reward_partial_score/mean": 0.7208983898162842,
      "rewards/symbolic_reward_partial_score/std": 0.19757305085659027,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0507006645202637,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 3.8391146659851074,
      "step": 1737
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.2659503221511841,
      "epoch": 2.78525641025641,
      "grad_norm": 0.006545082200318575,
      "learning_rate": 1e-06,
      "loss": 0.0058,
      "step": 1738
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.24851246178150177,
      "epoch": 2.7868589743589745,
      "grad_norm": 0.011122615076601505,
      "learning_rate": 1e-06,
      "loss": 0.0275,
      "step": 1739
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.26005537807941437,
      "epoch": 2.7884615384615383,
      "grad_norm": 0.01145921926945448,
      "learning_rate": 1e-06,
      "loss": -0.0071,
      "step": 1740
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3001.0,
      "completions/mean_length": 1462.2109375,
      "completions/mean_terminated_length": 1164.9642333984375,
      "completions/min_length": 367.0,
      "completions/min_terminated_length": 367.0,
      "entropy": 0.24173830449581146,
      "epoch": 2.7900641025641026,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 95.36475372314453,
      "learning_rate": 1e-06,
      "loss": 0.0115,
      "num_tokens": 1089973347.0,
      "reward": 0.2687704563140869,
      "reward_std": 0.03094436414539814,
      "rewards/progression_diversity/mean": -0.007230142131447792,
      "rewards/progression_diversity/std": 0.051848724484443665,
      "rewards/symbolic_reward_accuracy/mean": 0.08984375,
      "rewards/symbolic_reward_accuracy/std": 0.2862374484539032,
      "rewards/symbolic_reward_partial_score/mean": 0.7164551019668579,
      "rewards/symbolic_reward_partial_score/std": 0.19291111826896667,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.028411865234375,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 14.061389923095703,
      "step": 1741
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.24212167412042618,
      "epoch": 2.7916666666666665,
      "grad_norm": 0.023332608863711357,
      "learning_rate": 1e-06,
      "loss": 0.0322,
      "step": 1742
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.24151303619146347,
      "epoch": 2.793269230769231,
      "grad_norm": 0.02178528904914856,
      "learning_rate": 1e-06,
      "loss": 0.0041,
      "step": 1743
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.2449488863348961,
      "epoch": 2.7948717948717947,
      "grad_norm": 0.00785065721720457,
      "learning_rate": 1e-06,
      "loss": 0.0433,
      "step": 1744
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2171.0,
      "completions/mean_length": 1198.900390625,
      "completions/mean_terminated_length": 1139.35107421875,
      "completions/min_length": 294.0,
      "completions/min_terminated_length": 294.0,
      "entropy": 0.2584023177623749,
      "epoch": 2.796474358974359,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.017909616231918335,
      "learning_rate": 1e-06,
      "loss": 0.0197,
      "num_tokens": 1091436528.0,
      "reward": 0.4526541233062744,
      "reward_std": 0.034945450723171234,
      "rewards/progression_diversity/mean": -0.0016795285046100616,
      "rewards/progression_diversity/std": 0.023316482082009315,
      "rewards/symbolic_reward_accuracy/mean": 0.37109375,
      "rewards/symbolic_reward_accuracy/std": 0.4835699498653412,
      "rewards/symbolic_reward_partial_score/mean": 0.7667155265808105,
      "rewards/symbolic_reward_partial_score/std": 0.22165250778198242,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0529496669769287,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 3.4778664112091064,
      "step": 1745
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.25567905604839325,
      "epoch": 2.7980769230769234,
      "grad_norm": 1.9885333776474,
      "learning_rate": 1e-06,
      "loss": -0.001,
      "step": 1746
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.26071760058403015,
      "epoch": 2.7996794871794872,
      "grad_norm": 0.01138362381607294,
      "learning_rate": 1e-06,
      "loss": 0.0018,
      "step": 1747
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.2582208067178726,
      "epoch": 2.801282051282051,
      "grad_norm": 0.01637045480310917,
      "learning_rate": 1e-06,
      "loss": 0.0035,
      "step": 1748
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2307.0,
      "completions/mean_length": 1168.337890625,
      "completions/mean_terminated_length": 1078.658203125,
      "completions/min_length": 308.0,
      "completions/min_terminated_length": 308.0,
      "entropy": 0.26668301224708557,
      "epoch": 2.8028846153846154,
      "frac_reward_zero_std": 0.6875,
      "grad_norm": 0.014873064123094082,
      "learning_rate": 1e-06,
      "loss": -0.0023,
      "num_tokens": 1092912237.0,
      "reward": 0.3330345153808594,
      "reward_std": 0.009103953838348389,
      "rewards/progression_diversity/mean": -0.0026997888926416636,
      "rewards/progression_diversity/std": 0.03452081233263016,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.7358561158180237,
      "rewards/symbolic_reward_partial_score/std": 0.1908014863729477,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0509870052337646,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 4.842784881591797,
      "step": 1749
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0859375,
      "entropy": 0.26296547055244446,
      "epoch": 2.8044871794871797,
      "grad_norm": 0.014554469846189022,
      "learning_rate": 1e-06,
      "loss": 0.0075,
      "step": 1750
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.125,
      "entropy": 0.25916317105293274,
      "epoch": 2.8060897435897436,
      "grad_norm": 0.012337586842477322,
      "learning_rate": 1e-06,
      "loss": 0.0315,
      "step": 1751
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0390625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1171875,
      "entropy": 0.2664582431316376,
      "epoch": 2.8076923076923075,
      "grad_norm": 0.006437097210437059,
      "learning_rate": 1e-06,
      "loss": -0.0017,
      "step": 1752
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2030.0,
      "completions/max_terminated_length": 2030.0,
      "completions/mean_length": 1078.322265625,
      "completions/mean_terminated_length": 1078.322265625,
      "completions/min_length": 334.0,
      "completions/min_terminated_length": 334.0,
      "entropy": 0.26388947665691376,
      "epoch": 2.809294871794872,
      "frac_reward_zero_std": 0.6875,
      "grad_norm": 0.024757713079452515,
      "learning_rate": 1e-06,
      "loss": -0.0036,
      "num_tokens": 1094326178.0,
      "reward": 0.39402249455451965,
      "reward_std": 0.020639346912503242,
      "rewards/progression_diversity/mean": -9.485271584708244e-05,
      "rewards/progression_diversity/std": 0.0016233575297519565,
      "rewards/symbolic_reward_accuracy/mean": 0.275390625,
      "rewards/symbolic_reward_accuracy/std": 0.44714778661727905,
      "rewards/symbolic_reward_partial_score/mean": 0.7626302242279053,
      "rewards/symbolic_reward_partial_score/std": 0.1950385868549347,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0604701042175293,
      "sampling/importance_sampling_ratio/min": 0.00016087631229311228,
      "sampling/sampling_logp_difference/max": 8.734874725341797,
      "sampling/sampling_logp_difference/mean": 0.11699830740690231,
      "step": 1753
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.125,
      "entropy": 0.26401379704475403,
      "epoch": 2.810897435897436,
      "grad_norm": 0.006932654418051243,
      "learning_rate": 1e-06,
      "loss": 0.0043,
      "step": 1754
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1015625,
      "entropy": 0.2620585560798645,
      "epoch": 2.8125,
      "grad_norm": 0.00885852426290512,
      "learning_rate": 1e-06,
      "loss": 0.0047,
      "step": 1755
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1015625,
      "entropy": 0.26447711884975433,
      "epoch": 2.814102564102564,
      "grad_norm": 0.008724554441869259,
      "learning_rate": 1e-06,
      "loss": -0.005,
      "step": 1756
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2611.0,
      "completions/mean_length": 1278.806640625,
      "completions/mean_terminated_length": 1099.6937255859375,
      "completions/min_length": 360.0,
      "completions/min_terminated_length": 360.0,
      "entropy": 0.2583252787590027,
      "epoch": 2.815705128205128,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 0.008783966302871704,
      "learning_rate": 1e-06,
      "loss": 0.0674,
      "num_tokens": 1095875567.0,
      "reward": 0.38711655139923096,
      "reward_std": 0.0257473886013031,
      "rewards/progression_diversity/mean": -0.006121981889009476,
      "rewards/progression_diversity/std": 0.05701431632041931,
      "rewards/symbolic_reward_accuracy/mean": 0.27734375,
      "rewards/symbolic_reward_accuracy/std": 0.4481254518032074,
      "rewards/symbolic_reward_partial_score/mean": 0.7378580570220947,
      "rewards/symbolic_reward_partial_score/std": 0.21705490350723267,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0447036027908325,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 7.182675838470459,
      "step": 1757
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.273300439119339,
      "epoch": 2.8173076923076925,
      "grad_norm": 0.01463186927139759,
      "learning_rate": 1e-06,
      "loss": -0.0055,
      "step": 1758
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.125,
      "entropy": 0.2602022588253021,
      "epoch": 2.8189102564102564,
      "grad_norm": 0.013082200661301613,
      "learning_rate": 1e-06,
      "loss": 0.0554,
      "step": 1759
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.26440712809562683,
      "epoch": 2.8205128205128203,
      "grad_norm": 0.010828658938407898,
      "learning_rate": 1e-06,
      "loss": 0.0194,
      "step": 1760
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.001953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2079.0,
      "completions/mean_length": 1091.5390625,
      "completions/mean_terminated_length": 1061.612548828125,
      "completions/min_length": 333.0,
      "completions/min_terminated_length": 333.0,
      "entropy": 0.28021863102912903,
      "epoch": 2.8221153846153846,
      "frac_reward_zero_std": 0.65625,
      "grad_norm": 0.00968155823647976,
      "learning_rate": 1e-06,
      "loss": -0.0016,
      "num_tokens": 1097272771.0,
      "reward": 0.33711838722229004,
      "reward_std": 0.00528107350692153,
      "rewards/progression_diversity/mean": -0.0010524296667426825,
      "rewards/progression_diversity/std": 0.02093338593840599,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.7487630248069763,
      "rewards/symbolic_reward_partial_score/std": 0.18481901288032532,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0589430332183838,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 1.9359979629516602,
      "step": 1761
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.2707153856754303,
      "epoch": 2.823717948717949,
      "grad_norm": 0.007914133369922638,
      "learning_rate": 1e-06,
      "loss": 0.001,
      "step": 1762
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1171875,
      "entropy": 0.2671675831079483,
      "epoch": 2.8253205128205128,
      "grad_norm": 0.016453703865408897,
      "learning_rate": 1e-06,
      "loss": 0.4248,
      "step": 1763
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.27130477130413055,
      "epoch": 2.8269230769230766,
      "grad_norm": 0.009142329916357994,
      "learning_rate": 1e-06,
      "loss": -0.0026,
      "step": 1764
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2041.0,
      "completions/mean_length": 1321.56640625,
      "completions/mean_terminated_length": 1052.0595703125,
      "completions/min_length": 314.0,
      "completions/min_terminated_length": 314.0,
      "entropy": 0.25689977407455444,
      "epoch": 2.828525641025641,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 857.88134765625,
      "learning_rate": 1e-06,
      "loss": 0.0615,
      "num_tokens": 1098881909.0,
      "reward": 0.3542296588420868,
      "reward_std": 0.029991699382662773,
      "rewards/progression_diversity/mean": -0.008187920786440372,
      "rewards/progression_diversity/std": 0.060944657772779465,
      "rewards/symbolic_reward_accuracy/mean": 0.234375,
      "rewards/symbolic_reward_accuracy/std": 0.42402184009552,
      "rewards/symbolic_reward_partial_score/mean": 0.712939441204071,
      "rewards/symbolic_reward_partial_score/std": 0.2108824998140335,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0325154066085815,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 13.213213920593262,
      "step": 1765
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2575305104255676,
      "epoch": 2.8301282051282053,
      "grad_norm": 0.018781762570142746,
      "learning_rate": 1e-06,
      "loss": 0.0015,
      "step": 1766
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.2592518925666809,
      "epoch": 2.831730769230769,
      "grad_norm": 0.010047774761915207,
      "learning_rate": 1e-06,
      "loss": 0.0387,
      "step": 1767
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2506992667913437,
      "epoch": 2.8333333333333335,
      "grad_norm": 0.012858088128268719,
      "learning_rate": 1e-06,
      "loss": 0.0435,
      "step": 1768
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 1814.0,
      "completions/mean_length": 1162.033203125,
      "completions/mean_terminated_length": 1072.3162841796875,
      "completions/min_length": 345.0,
      "completions/min_terminated_length": 345.0,
      "entropy": 0.25953665375709534,
      "epoch": 2.8349358974358974,
      "frac_reward_zero_std": 0.625,
      "grad_norm": 402.998046875,
      "learning_rate": 1e-06,
      "loss": 0.0247,
      "num_tokens": 1100398246.0,
      "reward": 0.3336543142795563,
      "reward_std": 0.019593775272369385,
      "rewards/progression_diversity/mean": -0.002245709067210555,
      "rewards/progression_diversity/std": 0.029325280338525772,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.7450683116912842,
      "rewards/symbolic_reward_partial_score/std": 0.1828693449497223,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0506747961044312,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 4.43061637878418,
      "step": 1769
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.2642872482538223,
      "epoch": 2.8365384615384617,
      "grad_norm": 0.010171863250434399,
      "learning_rate": 1e-06,
      "loss": -0.0029,
      "step": 1770
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.26166392862796783,
      "epoch": 2.8381410256410255,
      "grad_norm": 0.009490652941167355,
      "learning_rate": 1e-06,
      "loss": 0.0014,
      "step": 1771
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2598698139190674,
      "epoch": 2.83974358974359,
      "grad_norm": 0.010799145326018333,
      "learning_rate": 1e-06,
      "loss": 0.0263,
      "step": 1772
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2031.0,
      "completions/mean_length": 1231.958984375,
      "completions/mean_terminated_length": 1052.29052734375,
      "completions/min_length": 339.0,
      "completions/min_terminated_length": 339.0,
      "entropy": 0.2563736140727997,
      "epoch": 2.8413461538461537,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.015560018830001354,
      "learning_rate": 1e-06,
      "loss": -0.0021,
      "num_tokens": 1102002417.0,
      "reward": 0.27941030263900757,
      "reward_std": 0.020234711468219757,
      "rewards/progression_diversity/mean": -0.005260556936264038,
      "rewards/progression_diversity/std": 0.04788186773657799,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.6821939945220947,
      "rewards/symbolic_reward_partial_score/std": 0.20531851053237915,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0375680923461914,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 9.631587028503418,
      "step": 1773
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.23910512030124664,
      "epoch": 2.842948717948718,
      "grad_norm": 0.008363538421690464,
      "learning_rate": 1e-06,
      "loss": 0.0743,
      "step": 1774
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.24283453822135925,
      "epoch": 2.844551282051282,
      "grad_norm": 49193.4765625,
      "learning_rate": 1e-06,
      "loss": 1.9294,
      "step": 1775
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2511303126811981,
      "epoch": 2.8461538461538463,
      "grad_norm": 0.016176847741007805,
      "learning_rate": 1e-06,
      "loss": -0.005,
      "step": 1776
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 1908.0,
      "completions/mean_length": 1422.087890625,
      "completions/mean_terminated_length": 1124.0418701171875,
      "completions/min_length": 381.0,
      "completions/min_terminated_length": 381.0,
      "entropy": 0.2574124187231064,
      "epoch": 2.84775641025641,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 261.94879150390625,
      "learning_rate": 1e-06,
      "loss": -0.0031,
      "num_tokens": 1103595310.0,
      "reward": 0.3236244320869446,
      "reward_std": 0.023357607424259186,
      "rewards/progression_diversity/mean": -0.0086508272215724,
      "rewards/progression_diversity/std": 0.062096692621707916,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.7079427242279053,
      "rewards/symbolic_reward_partial_score/std": 0.21444584429264069,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.028921127319336,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 764.0,
      "sampling/sampling_logp_difference/mean": 14.721025466918945,
      "step": 1777
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2630213350057602,
      "epoch": 2.8493589743589745,
      "grad_norm": 0.015530991367995739,
      "learning_rate": 1e-06,
      "loss": 0.0129,
      "step": 1778
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.25482146441936493,
      "epoch": 2.8509615384615383,
      "grad_norm": 2.474681854248047,
      "learning_rate": 1e-06,
      "loss": 0.0207,
      "step": 1779
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.24518878757953644,
      "epoch": 2.8525641025641026,
      "grad_norm": 122.15824890136719,
      "learning_rate": 1e-06,
      "loss": 0.073,
      "step": 1780
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 1969.0,
      "completions/mean_length": 1393.974609375,
      "completions/mean_terminated_length": 1125.763427734375,
      "completions/min_length": 419.0,
      "completions/min_terminated_length": 419.0,
      "entropy": 0.24117282778024673,
      "epoch": 2.8541666666666665,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 894.6455688476562,
      "learning_rate": 1e-06,
      "loss": 0.0781,
      "num_tokens": 1105132289.0,
      "reward": 0.3265843987464905,
      "reward_std": 0.029642684385180473,
      "rewards/progression_diversity/mean": -0.007576026022434235,
      "rewards/progression_diversity/std": 0.05679089576005936,
      "rewards/symbolic_reward_accuracy/mean": 0.181640625,
      "rewards/symbolic_reward_accuracy/std": 0.38592514395713806,
      "rewards/symbolic_reward_partial_score/mean": 0.7262369394302368,
      "rewards/symbolic_reward_partial_score/std": 0.19816945493221283,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0353561639785767,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 11.549863815307617,
      "step": 1781
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.26401175558567047,
      "epoch": 2.855769230769231,
      "grad_norm": 0.0201826523989439,
      "learning_rate": 1e-06,
      "loss": -0.0045,
      "step": 1782
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2670938968658447,
      "epoch": 2.8573717948717947,
      "grad_norm": 0.00789049081504345,
      "learning_rate": 1e-06,
      "loss": 0.0029,
      "step": 1783
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2526419162750244,
      "epoch": 2.858974358974359,
      "grad_norm": 5.186265468597412,
      "learning_rate": 1e-06,
      "loss": 0.0221,
      "step": 1784
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2701.0,
      "completions/mean_length": 1483.630859375,
      "completions/mean_terminated_length": 1186.810791015625,
      "completions/min_length": 405.0,
      "completions/min_terminated_length": 405.0,
      "entropy": 0.2505309730768204,
      "epoch": 2.8605769230769234,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 22.291501998901367,
      "learning_rate": 1e-06,
      "loss": 0.0005,
      "num_tokens": 1106690644.0,
      "reward": 0.33161357045173645,
      "reward_std": 0.03396005928516388,
      "rewards/progression_diversity/mean": -0.008566973730921745,
      "rewards/progression_diversity/std": 0.061884794384241104,
      "rewards/symbolic_reward_accuracy/mean": 0.17578125,
      "rewards/symbolic_reward_accuracy/std": 0.3810062110424042,
      "rewards/symbolic_reward_partial_score/mean": 0.7560546398162842,
      "rewards/symbolic_reward_partial_score/std": 0.18065473437309265,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0316046476364136,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 12.537492752075195,
      "step": 1785
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.23839128017425537,
      "epoch": 2.8621794871794872,
      "grad_norm": 0.012131825089454651,
      "learning_rate": 1e-06,
      "loss": 0.1023,
      "step": 1786
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2580532729625702,
      "epoch": 2.863782051282051,
      "grad_norm": 0.007426050491631031,
      "learning_rate": 1e-06,
      "loss": 0.0244,
      "step": 1787
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.258169487118721,
      "epoch": 2.8653846153846154,
      "grad_norm": 0.017984015867114067,
      "learning_rate": 1e-06,
      "loss": -0.003,
      "step": 1788
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2401.0,
      "completions/mean_length": 1486.4765625,
      "completions/mean_terminated_length": 1189.713134765625,
      "completions/min_length": 377.0,
      "completions/min_terminated_length": 377.0,
      "entropy": 0.23781320452690125,
      "epoch": 2.8669871794871797,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 482.7618713378906,
      "learning_rate": 1e-06,
      "loss": 0.0416,
      "num_tokens": 1108318152.0,
      "reward": 0.32290422916412354,
      "reward_std": 0.04777061939239502,
      "rewards/progression_diversity/mean": -0.008405622094869614,
      "rewards/progression_diversity/std": 0.060591381043195724,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.765429675579071,
      "rewards/symbolic_reward_partial_score/std": 0.16949279606342316,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0300309658050537,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 12.117807388305664,
      "step": 1789
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.24075324833393097,
      "epoch": 2.8685897435897436,
      "grad_norm": 0.012247085571289062,
      "learning_rate": 1e-06,
      "loss": 0.0271,
      "step": 1790
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.23328307271003723,
      "epoch": 2.8701923076923075,
      "grad_norm": 0.010141503997147083,
      "learning_rate": 1e-06,
      "loss": 0.0239,
      "step": 1791
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.23753970116376877,
      "epoch": 2.871794871794872,
      "grad_norm": 0.014962738379836082,
      "learning_rate": 1e-06,
      "loss": 0.0154,
      "step": 1792
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3149.0,
      "completions/mean_length": 1426.01953125,
      "completions/mean_terminated_length": 1188.59130859375,
      "completions/min_length": 372.0,
      "completions/min_terminated_length": 372.0,
      "entropy": 0.23268520832061768,
      "epoch": 2.873397435897436,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 389.8380432128906,
      "learning_rate": 1e-06,
      "loss": 0.042,
      "num_tokens": 1109934338.0,
      "reward": 0.4080950617790222,
      "reward_std": 0.04739544540643692,
      "rewards/progression_diversity/mean": -0.005925622768700123,
      "rewards/progression_diversity/std": 0.04622631147503853,
      "rewards/symbolic_reward_accuracy/mean": 0.29296875,
      "rewards/symbolic_reward_accuracy/std": 0.455569326877594,
      "rewards/symbolic_reward_partial_score/mean": 0.7745768427848816,
      "rewards/symbolic_reward_partial_score/std": 0.19915655255317688,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0313217639923096,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 11.25175666809082,
      "step": 1793
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.23954074829816818,
      "epoch": 2.875,
      "grad_norm": 0.014046828262507915,
      "learning_rate": 1e-06,
      "loss": 0.0466,
      "step": 1794
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.24329328536987305,
      "epoch": 2.876602564102564,
      "grad_norm": 0.012684703804552555,
      "learning_rate": 1e-06,
      "loss": -0.0077,
      "step": 1795
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.23238325119018555,
      "epoch": 2.878205128205128,
      "grad_norm": 0.020547054708003998,
      "learning_rate": 1e-06,
      "loss": 0.0476,
      "step": 1796
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.025390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2012.0,
      "completions/mean_length": 1598.23828125,
      "completions/mean_terminated_length": 1213.0380859375,
      "completions/min_length": 426.0,
      "completions/min_terminated_length": 426.0,
      "entropy": 0.24311521649360657,
      "epoch": 2.8798076923076925,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 125.55624389648438,
      "learning_rate": 1e-06,
      "loss": 0.0103,
      "num_tokens": 1111646460.0,
      "reward": 0.3389037549495697,
      "reward_std": 0.025876596570014954,
      "rewards/progression_diversity/mean": -0.0070867519825696945,
      "rewards/progression_diversity/std": 0.045192137360572815,
      "rewards/symbolic_reward_accuracy/mean": 0.19921875,
      "rewards/symbolic_reward_accuracy/std": 0.39980348944664,
      "rewards/symbolic_reward_partial_score/mean": 0.7314778566360474,
      "rewards/symbolic_reward_partial_score/std": 0.1943960338830948,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0247465372085571,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 14.375904083251953,
      "step": 1797
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.23341339081525803,
      "epoch": 2.8814102564102564,
      "grad_norm": 0.018049802631139755,
      "learning_rate": 1e-06,
      "loss": 0.1091,
      "step": 1798
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2326470911502838,
      "epoch": 2.8830128205128203,
      "grad_norm": 0.00664905272424221,
      "learning_rate": 1e-06,
      "loss": 0.0569,
      "step": 1799
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.2430347353219986,
      "epoch": 2.8846153846153846,
      "grad_norm": 0.00851722713559866,
      "learning_rate": 1e-06,
      "loss": 0.0132,
      "step": 1800
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2379.0,
      "completions/mean_length": 1422.78515625,
      "completions/mean_terminated_length": 1155.08935546875,
      "completions/min_length": 302.0,
      "completions/min_terminated_length": 302.0,
      "entropy": 0.24775538593530655,
      "epoch": 2.886217948717949,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 916.2537841796875,
      "learning_rate": 1e-06,
      "loss": 0.0312,
      "num_tokens": 1113154542.0,
      "reward": 0.4589552879333496,
      "reward_std": 0.024608189240098,
      "rewards/progression_diversity/mean": -0.0053527336567640305,
      "rewards/progression_diversity/std": 0.040160540491342545,
      "rewards/symbolic_reward_accuracy/mean": 0.361328125,
      "rewards/symbolic_reward_accuracy/std": 0.48085519671440125,
      "rewards/symbolic_reward_partial_score/mean": 0.8086751699447632,
      "rewards/symbolic_reward_partial_score/std": 0.18015848100185394,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0337169170379639,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 10.59257698059082,
      "step": 1801
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.24687150865793228,
      "epoch": 2.8878205128205128,
      "grad_norm": 829.5130615234375,
      "learning_rate": 1e-06,
      "loss": 0.0824,
      "step": 1802
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.25254474580287933,
      "epoch": 2.8894230769230766,
      "grad_norm": 2157.5517578125,
      "learning_rate": 1e-06,
      "loss": 0.3136,
      "step": 1803
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.24701299518346786,
      "epoch": 2.891025641025641,
      "grad_norm": 0.011435293592512608,
      "learning_rate": 1e-06,
      "loss": 0.0445,
      "step": 1804
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2067.0,
      "completions/mean_length": 1403.126953125,
      "completions/mean_terminated_length": 1195.4713134765625,
      "completions/min_length": 437.0,
      "completions/min_terminated_length": 437.0,
      "entropy": 0.23765964061021805,
      "epoch": 2.8926282051282053,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 398.1752014160156,
      "learning_rate": 1e-06,
      "loss": 0.0265,
      "num_tokens": 1114716159.0,
      "reward": 0.28466612100601196,
      "reward_std": 0.02723667398095131,
      "rewards/progression_diversity/mean": -0.004093291237950325,
      "rewards/progression_diversity/std": 0.03477524220943451,
      "rewards/symbolic_reward_accuracy/mean": 0.123046875,
      "rewards/symbolic_reward_accuracy/std": 0.32881227135658264,
      "rewards/symbolic_reward_partial_score/mean": 0.7035807371139526,
      "rewards/symbolic_reward_partial_score/std": 0.18481120467185974,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0375463962554932,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 8.100292205810547,
      "step": 1805
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.24171607196331024,
      "epoch": 2.894230769230769,
      "grad_norm": 16.580183029174805,
      "learning_rate": 1e-06,
      "loss": -0.0029,
      "step": 1806
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2345980554819107,
      "epoch": 2.8958333333333335,
      "grad_norm": 0.02494910918176174,
      "learning_rate": 1e-06,
      "loss": 0.0728,
      "step": 1807
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2448808252811432,
      "epoch": 2.8974358974358974,
      "grad_norm": 0.008014152757823467,
      "learning_rate": 1e-06,
      "loss": 0.0123,
      "step": 1808
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3170.0,
      "completions/mean_length": 1560.744140625,
      "completions/mean_terminated_length": 1204.986083984375,
      "completions/min_length": 486.0,
      "completions/min_terminated_length": 486.0,
      "entropy": 0.2300274297595024,
      "epoch": 2.8990384615384617,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 1100.1624755859375,
      "learning_rate": 1e-06,
      "loss": 0.0962,
      "num_tokens": 1116373148.0,
      "reward": 0.3304210603237152,
      "reward_std": 0.0337834395468235,
      "rewards/progression_diversity/mean": -0.007212708704173565,
      "rewards/progression_diversity/std": 0.04686911776661873,
      "rewards/symbolic_reward_accuracy/mean": 0.197265625,
      "rewards/symbolic_reward_accuracy/std": 0.3983237147331238,
      "rewards/symbolic_reward_partial_score/mean": 0.7084146738052368,
      "rewards/symbolic_reward_partial_score/std": 0.2101738452911377,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0344150066375732,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 9.156167030334473,
      "step": 1809
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.2424936443567276,
      "epoch": 2.9006410256410255,
      "grad_norm": 0.019458921626210213,
      "learning_rate": 1e-06,
      "loss": -0.0052,
      "step": 1810
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.23142275214195251,
      "epoch": 2.90224358974359,
      "grad_norm": 0.010163613595068455,
      "learning_rate": 1e-06,
      "loss": 0.2442,
      "step": 1811
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.24503599107265472,
      "epoch": 2.9038461538461537,
      "grad_norm": 0.008980165235698223,
      "learning_rate": 1e-06,
      "loss": 0.0205,
      "step": 1812
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.041015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2119.0,
      "completions/mean_length": 1808.796875,
      "completions/mean_terminated_length": 1185.4176025390625,
      "completions/min_length": 324.0,
      "completions/min_terminated_length": 324.0,
      "entropy": 0.23406819999217987,
      "epoch": 2.905448717948718,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 73.40351867675781,
      "learning_rate": 1e-06,
      "loss": 0.0672,
      "num_tokens": 1118158052.0,
      "reward": 0.42395147681236267,
      "reward_std": 0.051604464650154114,
      "rewards/progression_diversity/mean": -0.012569092214107513,
      "rewards/progression_diversity/std": 0.06235016882419586,
      "rewards/symbolic_reward_accuracy/mean": 0.330078125,
      "rewards/symbolic_reward_accuracy/std": 0.47070086002349854,
      "rewards/symbolic_reward_partial_score/mean": 0.7547363042831421,
      "rewards/symbolic_reward_partial_score/std": 0.2372274249792099,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0196201801300049,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 15.569795608520508,
      "step": 1813
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.222855344414711,
      "epoch": 2.907051282051282,
      "grad_norm": 1161.0462646484375,
      "learning_rate": 1e-06,
      "loss": 63.2965,
      "step": 1814
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2360314056277275,
      "epoch": 2.9086538461538463,
      "grad_norm": 173744.71875,
      "learning_rate": 1e-06,
      "loss": 10.0443,
      "step": 1815
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2379593700170517,
      "epoch": 2.91025641025641,
      "grad_norm": 0.012492096051573753,
      "learning_rate": 1e-06,
      "loss": 0.0198,
      "step": 1816
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2021.0,
      "completions/mean_length": 1775.76171875,
      "completions/mean_terminated_length": 1212.7667236328125,
      "completions/min_length": 438.0,
      "completions/min_terminated_length": 438.0,
      "entropy": 0.22906029224395752,
      "epoch": 2.9118589743589745,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 874.5888671875,
      "learning_rate": 1e-06,
      "loss": 0.0862,
      "num_tokens": 1119985066.0,
      "reward": 0.31860923767089844,
      "reward_std": 0.04233834519982338,
      "rewards/progression_diversity/mean": -0.011634357273578644,
      "rewards/progression_diversity/std": 0.059071075171232224,
      "rewards/symbolic_reward_accuracy/mean": 0.173828125,
      "rewards/symbolic_reward_accuracy/std": 0.3793322443962097,
      "rewards/symbolic_reward_partial_score/mean": 0.7160645127296448,
      "rewards/symbolic_reward_partial_score/std": 0.22811061143875122,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0222280025482178,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 15.096211433410645,
      "step": 1817
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.23098096251487732,
      "epoch": 2.9134615384615383,
      "grad_norm": 0.011047611013054848,
      "learning_rate": 1e-06,
      "loss": 0.0751,
      "step": 1818
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.22944733500480652,
      "epoch": 2.9150641025641026,
      "grad_norm": 0.008758867159485817,
      "learning_rate": 1e-06,
      "loss": 0.0656,
      "step": 1819
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.23350029438734055,
      "epoch": 2.9166666666666665,
      "grad_norm": 0.009284427389502525,
      "learning_rate": 1e-06,
      "loss": 0.0732,
      "step": 1820
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2415.0,
      "completions/mean_length": 1809.248046875,
      "completions/mean_terminated_length": 1216.7784423828125,
      "completions/min_length": 396.0,
      "completions/min_terminated_length": 396.0,
      "entropy": 0.23025192320346832,
      "epoch": 2.918269230769231,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 926.9324951171875,
      "learning_rate": 1e-06,
      "loss": 0.0657,
      "num_tokens": 1121805241.0,
      "reward": 0.36059021949768066,
      "reward_std": 0.031355924904346466,
      "rewards/progression_diversity/mean": -0.0122678866609931,
      "rewards/progression_diversity/std": 0.062373507767915726,
      "rewards/symbolic_reward_accuracy/mean": 0.24609375,
      "rewards/symbolic_reward_accuracy/std": 0.4311550557613373,
      "rewards/symbolic_reward_partial_score/mean": 0.7114908695220947,
      "rewards/symbolic_reward_partial_score/std": 0.22253166139125824,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0174920558929443,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 764.0,
      "sampling/sampling_logp_difference/mean": 17.77764892578125,
      "step": 1821
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23264270275831223,
      "epoch": 2.9198717948717947,
      "grad_norm": 0.008918493054807186,
      "learning_rate": 1e-06,
      "loss": 0.0674,
      "step": 1822
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.22309153527021408,
      "epoch": 2.921474358974359,
      "grad_norm": 0.0191388800740242,
      "learning_rate": 1e-06,
      "loss": 0.1132,
      "step": 1823
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.2341051548719406,
      "epoch": 2.9230769230769234,
      "grad_norm": 0.009794514626264572,
      "learning_rate": 1e-06,
      "loss": 0.0327,
      "step": 1824
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.021484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2205.0,
      "completions/mean_length": 1469.75,
      "completions/mean_terminated_length": 1142.2913818359375,
      "completions/min_length": 323.0,
      "completions/min_terminated_length": 323.0,
      "entropy": 0.24822328984737396,
      "epoch": 2.9246794871794872,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.017211588099598885,
      "learning_rate": 1e-06,
      "loss": -0.0109,
      "num_tokens": 1123429833.0,
      "reward": 0.3736184239387512,
      "reward_std": 0.024019591510295868,
      "rewards/progression_diversity/mean": -0.007301281206309795,
      "rewards/progression_diversity/std": 0.050444815307855606,
      "rewards/symbolic_reward_accuracy/mean": 0.248046875,
      "rewards/symbolic_reward_accuracy/std": 0.4323015511035919,
      "rewards/symbolic_reward_partial_score/mean": 0.7514973878860474,
      "rewards/symbolic_reward_partial_score/std": 0.1985999196767807,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0336096286773682,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 764.0,
      "sampling/sampling_logp_difference/mean": 10.840036392211914,
      "step": 1825
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.23915088176727295,
      "epoch": 2.926282051282051,
      "grad_norm": 0.015742426738142967,
      "learning_rate": 1e-06,
      "loss": 0.0593,
      "step": 1826
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.23261761665344238,
      "epoch": 2.9278846153846154,
      "grad_norm": 0.004535183776170015,
      "learning_rate": 1e-06,
      "loss": 0.0502,
      "step": 1827
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.23625855147838593,
      "epoch": 2.9294871794871797,
      "grad_norm": 0.011525583453476429,
      "learning_rate": 1e-06,
      "loss": 0.0662,
      "step": 1828
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 1917.0,
      "completions/mean_length": 1803.416015625,
      "completions/mean_terminated_length": 1210.7093505859375,
      "completions/min_length": 412.0,
      "completions/min_terminated_length": 412.0,
      "entropy": 0.22059651464223862,
      "epoch": 2.9310897435897436,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 478.7748718261719,
      "learning_rate": 1e-06,
      "loss": 0.1023,
      "num_tokens": 1125201182.0,
      "reward": 0.32501205801963806,
      "reward_std": 0.03538961708545685,
      "rewards/progression_diversity/mean": -0.011980684474110603,
      "rewards/progression_diversity/std": 0.05950572341680527,
      "rewards/symbolic_reward_accuracy/mean": 0.166015625,
      "rewards/symbolic_reward_accuracy/std": 0.3724585771560669,
      "rewards/symbolic_reward_partial_score/mean": 0.752392590045929,
      "rewards/symbolic_reward_partial_score/std": 0.18106764554977417,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.018181324005127,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 764.0,
      "sampling/sampling_logp_difference/mean": 17.928653717041016,
      "step": 1829
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.22678129374980927,
      "epoch": 2.9326923076923075,
      "grad_norm": 0.018833013251423836,
      "learning_rate": 1e-06,
      "loss": 0.0311,
      "step": 1830
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.22647760808467865,
      "epoch": 2.934294871794872,
      "grad_norm": 17.264345169067383,
      "learning_rate": 1e-06,
      "loss": 0.0067,
      "step": 1831
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.21786439418792725,
      "epoch": 2.935897435897436,
      "grad_norm": 0.15827719867229462,
      "learning_rate": 1e-06,
      "loss": 0.0264,
      "step": 1832
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.076171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2052.0,
      "completions/mean_length": 2331.796875,
      "completions/mean_terminated_length": 1173.1585693359375,
      "completions/min_length": 314.0,
      "completions/min_terminated_length": 314.0,
      "entropy": 0.21573705226182938,
      "epoch": 2.9375,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 590.8438110351562,
      "learning_rate": 1e-06,
      "loss": 0.0605,
      "num_tokens": 1127233478.0,
      "reward": 0.3646172285079956,
      "reward_std": 0.0675165057182312,
      "rewards/progression_diversity/mean": -0.02313985861837864,
      "rewards/progression_diversity/std": 0.08131039142608643,
      "rewards/symbolic_reward_accuracy/mean": 0.2265625,
      "rewards/symbolic_reward_accuracy/std": 0.4190165400505066,
      "rewards/symbolic_reward_partial_score/mean": 0.7630370855331421,
      "rewards/symbolic_reward_partial_score/std": 0.2085689753293991,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9808230996131897,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 764.0,
      "sampling/sampling_logp_difference/mean": 36.245513916015625,
      "step": 1833
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2271483689546585,
      "epoch": 2.939102564102564,
      "grad_norm": 0.014887535013258457,
      "learning_rate": 1e-06,
      "loss": 0.0405,
      "step": 1834
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.21168261021375656,
      "epoch": 2.940705128205128,
      "grad_norm": 0.017205113545060158,
      "learning_rate": 1e-06,
      "loss": 0.1176,
      "step": 1835
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.20247046649456024,
      "epoch": 2.9423076923076925,
      "grad_norm": 0.014779583550989628,
      "learning_rate": 1e-06,
      "loss": 0.1315,
      "step": 1836
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.07421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2314.0,
      "completions/mean_length": 2300.94921875,
      "completions/mean_terminated_length": 1171.92822265625,
      "completions/min_length": 449.0,
      "completions/min_terminated_length": 449.0,
      "entropy": 0.20949921756982803,
      "epoch": 2.9439102564102564,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 870.5208740234375,
      "learning_rate": 1e-06,
      "loss": 0.0751,
      "num_tokens": 1129310636.0,
      "reward": 0.45326170325279236,
      "reward_std": 0.044423650950193405,
      "rewards/progression_diversity/mean": -0.02392907440662384,
      "rewards/progression_diversity/std": 0.08498639613389969,
      "rewards/symbolic_reward_accuracy/mean": 0.365234375,
      "rewards/symbolic_reward_accuracy/std": 0.4819667339324951,
      "rewards/symbolic_reward_partial_score/mean": 0.781201183795929,
      "rewards/symbolic_reward_partial_score/std": 0.22568704187870026,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9865758419036865,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 764.0,
      "sampling/sampling_logp_difference/mean": 33.854530334472656,
      "step": 1837
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.21899111568927765,
      "epoch": 2.9455128205128203,
      "grad_norm": 0.012719093821942806,
      "learning_rate": 1e-06,
      "loss": 0.0941,
      "step": 1838
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.22411292791366577,
      "epoch": 2.9471153846153846,
      "grad_norm": 0.021860910579562187,
      "learning_rate": 1e-06,
      "loss": 0.0825,
      "step": 1839
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2084907665848732,
      "epoch": 2.948717948717949,
      "grad_norm": 0.005877522751688957,
      "learning_rate": 1e-06,
      "loss": 0.0637,
      "step": 1840
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2297.0,
      "completions/mean_length": 2534.21484375,
      "completions/mean_terminated_length": 1232.098388671875,
      "completions/min_length": 392.0,
      "completions/min_terminated_length": 392.0,
      "entropy": 0.20938124507665634,
      "epoch": 2.9503205128205128,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 560.36669921875,
      "learning_rate": 1e-06,
      "loss": 0.12,
      "num_tokens": 1131555370.0,
      "reward": 0.32735151052474976,
      "reward_std": 0.05115168169140816,
      "rewards/progression_diversity/mean": -0.03047451376914978,
      "rewards/progression_diversity/std": 0.09939130395650864,
      "rewards/symbolic_reward_accuracy/mean": 0.19140625,
      "rewards/symbolic_reward_accuracy/std": 0.3937928080558777,
      "rewards/symbolic_reward_partial_score/mean": 0.7093750238418579,
      "rewards/symbolic_reward_partial_score/std": 0.2335452288389206,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9635010361671448,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 45.37931442260742,
      "step": 1841
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2120969444513321,
      "epoch": 2.9519230769230766,
      "grad_norm": 381.89239501953125,
      "learning_rate": 1e-06,
      "loss": 821.5478,
      "step": 1842
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2079833373427391,
      "epoch": 2.953525641025641,
      "grad_norm": 0.010464330203831196,
      "learning_rate": 1e-06,
      "loss": 32.2227,
      "step": 1843
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.21436134725809097,
      "epoch": 2.9551282051282053,
      "grad_norm": 0.02022792212665081,
      "learning_rate": 1e-06,
      "loss": 0.1652,
      "step": 1844
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.14453125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2207.0,
      "completions/mean_length": 3449.130859375,
      "completions/mean_terminated_length": 1263.78759765625,
      "completions/min_length": 482.0,
      "completions/min_terminated_length": 482.0,
      "entropy": 0.21006950736045837,
      "epoch": 2.956730769230769,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 1419.5054931640625,
      "learning_rate": 1e-06,
      "loss": 0.1083,
      "num_tokens": 1134191421.0,
      "reward": 0.2933524250984192,
      "reward_std": 0.04929710179567337,
      "rewards/progression_diversity/mean": -0.05147593468427658,
      "rewards/progression_diversity/std": 0.12628963589668274,
      "rewards/symbolic_reward_accuracy/mean": 0.150390625,
      "rewards/symbolic_reward_accuracy/std": 0.35780346393585205,
      "rewards/symbolic_reward_partial_score/mean": 0.6800781488418579,
      "rewards/symbolic_reward_partial_score/std": 0.23385514318943024,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9397692680358887,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 57.66311264038086,
      "step": 1845
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.1966499239206314,
      "epoch": 2.9583333333333335,
      "grad_norm": 16499.4140625,
      "learning_rate": 1e-06,
      "loss": 5.3166,
      "step": 1846
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.19310440123081207,
      "epoch": 2.9599358974358974,
      "grad_norm": 4365.603515625,
      "learning_rate": 1e-06,
      "loss": 0.5053,
      "step": 1847
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.1920364871621132,
      "epoch": 2.9615384615384617,
      "grad_norm": 0.009575553238391876,
      "learning_rate": 1e-06,
      "loss": 0.1589,
      "step": 1848
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.12890625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2035.0,
      "completions/mean_length": 3208.90625,
      "completions/mean_terminated_length": 1259.228759765625,
      "completions/min_length": 470.0,
      "completions/min_terminated_length": 470.0,
      "entropy": 0.2075439766049385,
      "epoch": 2.9631410256410255,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 574.5317993164062,
      "learning_rate": 1e-06,
      "loss": 0.0662,
      "num_tokens": 1136791005.0,
      "reward": 0.28958579897880554,
      "reward_std": 0.05835293233394623,
      "rewards/progression_diversity/mean": -0.04337439686059952,
      "rewards/progression_diversity/std": 0.11332310736179352,
      "rewards/symbolic_reward_accuracy/mean": 0.142578125,
      "rewards/symbolic_reward_accuracy/std": 0.3499840497970581,
      "rewards/symbolic_reward_partial_score/mean": 0.6835286617279053,
      "rewards/symbolic_reward_partial_score/std": 0.22093914449214935,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9411171674728394,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 56.884559631347656,
      "step": 1849
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.359375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.19637862592935562,
      "epoch": 2.96474358974359,
      "grad_norm": 0.008334570564329624,
      "learning_rate": 1e-06,
      "loss": 0.1807,
      "step": 1850
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.390625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.20494883507490158,
      "epoch": 2.9663461538461537,
      "grad_norm": 0.007710518781095743,
      "learning_rate": 1e-06,
      "loss": 0.1235,
      "step": 1851
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.18237695842981339,
      "epoch": 2.967948717948718,
      "grad_norm": 0.011762432754039764,
      "learning_rate": 1e-06,
      "loss": 0.1953,
      "step": 1852
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.09765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2194.0,
      "completions/mean_length": 2733.078125,
      "completions/mean_terminated_length": 1255.70556640625,
      "completions/min_length": 351.0,
      "completions/min_terminated_length": 351.0,
      "entropy": 0.2183959111571312,
      "epoch": 2.969551282051282,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 1045.5770263671875,
      "learning_rate": 1e-06,
      "loss": 0.0877,
      "num_tokens": 1139114341.0,
      "reward": 0.33328425884246826,
      "reward_std": 0.06576712429523468,
      "rewards/progression_diversity/mean": -0.030461102724075317,
      "rewards/progression_diversity/std": 0.09234879910945892,
      "rewards/symbolic_reward_accuracy/mean": 0.203125,
      "rewards/symbolic_reward_accuracy/std": 0.4027182459831238,
      "rewards/symbolic_reward_partial_score/mean": 0.7057129144668579,
      "rewards/symbolic_reward_partial_score/std": 0.23553626239299774,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9779849052429199,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 39.03363800048828,
      "step": 1853
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.2225719690322876,
      "epoch": 2.9711538461538463,
      "grad_norm": 0.00991111621260643,
      "learning_rate": 1e-06,
      "loss": 0.1348,
      "step": 1854
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.19753701984882355,
      "epoch": 2.97275641025641,
      "grad_norm": 0.006589308846741915,
      "learning_rate": 1e-06,
      "loss": 0.145,
      "step": 1855
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2158377543091774,
      "epoch": 2.9743589743589745,
      "grad_norm": 0.009189927950501442,
      "learning_rate": 1e-06,
      "loss": 0.1266,
      "step": 1856
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2904.0,
      "completions/mean_length": 2554.814453125,
      "completions/mean_terminated_length": 1254.6346435546875,
      "completions/min_length": 385.0,
      "completions/min_terminated_length": 385.0,
      "entropy": 0.20936352014541626,
      "epoch": 2.9759615384615383,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 463.4955139160156,
      "learning_rate": 1e-06,
      "loss": 0.0787,
      "num_tokens": 1141373350.0,
      "reward": 0.3864889144897461,
      "reward_std": 0.0557912215590477,
      "rewards/progression_diversity/mean": -0.02347256988286972,
      "rewards/progression_diversity/std": 0.07694897800683975,
      "rewards/symbolic_reward_accuracy/mean": 0.2734375,
      "rewards/symbolic_reward_accuracy/std": 0.4461594223976135,
      "rewards/symbolic_reward_partial_score/mean": 0.7422037720680237,
      "rewards/symbolic_reward_partial_score/std": 0.24117407202720642,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9877347946166992,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 33.34986114501953,
      "step": 1857
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.22580835223197937,
      "epoch": 2.9775641025641026,
      "grad_norm": 0.01121189258992672,
      "learning_rate": 1e-06,
      "loss": 0.0885,
      "step": 1858
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2194606065750122,
      "epoch": 2.9791666666666665,
      "grad_norm": 0.018390899524092674,
      "learning_rate": 1e-06,
      "loss": 0.1005,
      "step": 1859
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.20080887526273727,
      "epoch": 2.980769230769231,
      "grad_norm": 0.009253687225282192,
      "learning_rate": 1e-06,
      "loss": 0.1674,
      "step": 1860
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2504.0,
      "completions/mean_length": 2508.169921875,
      "completions/mean_terminated_length": 1268.2021484375,
      "completions/min_length": 455.0,
      "completions/min_terminated_length": 455.0,
      "entropy": 0.23013577610254288,
      "epoch": 2.9823717948717947,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 397.2406311035156,
      "learning_rate": 1e-06,
      "loss": 0.0351,
      "num_tokens": 1143492301.0,
      "reward": 0.39299777150154114,
      "reward_std": 0.05584446340799332,
      "rewards/progression_diversity/mean": -0.023467358201742172,
      "rewards/progression_diversity/std": 0.07824753224849701,
      "rewards/symbolic_reward_accuracy/mean": 0.291015625,
      "rewards/symbolic_reward_accuracy/std": 0.45467492938041687,
      "rewards/symbolic_reward_partial_score/mean": 0.7287434935569763,
      "rewards/symbolic_reward_partial_score/std": 0.24250133335590363,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9919546842575073,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 31.701099395751953,
      "step": 1861
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2217669039964676,
      "epoch": 2.983974358974359,
      "grad_norm": 53.626285552978516,
      "learning_rate": 1e-06,
      "loss": 0.0267,
      "step": 1862
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.21721980720758438,
      "epoch": 2.9855769230769234,
      "grad_norm": 0.032945115119218826,
      "learning_rate": 1e-06,
      "loss": 0.172,
      "step": 1863
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.22124967724084854,
      "epoch": 2.9871794871794872,
      "grad_norm": 0.01755567640066147,
      "learning_rate": 1e-06,
      "loss": 0.093,
      "step": 1864
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.064453125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2574.0,
      "completions/mean_length": 2252.9453125,
      "completions/mean_terminated_length": 1279.4071044921875,
      "completions/min_length": 314.0,
      "completions/min_terminated_length": 314.0,
      "entropy": 0.2342042252421379,
      "epoch": 2.988782051282051,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 481.52252197265625,
      "learning_rate": 1e-06,
      "loss": 0.1885,
      "num_tokens": 1145451937.0,
      "reward": 0.3523566424846649,
      "reward_std": 0.04059898108243942,
      "rewards/progression_diversity/mean": -0.01775522157549858,
      "rewards/progression_diversity/std": 0.06716568768024445,
      "rewards/symbolic_reward_accuracy/mean": 0.212890625,
      "rewards/symbolic_reward_accuracy/std": 0.409751296043396,
      "rewards/symbolic_reward_partial_score/mean": 0.7493326663970947,
      "rewards/symbolic_reward_partial_score/std": 0.22399716079235077,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0028358697891235,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 27.260162353515625,
      "step": 1865
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.24164949357509613,
      "epoch": 2.9903846153846154,
      "grad_norm": 14.552428245544434,
      "learning_rate": 1e-06,
      "loss": 0.0096,
      "step": 1866
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.23062748461961746,
      "epoch": 2.9919871794871797,
      "grad_norm": 0.030370866879820824,
      "learning_rate": 1e-06,
      "loss": 0.1004,
      "step": 1867
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.24224235862493515,
      "epoch": 2.9935897435897436,
      "grad_norm": 0.011348499916493893,
      "learning_rate": 1e-06,
      "loss": 0.1191,
      "step": 1868
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.091796875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2234.0,
      "completions/mean_length": 2649.326171875,
      "completions/mean_terminated_length": 1261.09033203125,
      "completions/min_length": 413.0,
      "completions/min_terminated_length": 413.0,
      "entropy": 0.24135691672563553,
      "epoch": 2.9951923076923075,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 1024.800537109375,
      "learning_rate": 1e-06,
      "loss": 0.0707,
      "num_tokens": 1147683176.0,
      "reward": 0.3426709771156311,
      "reward_std": 0.04791571944952011,
      "rewards/progression_diversity/mean": -0.025381293147802353,
      "rewards/progression_diversity/std": 0.07976257801055908,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.7212076187133789,
      "rewards/symbolic_reward_partial_score/std": 0.21549275517463684,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9791022539138794,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 38.175758361816406,
      "step": 1869
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3671875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2204393520951271,
      "epoch": 2.996794871794872,
      "grad_norm": 1905.4779052734375,
      "learning_rate": 1e-06,
      "loss": 0.2076,
      "step": 1870
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.21573010832071304,
      "epoch": 2.998397435897436,
      "grad_norm": 12149.6005859375,
      "learning_rate": 1e-06,
      "loss": 1.1185,
      "step": 1871
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3671875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.2214333862066269,
      "epoch": 3.0,
      "grad_norm": 0.018644830211997032,
      "learning_rate": 1e-06,
      "loss": 0.1475,
      "step": 1872
    },
    {
      "epoch": 3.0,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.080078125,
      "eval_completions/max_length": 16384.0,
      "eval_completions/max_terminated_length": 2229.46875,
      "eval_completions/mean_length": 2536.551025390625,
      "eval_completions/mean_terminated_length": 1331.3045349121094,
      "eval_completions/min_length": 594.8125,
      "eval_completions/min_terminated_length": 594.8125,
      "eval_entropy": 0.20870172325521708,
      "eval_frac_reward_zero_std": 0.1796875,
      "eval_loss": 0.04187625274062157,
      "eval_num_tokens": 1147683176.0,
      "eval_reward": 0.2534480579197407,
      "eval_reward_std": 0.0388854734483175,
      "eval_rewards/progression_diversity/mean": -0.021589503419818357,
      "eval_rewards/progression_diversity/std": 0.07232485536951572,
      "eval_rewards/symbolic_reward_accuracy/mean": 0.09130859375,
      "eval_rewards/symbolic_reward_accuracy/std": 0.2087636678479612,
      "eval_rewards/symbolic_reward_partial_score/mean": 0.6634989399462938,
      "eval_rewards/symbolic_reward_partial_score/std": 0.1970880120061338,
      "eval_rewards/tag_count_reward/mean": -0.001708984375,
      "eval_rewards/tag_count_reward/std": 0.015794883016496897,
      "eval_runtime": 4301.5149,
      "eval_samples_per_second": 0.058,
      "eval_sampling/importance_sampling_ratio/max": 2.0,
      "eval_sampling/importance_sampling_ratio/mean": 1.021749384701252,
      "eval_sampling/importance_sampling_ratio/min": 0.0,
      "eval_sampling/sampling_logp_difference/max": 768.0,
      "eval_sampling/sampling_logp_difference/mean": 19.073409140110016,
      "eval_steps_per_second": 0.0,
      "step": 1872
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.072265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2393.0,
      "completions/mean_length": 2402.98046875,
      "completions/mean_terminated_length": 1313.9326171875,
      "completions/min_length": 359.0,
      "completions/min_terminated_length": 359.0,
      "entropy": 0.23216699063777924,
      "epoch": 3.0016025641025643,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 1125.79931640625,
      "learning_rate": 1e-06,
      "loss": 0.0865,
      "num_tokens": 1149811486.0,
      "reward": 0.3975224494934082,
      "reward_std": 0.04312722384929657,
      "rewards/progression_diversity/mean": -0.02021792158484459,
      "rewards/progression_diversity/std": 0.07212448120117188,
      "rewards/symbolic_reward_accuracy/mean": 0.3046875,
      "rewards/symbolic_reward_accuracy/std": 0.4607250988483429,
      "rewards/symbolic_reward_partial_score/mean": 0.7163736820220947,
      "rewards/symbolic_reward_partial_score/std": 0.23637627065181732,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9978500604629517,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 30.130069732666016,
      "step": 1873
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.23934829235076904,
      "epoch": 3.003205128205128,
      "grad_norm": 0.007034912705421448,
      "learning_rate": 1e-06,
      "loss": 0.1083,
      "step": 1874
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.22745058685541153,
      "epoch": 3.0048076923076925,
      "grad_norm": 0.00796093326061964,
      "learning_rate": 1e-06,
      "loss": 0.0727,
      "step": 1875
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.2292134240269661,
      "epoch": 3.0064102564102564,
      "grad_norm": 0.019052157178521156,
      "learning_rate": 1e-06,
      "loss": 0.1367,
      "step": 1876
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2208.0,
      "completions/mean_length": 1881.35546875,
      "completions/mean_terminated_length": 1322.429931640625,
      "completions/min_length": 434.0,
      "completions/min_terminated_length": 434.0,
      "entropy": 0.2549893334507942,
      "epoch": 3.0080128205128207,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 0.024593627080321312,
      "learning_rate": 1e-06,
      "loss": 0.0391,
      "num_tokens": 1151633332.0,
      "reward": 0.2892531454563141,
      "reward_std": 0.039773762226104736,
      "rewards/progression_diversity/mean": -0.010232724249362946,
      "rewards/progression_diversity/std": 0.055212657898664474,
      "rewards/symbolic_reward_accuracy/mean": 0.119140625,
      "rewards/symbolic_reward_accuracy/std": 0.32427072525024414,
      "rewards/symbolic_reward_partial_score/mean": 0.7275390625,
      "rewards/symbolic_reward_partial_score/std": 0.19641563296318054,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0221221446990967,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 19.10765838623047,
      "step": 1877
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.24741330742835999,
      "epoch": 3.0096153846153846,
      "grad_norm": 0.01947159692645073,
      "learning_rate": 1e-06,
      "loss": 0.1312,
      "step": 1878
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.24799786508083344,
      "epoch": 3.011217948717949,
      "grad_norm": 0.02338356338441372,
      "learning_rate": 1e-06,
      "loss": 0.0404,
      "step": 1879
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2448197305202484,
      "epoch": 3.0128205128205128,
      "grad_norm": 0.014209832064807415,
      "learning_rate": 1e-06,
      "loss": 0.0894,
      "step": 1880
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.033203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3167.0,
      "completions/mean_length": 1858.625,
      "completions/mean_terminated_length": 1359.7738037109375,
      "completions/min_length": 445.0,
      "completions/min_terminated_length": 445.0,
      "entropy": 0.25408633053302765,
      "epoch": 3.014423076923077,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 454.2471618652344,
      "learning_rate": 1e-06,
      "loss": 0.0447,
      "num_tokens": 1153465092.0,
      "reward": 0.3250887095928192,
      "reward_std": 0.015932850539684296,
      "rewards/progression_diversity/mean": -0.011149164289236069,
      "rewards/progression_diversity/std": 0.05865437537431717,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.7096517086029053,
      "rewards/symbolic_reward_partial_score/std": 0.19842353463172913,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.028539776802063,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 17.165746688842773,
      "step": 1881
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2505646124482155,
      "epoch": 3.016025641025641,
      "grad_norm": 0.6798326373100281,
      "learning_rate": 1e-06,
      "loss": 0.0431,
      "step": 1882
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.24388013035058975,
      "epoch": 3.0176282051282053,
      "grad_norm": 1.9514085054397583,
      "learning_rate": 1e-06,
      "loss": 0.1241,
      "step": 1883
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.26686759293079376,
      "epoch": 3.019230769230769,
      "grad_norm": 0.012397652491927147,
      "learning_rate": 1e-06,
      "loss": -0.0004,
      "step": 1884
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.025390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2108.0,
      "completions/mean_length": 1794.703125,
      "completions/mean_terminated_length": 1414.6212158203125,
      "completions/min_length": 480.0,
      "completions/min_terminated_length": 480.0,
      "entropy": 0.2557114362716675,
      "epoch": 3.0208333333333335,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 544.6851196289062,
      "learning_rate": 1e-06,
      "loss": 0.0255,
      "num_tokens": 1155145164.0,
      "reward": 0.37460383772850037,
      "reward_std": 0.016295883804559708,
      "rewards/progression_diversity/mean": -0.008367680944502354,
      "rewards/progression_diversity/std": 0.05215869098901749,
      "rewards/symbolic_reward_accuracy/mean": 0.25,
      "rewards/symbolic_reward_accuracy/std": 0.43343618512153625,
      "rewards/symbolic_reward_partial_score/mean": 0.7502604126930237,
      "rewards/symbolic_reward_partial_score/std": 0.1986854523420334,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.030568242073059,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 16.76722526550293,
      "step": 1885
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.26103954017162323,
      "epoch": 3.0224358974358974,
      "grad_norm": 91.75881958007812,
      "learning_rate": 1e-06,
      "loss": 0.2128,
      "step": 1886
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2607679069042206,
      "epoch": 3.0240384615384617,
      "grad_norm": 0.012660062871873379,
      "learning_rate": 1e-06,
      "loss": 0.0615,
      "step": 1887
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.26138727366924286,
      "epoch": 3.0256410256410255,
      "grad_norm": 0.01078207790851593,
      "learning_rate": 1e-06,
      "loss": 0.0361,
      "step": 1888
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2198.0,
      "completions/mean_length": 1951.89453125,
      "completions/mean_terminated_length": 1395.6876220703125,
      "completions/min_length": 593.0,
      "completions/min_terminated_length": 593.0,
      "entropy": 0.2467678338289261,
      "epoch": 3.02724358974359,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 572.4014282226562,
      "learning_rate": 1e-06,
      "loss": 0.0566,
      "num_tokens": 1157118710.0,
      "reward": 0.226592555642128,
      "reward_std": 0.021917399019002914,
      "rewards/progression_diversity/mean": -0.01262015849351883,
      "rewards/progression_diversity/std": 0.06260491162538528,
      "rewards/symbolic_reward_accuracy/mean": 0.05859375,
      "rewards/symbolic_reward_accuracy/std": 0.23509246110916138,
      "rewards/symbolic_reward_partial_score/mean": 0.6385416984558105,
      "rewards/symbolic_reward_partial_score/std": 0.20352140069007874,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0226848125457764,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 19.33092498779297,
      "step": 1889
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.23328816145658493,
      "epoch": 3.0288461538461537,
      "grad_norm": 20633.263671875,
      "learning_rate": 1e-06,
      "loss": 4.1374,
      "step": 1890
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2414126992225647,
      "epoch": 3.030448717948718,
      "grad_norm": 6596.80908203125,
      "learning_rate": 1e-06,
      "loss": 0.5508,
      "step": 1891
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2591097354888916,
      "epoch": 3.032051282051282,
      "grad_norm": 0.01172653678804636,
      "learning_rate": 1e-06,
      "loss": 0.0182,
      "step": 1892
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2715.0,
      "completions/mean_length": 1600.69921875,
      "completions/mean_terminated_length": 1366.043701171875,
      "completions/min_length": 407.0,
      "completions/min_terminated_length": 407.0,
      "entropy": 0.26757121086120605,
      "epoch": 3.0336538461538463,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.026514563709497452,
      "learning_rate": 1e-06,
      "loss": 0.0088,
      "num_tokens": 1158840796.0,
      "reward": 0.22522126138210297,
      "reward_std": 0.04094214737415314,
      "rewards/progression_diversity/mean": -0.005220310762524605,
      "rewards/progression_diversity/std": 0.04069770127534866,
      "rewards/symbolic_reward_accuracy/mean": 0.0390625,
      "rewards/symbolic_reward_accuracy/std": 0.1939331740140915,
      "rewards/symbolic_reward_partial_score/mean": 0.6740885972976685,
      "rewards/symbolic_reward_partial_score/std": 0.1974918395280838,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.041705846786499,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 10.610610008239746,
      "step": 1893
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.25377361476421356,
      "epoch": 3.03525641025641,
      "grad_norm": 10585.2099609375,
      "learning_rate": 1e-06,
      "loss": 0.5272,
      "step": 1894
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.26954758167266846,
      "epoch": 3.0368589743589745,
      "grad_norm": 0.022427378222346306,
      "learning_rate": 1e-06,
      "loss": -0.0047,
      "step": 1895
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2479778230190277,
      "epoch": 3.0384615384615383,
      "grad_norm": 0.013517378829419613,
      "learning_rate": 1e-06,
      "loss": 0.2005,
      "step": 1896
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2292.0,
      "completions/mean_length": 1637.05859375,
      "completions/mean_terminated_length": 1373.19677734375,
      "completions/min_length": 425.0,
      "completions/min_terminated_length": 425.0,
      "entropy": 0.2555150091648102,
      "epoch": 3.0400641025641026,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 30.618288040161133,
      "learning_rate": 1e-06,
      "loss": 0.0226,
      "num_tokens": 1160526426.0,
      "reward": 0.4024108052253723,
      "reward_std": 0.02461879327893257,
      "rewards/progression_diversity/mean": -0.005499871913343668,
      "rewards/progression_diversity/std": 0.04200981557369232,
      "rewards/symbolic_reward_accuracy/mean": 0.275390625,
      "rewards/symbolic_reward_accuracy/std": 0.44714778661727905,
      "rewards/symbolic_reward_partial_score/mean": 0.790771484375,
      "rewards/symbolic_reward_partial_score/std": 0.18546874821186066,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0372226238250732,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 12.745841979980469,
      "step": 1897
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.2539917603135109,
      "epoch": 3.0416666666666665,
      "grad_norm": 44032.421875,
      "learning_rate": 1e-06,
      "loss": 1.0033,
      "step": 1898
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.25875651836395264,
      "epoch": 3.043269230769231,
      "grad_norm": 0.02200913429260254,
      "learning_rate": 1e-06,
      "loss": 0.0217,
      "step": 1899
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.25009360164403915,
      "epoch": 3.0448717948717947,
      "grad_norm": 0.015072772279381752,
      "learning_rate": 1e-06,
      "loss": 0.0296,
      "step": 1900
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.033203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2233.0,
      "completions/mean_length": 1893.123046875,
      "completions/mean_terminated_length": 1395.4566650390625,
      "completions/min_length": 557.0,
      "completions/min_terminated_length": 557.0,
      "entropy": 0.25338225811719894,
      "epoch": 3.046474358974359,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.014189253561198711,
      "learning_rate": 1e-06,
      "loss": 0.0664,
      "num_tokens": 1162334969.0,
      "reward": 0.31671467423439026,
      "reward_std": 0.037882737815380096,
      "rewards/progression_diversity/mean": -0.010173956863582134,
      "rewards/progression_diversity/std": 0.055442675948143005,
      "rewards/symbolic_reward_accuracy/mean": 0.173828125,
      "rewards/symbolic_reward_accuracy/std": 0.3793322443962097,
      "rewards/symbolic_reward_partial_score/mean": 0.7090494632720947,
      "rewards/symbolic_reward_partial_score/std": 0.20027847588062286,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.020569920539856,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 20.987850189208984,
      "step": 1901
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.25274112075567245,
      "epoch": 3.048076923076923,
      "grad_norm": 0.008120290003716946,
      "learning_rate": 1e-06,
      "loss": 0.0418,
      "step": 1902
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.24684013426303864,
      "epoch": 3.0496794871794872,
      "grad_norm": 0.010809239000082016,
      "learning_rate": 1e-06,
      "loss": 0.0503,
      "step": 1903
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.25020767748355865,
      "epoch": 3.051282051282051,
      "grad_norm": 0.0168289951980114,
      "learning_rate": 1e-06,
      "loss": 0.0331,
      "step": 1904
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2925.0,
      "completions/mean_length": 1549.30078125,
      "completions/mean_terminated_length": 1343.67138671875,
      "completions/min_length": 443.0,
      "completions/min_terminated_length": 443.0,
      "entropy": 0.2655075490474701,
      "epoch": 3.0528846153846154,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.02596384286880493,
      "learning_rate": 1e-06,
      "loss": 0.0209,
      "num_tokens": 1164024995.0,
      "reward": 0.3695905804634094,
      "reward_std": 0.033132217824459076,
      "rewards/progression_diversity/mean": -0.004322149325162172,
      "rewards/progression_diversity/std": 0.03666413947939873,
      "rewards/symbolic_reward_accuracy/mean": 0.244140625,
      "rewards/symbolic_reward_accuracy/std": 0.42999663949012756,
      "rewards/symbolic_reward_partial_score/mean": 0.7438313961029053,
      "rewards/symbolic_reward_partial_score/std": 0.20166073739528656,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0442430973052979,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 9.987215995788574,
      "step": 1905
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.25963981449604034,
      "epoch": 3.0544871794871793,
      "grad_norm": 140.0711669921875,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "step": 1906
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.26139749586582184,
      "epoch": 3.0560897435897436,
      "grad_norm": 0.00938948430120945,
      "learning_rate": 1e-06,
      "loss": 0.0255,
      "step": 1907
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2722902148962021,
      "epoch": 3.0576923076923075,
      "grad_norm": 0.011267850175499916,
      "learning_rate": 1e-06,
      "loss": 0.0042,
      "step": 1908
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2150.0,
      "completions/mean_length": 1744.671875,
      "completions/mean_terminated_length": 1333.1243896484375,
      "completions/min_length": 520.0,
      "completions/min_terminated_length": 520.0,
      "entropy": 0.2618495374917984,
      "epoch": 3.059294871794872,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 652.4515380859375,
      "learning_rate": 1e-06,
      "loss": 0.0214,
      "num_tokens": 1165798987.0,
      "reward": 0.4160562753677368,
      "reward_std": 0.061300039291381836,
      "rewards/progression_diversity/mean": -0.009119954891502857,
      "rewards/progression_diversity/std": 0.05495789274573326,
      "rewards/symbolic_reward_accuracy/mean": 0.314453125,
      "rewards/symbolic_reward_accuracy/std": 0.4647517800331116,
      "rewards/symbolic_reward_partial_score/mean": 0.7589030265808105,
      "rewards/symbolic_reward_partial_score/std": 0.2069811373949051,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0315742492675781,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 16.712371826171875,
      "step": 1909
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.2622825354337692,
      "epoch": 3.0608974358974357,
      "grad_norm": 0.021915055811405182,
      "learning_rate": 1e-06,
      "loss": -0.0031,
      "step": 1910
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2576129883527756,
      "epoch": 3.0625,
      "grad_norm": 0.013906807638704777,
      "learning_rate": 1e-06,
      "loss": 0.0443,
      "step": 1911
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2665242552757263,
      "epoch": 3.064102564102564,
      "grad_norm": 0.012783014215528965,
      "learning_rate": 1e-06,
      "loss": 0.0385,
      "step": 1912
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.021484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2434.0,
      "completions/mean_length": 1683.671875,
      "completions/mean_terminated_length": 1360.91015625,
      "completions/min_length": 601.0,
      "completions/min_terminated_length": 601.0,
      "entropy": 0.25943586230278015,
      "epoch": 3.065705128205128,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 292.14520263671875,
      "learning_rate": 1e-06,
      "loss": 0.0416,
      "num_tokens": 1167423891.0,
      "reward": 0.37675100564956665,
      "reward_std": 0.06550440192222595,
      "rewards/progression_diversity/mean": -0.005566404201090336,
      "rewards/progression_diversity/std": 0.03809665888547897,
      "rewards/symbolic_reward_accuracy/mean": 0.2578125,
      "rewards/symbolic_reward_accuracy/std": 0.43785804510116577,
      "rewards/symbolic_reward_partial_score/mean": 0.74169921875,
      "rewards/symbolic_reward_partial_score/std": 0.21923932433128357,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.039749026298523,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 12.124536514282227,
      "step": 1913
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.27193763852119446,
      "epoch": 3.0673076923076925,
      "grad_norm": 0.013166146352887154,
      "learning_rate": 1e-06,
      "loss": 0.0071,
      "step": 1914
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2520969808101654,
      "epoch": 3.0689102564102564,
      "grad_norm": 3.3681511878967285,
      "learning_rate": 1e-06,
      "loss": 0.057,
      "step": 1915
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.26810455322265625,
      "epoch": 3.0705128205128207,
      "grad_norm": 0.6854636073112488,
      "learning_rate": 1e-06,
      "loss": 0.0013,
      "step": 1916
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.033203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2121.0,
      "completions/mean_length": 1847.66796875,
      "completions/mean_terminated_length": 1348.4404296875,
      "completions/min_length": 637.0,
      "completions/min_terminated_length": 637.0,
      "entropy": 0.261865496635437,
      "epoch": 3.0721153846153846,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 45.9178581237793,
      "learning_rate": 1e-06,
      "loss": -0.0052,
      "num_tokens": 1169237785.0,
      "reward": 0.32650455832481384,
      "reward_std": 0.09211001545190811,
      "rewards/progression_diversity/mean": -0.009701091796159744,
      "rewards/progression_diversity/std": 0.052368562668561935,
      "rewards/symbolic_reward_accuracy/mean": 0.169921875,
      "rewards/symbolic_reward_accuracy/std": 0.3759314715862274,
      "rewards/symbolic_reward_partial_score/mean": 0.750781238079071,
      "rewards/symbolic_reward_partial_score/std": 0.19588179886341095,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0248017311096191,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 18.72113800048828,
      "step": 1917
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.24243663251399994,
      "epoch": 3.073717948717949,
      "grad_norm": 1099.10400390625,
      "learning_rate": 1e-06,
      "loss": 0.1447,
      "step": 1918
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2528642266988754,
      "epoch": 3.0753205128205128,
      "grad_norm": 0.01181944552809,
      "learning_rate": 1e-06,
      "loss": 0.0237,
      "step": 1919
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.25511983036994934,
      "epoch": 3.076923076923077,
      "grad_norm": 0.016715671867132187,
      "learning_rate": 1e-06,
      "loss": 0.0151,
      "step": 1920
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2228.0,
      "completions/mean_length": 1621.16796875,
      "completions/mean_terminated_length": 1327.087646484375,
      "completions/min_length": 613.0,
      "completions/min_terminated_length": 613.0,
      "entropy": 0.2685047686100006,
      "epoch": 3.078525641025641,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 0.027836158871650696,
      "learning_rate": 1e-06,
      "loss": 0.0567,
      "num_tokens": 1170851103.0,
      "reward": 0.36298680305480957,
      "reward_std": 0.06454600393772125,
      "rewards/progression_diversity/mean": -0.006008030381053686,
      "rewards/progression_diversity/std": 0.04314760863780975,
      "rewards/symbolic_reward_accuracy/mean": 0.228515625,
      "rewards/symbolic_reward_accuracy/std": 0.4202871024608612,
      "rewards/symbolic_reward_partial_score/mean": 0.7537760734558105,
      "rewards/symbolic_reward_partial_score/std": 0.1955997198820114,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0417566299438477,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 11.107623100280762,
      "step": 1921
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.25865650177001953,
      "epoch": 3.0801282051282053,
      "grad_norm": 3558.986328125,
      "learning_rate": 1e-06,
      "loss": 0.0839,
      "step": 1922
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.2664439529180527,
      "epoch": 3.081730769230769,
      "grad_norm": 0.014974008314311504,
      "learning_rate": 1e-06,
      "loss": -0.005,
      "step": 1923
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.25956542789936066,
      "epoch": 3.0833333333333335,
      "grad_norm": 0.009708881378173828,
      "learning_rate": 1e-06,
      "loss": 0.0435,
      "step": 1924
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 1970.0,
      "completions/mean_length": 2362.03515625,
      "completions/mean_terminated_length": 1301.550537109375,
      "completions/min_length": 400.0,
      "completions/min_terminated_length": 400.0,
      "entropy": 0.24602339416742325,
      "epoch": 3.0849358974358974,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 400.1549072265625,
      "learning_rate": 1e-06,
      "loss": 0.0486,
      "num_tokens": 1172988833.0,
      "reward": 0.3911093473434448,
      "reward_std": 0.03172388672828674,
      "rewards/progression_diversity/mean": -0.020903022959828377,
      "rewards/progression_diversity/std": 0.0763976201415062,
      "rewards/symbolic_reward_accuracy/mean": 0.2734375,
      "rewards/symbolic_reward_accuracy/std": 0.4461594223976135,
      "rewards/symbolic_reward_partial_score/mean": 0.759472668170929,
      "rewards/symbolic_reward_partial_score/std": 0.20224013924598694,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9978427886962891,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 31.93221664428711,
      "step": 1925
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.25485508143901825,
      "epoch": 3.0865384615384617,
      "grad_norm": 0.007848401553928852,
      "learning_rate": 1e-06,
      "loss": 0.0098,
      "step": 1926
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.23554416000843048,
      "epoch": 3.0881410256410255,
      "grad_norm": 0.011878496035933495,
      "learning_rate": 1e-06,
      "loss": 0.092,
      "step": 1927
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.21972551941871643,
      "epoch": 3.08974358974359,
      "grad_norm": 0.0162891186773777,
      "learning_rate": 1e-06,
      "loss": 0.1825,
      "step": 1928
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.111328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2478.0,
      "completions/mean_length": 3023.3359375,
      "completions/mean_terminated_length": 1349.58251953125,
      "completions/min_length": 533.0,
      "completions/min_terminated_length": 533.0,
      "entropy": 0.220417819917202,
      "epoch": 3.0913461538461537,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 1063.9591064453125,
      "learning_rate": 1e-06,
      "loss": 0.1881,
      "num_tokens": 1175419421.0,
      "reward": 0.32698261737823486,
      "reward_std": 0.024275176227092743,
      "rewards/progression_diversity/mean": -0.031721051782369614,
      "rewards/progression_diversity/std": 0.09057775884866714,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.7316243648529053,
      "rewards/symbolic_reward_partial_score/std": 0.19131198525428772,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9758140444755554,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 42.80847930908203,
      "step": 1929
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2313474789261818,
      "epoch": 3.092948717948718,
      "grad_norm": 0.009462164714932442,
      "learning_rate": 1e-06,
      "loss": 0.0303,
      "step": 1930
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.23852744698524475,
      "epoch": 3.094551282051282,
      "grad_norm": 0.00927420798689127,
      "learning_rate": 1e-06,
      "loss": 0.0868,
      "step": 1931
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.21710015833377838,
      "epoch": 3.0961538461538463,
      "grad_norm": 0.015988515689969063,
      "learning_rate": 1e-06,
      "loss": 0.1895,
      "step": 1932
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.072265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2227.0,
      "completions/mean_length": 2461.416015625,
      "completions/mean_terminated_length": 1376.919921875,
      "completions/min_length": 707.0,
      "completions/min_terminated_length": 707.0,
      "entropy": 0.24317217618227005,
      "epoch": 3.09775641025641,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1258.6036376953125,
      "learning_rate": 1e-06,
      "loss": 0.09,
      "num_tokens": 1177535618.0,
      "reward": 0.2823217809200287,
      "reward_std": 0.015544566325843334,
      "rewards/progression_diversity/mean": -0.020752854645252228,
      "rewards/progression_diversity/std": 0.07482607662677765,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.6937174797058105,
      "rewards/symbolic_reward_partial_score/std": 0.17785848677158356,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0054988861083984,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 28.267414093017578,
      "step": 1933
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.24698983132839203,
      "epoch": 3.0993589743589745,
      "grad_norm": 2.280089855194092,
      "learning_rate": 1e-06,
      "loss": 0.063,
      "step": 1934
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.24258919060230255,
      "epoch": 3.1009615384615383,
      "grad_norm": 0.01370900496840477,
      "learning_rate": 1e-06,
      "loss": 0.0806,
      "step": 1935
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.24202165007591248,
      "epoch": 3.1025641025641026,
      "grad_norm": 0.010354790836572647,
      "learning_rate": 1e-06,
      "loss": 0.0415,
      "step": 1936
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.091796875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2383.0,
      "completions/mean_length": 2724.0078125,
      "completions/mean_terminated_length": 1343.3204345703125,
      "completions/min_length": 516.0,
      "completions/min_terminated_length": 516.0,
      "entropy": 0.23228947818279266,
      "epoch": 3.1041666666666665,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 134.23583984375,
      "learning_rate": 1e-06,
      "loss": 0.0476,
      "num_tokens": 1179819430.0,
      "reward": 0.35509487986564636,
      "reward_std": 0.032482314854860306,
      "rewards/progression_diversity/mean": -0.02615930885076523,
      "rewards/progression_diversity/std": 0.08310631662607193,
      "rewards/symbolic_reward_accuracy/mean": 0.21484375,
      "rewards/symbolic_reward_accuracy/std": 0.4111155867576599,
      "rewards/symbolic_reward_partial_score/mean": 0.7567870616912842,
      "rewards/symbolic_reward_partial_score/std": 0.19756251573562622,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9856817722320557,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 37.22113800048828,
      "step": 1937
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.21816729754209518,
      "epoch": 3.105769230769231,
      "grad_norm": 22.10169792175293,
      "learning_rate": 1e-06,
      "loss": 0.1096,
      "step": 1938
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.23560451716184616,
      "epoch": 3.1073717948717947,
      "grad_norm": 861.4119873046875,
      "learning_rate": 1e-06,
      "loss": 0.1194,
      "step": 1939
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.237302266061306,
      "epoch": 3.108974358974359,
      "grad_norm": 0.013657595030963421,
      "learning_rate": 1e-06,
      "loss": 0.0515,
      "step": 1940
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.103515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2743.0,
      "completions/mean_length": 2930.619140625,
      "completions/mean_terminated_length": 1377.1785888671875,
      "completions/min_length": 395.0,
      "completions/min_terminated_length": 395.0,
      "entropy": 0.23268328607082367,
      "epoch": 3.110576923076923,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 118.8416976928711,
      "learning_rate": 1e-06,
      "loss": 0.0817,
      "num_tokens": 1182230323.0,
      "reward": 0.37685173749923706,
      "reward_std": 0.045152001082897186,
      "rewards/progression_diversity/mean": -0.029186122119426727,
      "rewards/progression_diversity/std": 0.08660390228033066,
      "rewards/symbolic_reward_accuracy/mean": 0.2578125,
      "rewards/symbolic_reward_accuracy/std": 0.43785804510116577,
      "rewards/symbolic_reward_partial_score/mean": 0.7434733510017395,
      "rewards/symbolic_reward_partial_score/std": 0.20847611129283905,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9919402599334717,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 33.77556610107422,
      "step": 1941
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.2476828247308731,
      "epoch": 3.1121794871794872,
      "grad_norm": 2530.47900390625,
      "learning_rate": 1e-06,
      "loss": 0.222,
      "step": 1942
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.22099629044532776,
      "epoch": 3.113782051282051,
      "grad_norm": 0.0288047194480896,
      "learning_rate": 1e-06,
      "loss": 0.1299,
      "step": 1943
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2267403081059456,
      "epoch": 3.1153846153846154,
      "grad_norm": 0.01607387140393257,
      "learning_rate": 1e-06,
      "loss": 0.1327,
      "step": 1944
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.091796875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2399.0,
      "completions/mean_length": 2772.017578125,
      "completions/mean_terminated_length": 1396.182861328125,
      "completions/min_length": 611.0,
      "completions/min_terminated_length": 611.0,
      "entropy": 0.2250806912779808,
      "epoch": 3.1169871794871793,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 1792.7398681640625,
      "learning_rate": 1e-06,
      "loss": 0.1175,
      "num_tokens": 1184533004.0,
      "reward": 0.28835493326187134,
      "reward_std": 0.013276607729494572,
      "rewards/progression_diversity/mean": -0.02485804632306099,
      "rewards/progression_diversity/std": 0.07890177518129349,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.7126628160476685,
      "rewards/symbolic_reward_partial_score/std": 0.17205090820789337,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.004730463027954,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 27.43810272216797,
      "step": 1945
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2443278580904007,
      "epoch": 3.1185897435897436,
      "grad_norm": 4188.24609375,
      "learning_rate": 1e-06,
      "loss": 0.5792,
      "step": 1946
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2495875358581543,
      "epoch": 3.1201923076923075,
      "grad_norm": 0.09753794223070145,
      "learning_rate": 1e-06,
      "loss": 0.0398,
      "step": 1947
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.24468916654586792,
      "epoch": 3.121794871794872,
      "grad_norm": 0.011647003702819347,
      "learning_rate": 1e-06,
      "loss": 0.0852,
      "step": 1948
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.083984375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2344.0,
      "completions/mean_length": 2688.154296875,
      "completions/mean_terminated_length": 1432.45849609375,
      "completions/min_length": 568.0,
      "completions/min_terminated_length": 568.0,
      "entropy": 0.2420196533203125,
      "epoch": 3.123397435897436,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 359.50616455078125,
      "learning_rate": 1e-06,
      "loss": 0.1009,
      "num_tokens": 1186778155.0,
      "reward": 0.32234811782836914,
      "reward_std": 0.037715356796979904,
      "rewards/progression_diversity/mean": -0.022023096680641174,
      "rewards/progression_diversity/std": 0.07346688210964203,
      "rewards/symbolic_reward_accuracy/mean": 0.177734375,
      "rewards/symbolic_reward_accuracy/std": 0.3826628625392914,
      "rewards/symbolic_reward_partial_score/mean": 0.7197591066360474,
      "rewards/symbolic_reward_partial_score/std": 0.20028026401996613,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.001284122467041,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 28.891010284423828,
      "step": 1949
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.24103453755378723,
      "epoch": 3.125,
      "grad_norm": 589.58740234375,
      "learning_rate": 1e-06,
      "loss": 2.001,
      "step": 1950
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.24162401258945465,
      "epoch": 3.126602564102564,
      "grad_norm": 5572.74560546875,
      "learning_rate": 1e-06,
      "loss": 0.4189,
      "step": 1951
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.234800323843956,
      "epoch": 3.128205128205128,
      "grad_norm": 475.8064270019531,
      "learning_rate": 1e-06,
      "loss": 0.2053,
      "step": 1952
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.11328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2726.0,
      "completions/mean_length": 3146.59375,
      "completions/mean_terminated_length": 1455.4713134765625,
      "completions/min_length": 452.0,
      "completions/min_terminated_length": 452.0,
      "entropy": 0.2195385992527008,
      "epoch": 3.1298076923076925,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 503.7859802246094,
      "learning_rate": 1e-06,
      "loss": 0.1427,
      "num_tokens": 1189316267.0,
      "reward": 0.269339382648468,
      "reward_std": 0.020436681807041168,
      "rewards/progression_diversity/mean": -0.027488097548484802,
      "rewards/progression_diversity/std": 0.0782787874341011,
      "rewards/symbolic_reward_accuracy/mean": 0.09765625,
      "rewards/symbolic_reward_accuracy/std": 0.29713961482048035,
      "rewards/symbolic_reward_partial_score/mean": 0.7034016847610474,
      "rewards/symbolic_reward_partial_score/std": 0.17726561427116394,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9874823093414307,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 35.55792999267578,
      "step": 1953
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.328125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.25392740964889526,
      "epoch": 3.1314102564102564,
      "grad_norm": 6332.8427734375,
      "learning_rate": 1e-06,
      "loss": 0.1834,
      "step": 1954
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2183244675397873,
      "epoch": 3.1330128205128207,
      "grad_norm": 0.011082753539085388,
      "learning_rate": 1e-06,
      "loss": 0.1257,
      "step": 1955
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.21788176894187927,
      "epoch": 3.1346153846153846,
      "grad_norm": 0.012296471744775772,
      "learning_rate": 1e-06,
      "loss": 0.1528,
      "step": 1956
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.103515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2483.0,
      "completions/mean_length": 3008.599609375,
      "completions/mean_terminated_length": 1464.163330078125,
      "completions/min_length": 584.0,
      "completions/min_terminated_length": 584.0,
      "entropy": 0.2399488091468811,
      "epoch": 3.136217948717949,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 634.225341796875,
      "learning_rate": 1e-06,
      "loss": 0.1405,
      "num_tokens": 1191638702.0,
      "reward": 0.31707698106765747,
      "reward_std": 0.049435004591941833,
      "rewards/progression_diversity/mean": -0.02521352283656597,
      "rewards/progression_diversity/std": 0.07498207688331604,
      "rewards/symbolic_reward_accuracy/mean": 0.169921875,
      "rewards/symbolic_reward_accuracy/std": 0.3759314715862274,
      "rewards/symbolic_reward_partial_score/mean": 0.7185709476470947,
      "rewards/symbolic_reward_partial_score/std": 0.21633154153823853,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.985991895198822,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 37.30690002441406,
      "step": 1957
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.2299083024263382,
      "epoch": 3.1378205128205128,
      "grad_norm": 0.010621101595461369,
      "learning_rate": 1e-06,
      "loss": 0.1298,
      "step": 1958
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2408609539270401,
      "epoch": 3.139423076923077,
      "grad_norm": 0.011294880881905556,
      "learning_rate": 1e-06,
      "loss": 0.0936,
      "step": 1959
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.23295968770980835,
      "epoch": 3.141025641025641,
      "grad_norm": 0.014650301076471806,
      "learning_rate": 1e-06,
      "loss": 0.0687,
      "step": 1960
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.11328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2611.0,
      "completions/mean_length": 3111.46875,
      "completions/mean_terminated_length": 1415.8590087890625,
      "completions/min_length": 392.0,
      "completions/min_terminated_length": 392.0,
      "entropy": 0.24924977123737335,
      "epoch": 3.1426282051282053,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 237.02276611328125,
      "learning_rate": 1e-06,
      "loss": 0.0949,
      "num_tokens": 1194090830.0,
      "reward": 0.3610386848449707,
      "reward_std": 0.037800274789333344,
      "rewards/progression_diversity/mean": -0.028455514460802078,
      "rewards/progression_diversity/std": 0.08039110153913498,
      "rewards/symbolic_reward_accuracy/mean": 0.224609375,
      "rewards/symbolic_reward_accuracy/std": 0.41773295402526855,
      "rewards/symbolic_reward_partial_score/mean": 0.7551920413970947,
      "rewards/symbolic_reward_partial_score/std": 0.19215147197246552,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9815695285797119,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 40.95595169067383,
      "step": 1961
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.22469578683376312,
      "epoch": 3.144230769230769,
      "grad_norm": 0.009706591255962849,
      "learning_rate": 1e-06,
      "loss": 0.1529,
      "step": 1962
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.22220056504011154,
      "epoch": 3.1458333333333335,
      "grad_norm": 0.02369273081421852,
      "learning_rate": 1e-06,
      "loss": 0.1648,
      "step": 1963
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.25326360762119293,
      "epoch": 3.1474358974358974,
      "grad_norm": 0.015049627050757408,
      "learning_rate": 1e-06,
      "loss": 0.051,
      "step": 1964
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.07421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2528.0,
      "completions/mean_length": 2517.6328125,
      "completions/mean_terminated_length": 1405.9830322265625,
      "completions/min_length": 442.0,
      "completions/min_terminated_length": 442.0,
      "entropy": 0.25177982449531555,
      "epoch": 3.1490384615384617,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 872.0093994140625,
      "learning_rate": 1e-06,
      "loss": 0.0878,
      "num_tokens": 1196269138.0,
      "reward": 0.3650954067707062,
      "reward_std": 0.032753195613622665,
      "rewards/progression_diversity/mean": -0.019758375361561775,
      "rewards/progression_diversity/std": 0.07037756592035294,
      "rewards/symbolic_reward_accuracy/mean": 0.23828125,
      "rewards/symbolic_reward_accuracy/std": 0.42644867300987244,
      "rewards/symbolic_reward_partial_score/mean": 0.7410807013511658,
      "rewards/symbolic_reward_partial_score/std": 0.20055758953094482,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.004058599472046,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 30.5396671295166,
      "step": 1965
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.25476063042879105,
      "epoch": 3.1506410256410255,
      "grad_norm": 0.01801050268113613,
      "learning_rate": 1e-06,
      "loss": 0.0639,
      "step": 1966
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.24143607169389725,
      "epoch": 3.15224358974359,
      "grad_norm": 0.009572253562510014,
      "learning_rate": 1e-06,
      "loss": 0.114,
      "step": 1967
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2614782154560089,
      "epoch": 3.1538461538461537,
      "grad_norm": 0.02647644467651844,
      "learning_rate": 1e-06,
      "loss": 0.04,
      "step": 1968
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.056640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2685.0,
      "completions/mean_length": 2340.412109375,
      "completions/mean_terminated_length": 1497.21533203125,
      "completions/min_length": 473.0,
      "completions/min_terminated_length": 473.0,
      "entropy": 0.2505124807357788,
      "epoch": 3.155448717948718,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 1409.9781494140625,
      "learning_rate": 1e-06,
      "loss": 0.1449,
      "num_tokens": 1198283685.0,
      "reward": 0.4389854073524475,
      "reward_std": 0.043696288019418716,
      "rewards/progression_diversity/mean": -0.015524028800427914,
      "rewards/progression_diversity/std": 0.06471197307109833,
      "rewards/symbolic_reward_accuracy/mean": 0.34765625,
      "rewards/symbolic_reward_accuracy/std": 0.47669193148612976,
      "rewards/symbolic_reward_partial_score/mean": 0.7691406011581421,
      "rewards/symbolic_reward_partial_score/std": 0.21009880304336548,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0053313970565796,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 29.34543800354004,
      "step": 1969
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2559570074081421,
      "epoch": 3.157051282051282,
      "grad_norm": 1054.045654296875,
      "learning_rate": 1e-06,
      "loss": 0.082,
      "step": 1970
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2517922967672348,
      "epoch": 3.1586538461538463,
      "grad_norm": 10.967491149902344,
      "learning_rate": 1e-06,
      "loss": -0.0051,
      "step": 1971
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.24725372344255447,
      "epoch": 3.16025641025641,
      "grad_norm": 0.010240145027637482,
      "learning_rate": 1e-06,
      "loss": 0.0999,
      "step": 1972
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.09765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2716.0,
      "completions/mean_length": 2950.55859375,
      "completions/mean_terminated_length": 1496.722900390625,
      "completions/min_length": 423.0,
      "completions/min_terminated_length": 423.0,
      "entropy": 0.25035959482192993,
      "epoch": 3.1618589743589745,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 633.6778564453125,
      "learning_rate": 1e-06,
      "loss": 0.0946,
      "num_tokens": 1200602867.0,
      "reward": 0.33387240767478943,
      "reward_std": 0.026298727840185165,
      "rewards/progression_diversity/mean": -0.02584683895111084,
      "rewards/progression_diversity/std": 0.07925648987293243,
      "rewards/symbolic_reward_accuracy/mean": 0.197265625,
      "rewards/symbolic_reward_accuracy/std": 0.3983237147331238,
      "rewards/symbolic_reward_partial_score/mean": 0.71923828125,
      "rewards/symbolic_reward_partial_score/std": 0.1927872896194458,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9945404529571533,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 35.76648712158203,
      "step": 1973
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.25055961310863495,
      "epoch": 3.1634615384615383,
      "grad_norm": 278.57867431640625,
      "learning_rate": 1e-06,
      "loss": 0.0723,
      "step": 1974
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.23513774573802948,
      "epoch": 3.1650641025641026,
      "grad_norm": 0.15121746063232422,
      "learning_rate": 1e-06,
      "loss": 0.1125,
      "step": 1975
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2441992089152336,
      "epoch": 3.1666666666666665,
      "grad_norm": 0.013731342740356922,
      "learning_rate": 1e-06,
      "loss": 0.0602,
      "step": 1976
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.11328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3033.0,
      "completions/mean_length": 3148.57421875,
      "completions/mean_terminated_length": 1457.704833984375,
      "completions/min_length": 573.0,
      "completions/min_terminated_length": 573.0,
      "entropy": 0.21663058549165726,
      "epoch": 3.168269230769231,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 515.6699829101562,
      "learning_rate": 1e-06,
      "loss": 0.0905,
      "num_tokens": 1203158377.0,
      "reward": 0.2782261371612549,
      "reward_std": 0.03023504465818405,
      "rewards/progression_diversity/mean": -0.029924802482128143,
      "rewards/progression_diversity/std": 0.08487491309642792,
      "rewards/symbolic_reward_accuracy/mean": 0.119140625,
      "rewards/symbolic_reward_accuracy/std": 0.32427072525024414,
      "rewards/symbolic_reward_partial_score/mean": 0.6914387941360474,
      "rewards/symbolic_reward_partial_score/std": 0.19170799851417542,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9659879207611084,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 49.51683807373047,
      "step": 1977
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2332163155078888,
      "epoch": 3.1698717948717947,
      "grad_norm": 2.1581897735595703,
      "learning_rate": 1e-06,
      "loss": 0.0941,
      "step": 1978
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.21122775226831436,
      "epoch": 3.171474358974359,
      "grad_norm": 0.049657199531793594,
      "learning_rate": 1e-06,
      "loss": 0.1684,
      "step": 1979
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.24628569930791855,
      "epoch": 3.173076923076923,
      "grad_norm": 0.019608579576015472,
      "learning_rate": 1e-06,
      "loss": 0.0439,
      "step": 1980
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.083984375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3310.0,
      "completions/mean_length": 2705.47265625,
      "completions/mean_terminated_length": 1451.3646240234375,
      "completions/min_length": 542.0,
      "completions/min_terminated_length": 542.0,
      "entropy": 0.26659801602363586,
      "epoch": 3.1746794871794872,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 256.61602783203125,
      "learning_rate": 1e-06,
      "loss": 0.0419,
      "num_tokens": 1205377627.0,
      "reward": 0.40071532130241394,
      "reward_std": 0.029286310076713562,
      "rewards/progression_diversity/mean": -0.022705569863319397,
      "rewards/progression_diversity/std": 0.07598486542701721,
      "rewards/symbolic_reward_accuracy/mean": 0.27734375,
      "rewards/symbolic_reward_accuracy/std": 0.4481254518032074,
      "rewards/symbolic_reward_partial_score/mean": 0.7830891609191895,
      "rewards/symbolic_reward_partial_score/std": 0.1975134015083313,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9810075759887695,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 43.57196044921875,
      "step": 1981
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.23883583396673203,
      "epoch": 3.176282051282051,
      "grad_norm": 2.147189140319824,
      "learning_rate": 1e-06,
      "loss": 0.0706,
      "step": 1982
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.26977184414863586,
      "epoch": 3.1778846153846154,
      "grad_norm": 78.77035522460938,
      "learning_rate": 1e-06,
      "loss": 0.0521,
      "step": 1983
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23553596436977386,
      "epoch": 3.1794871794871793,
      "grad_norm": 0.018339024856686592,
      "learning_rate": 1e-06,
      "loss": 0.5282,
      "step": 1984
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.068359375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2776.0,
      "completions/mean_length": 2496.416015625,
      "completions/mean_terminated_length": 1477.410888671875,
      "completions/min_length": 393.0,
      "completions/min_terminated_length": 393.0,
      "entropy": 0.2516746520996094,
      "epoch": 3.1810897435897436,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 319.26788330078125,
      "learning_rate": 1e-06,
      "loss": 0.0477,
      "num_tokens": 1207524576.0,
      "reward": 0.3050941824913025,
      "reward_std": 0.039040304720401764,
      "rewards/progression_diversity/mean": -0.017927538603544235,
      "rewards/progression_diversity/std": 0.06751146912574768,
      "rewards/symbolic_reward_accuracy/mean": 0.146484375,
      "rewards/symbolic_reward_accuracy/std": 0.35393697023391724,
      "rewards/symbolic_reward_partial_score/mean": 0.7259114980697632,
      "rewards/symbolic_reward_partial_score/std": 0.20375892519950867,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0033478736877441,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 31.887889862060547,
      "step": 1985
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.26808346807956696,
      "epoch": 3.1826923076923075,
      "grad_norm": 0.020371215417981148,
      "learning_rate": 1e-06,
      "loss": 0.0406,
      "step": 1986
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.24439021944999695,
      "epoch": 3.184294871794872,
      "grad_norm": 0.026235945522785187,
      "learning_rate": 1e-06,
      "loss": 0.0765,
      "step": 1987
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.25858327746391296,
      "epoch": 3.185897435897436,
      "grad_norm": 0.016966918483376503,
      "learning_rate": 1e-06,
      "loss": 0.0628,
      "step": 1988
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2761.0,
      "completions/mean_length": 1891.138671875,
      "completions/mean_terminated_length": 1363.0587158203125,
      "completions/min_length": 651.0,
      "completions/min_terminated_length": 651.0,
      "entropy": 0.28316083550453186,
      "epoch": 3.1875,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 3698.61865234375,
      "learning_rate": 1e-06,
      "loss": 0.0706,
      "num_tokens": 1209406615.0,
      "reward": 0.3008219003677368,
      "reward_std": 0.029717685654759407,
      "rewards/progression_diversity/mean": -0.010097447782754898,
      "rewards/progression_diversity/std": 0.053449101746082306,
      "rewards/symbolic_reward_accuracy/mean": 0.142578125,
      "rewards/symbolic_reward_accuracy/std": 0.3499840497970581,
      "rewards/symbolic_reward_partial_score/mean": 0.7185709476470947,
      "rewards/symbolic_reward_partial_score/std": 0.17875374853610992,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0198476314544678,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 25.146913528442383,
      "step": 1989
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.2714526355266571,
      "epoch": 3.189102564102564,
      "grad_norm": 7592.54052734375,
      "learning_rate": 1e-06,
      "loss": 0.8777,
      "step": 1990
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.28693751990795135,
      "epoch": 3.190705128205128,
      "grad_norm": 20502.578125,
      "learning_rate": 1e-06,
      "loss": 2.4146,
      "step": 1991
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.283732146024704,
      "epoch": 3.1923076923076925,
      "grad_norm": 0.011515190824866295,
      "learning_rate": 1e-06,
      "loss": 0.043,
      "step": 1992
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2856.0,
      "completions/mean_length": 1891.53125,
      "completions/mean_terminated_length": 1363.465576171875,
      "completions/min_length": 405.0,
      "completions/min_terminated_length": 405.0,
      "entropy": 0.2893388271331787,
      "epoch": 3.1939102564102564,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 136.03277587890625,
      "learning_rate": 1e-06,
      "loss": 0.0309,
      "num_tokens": 1211197511.0,
      "reward": 0.38298678398132324,
      "reward_std": 0.05362379923462868,
      "rewards/progression_diversity/mean": -0.010407653637230396,
      "rewards/progression_diversity/std": 0.05494730547070503,
      "rewards/symbolic_reward_accuracy/mean": 0.26171875,
      "rewards/symbolic_reward_accuracy/std": 0.44000017642974854,
      "rewards/symbolic_reward_partial_score/mean": 0.754833996295929,
      "rewards/symbolic_reward_partial_score/std": 0.21281343698501587,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0291106700897217,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 21.492839813232422,
      "step": 1993
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.2943323999643326,
      "epoch": 3.1955128205128207,
      "grad_norm": 0.021322406828403473,
      "learning_rate": 1e-06,
      "loss": 0.033,
      "step": 1994
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.29063861072063446,
      "epoch": 3.1971153846153846,
      "grad_norm": 0.012699613347649574,
      "learning_rate": 1e-06,
      "loss": 0.0278,
      "step": 1995
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.28081028163433075,
      "epoch": 3.198717948717949,
      "grad_norm": 0.00812804140150547,
      "learning_rate": 1e-06,
      "loss": 0.0823,
      "step": 1996
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2786.0,
      "completions/mean_length": 1646.9921875,
      "completions/mean_terminated_length": 1353.4263916015625,
      "completions/min_length": 463.0,
      "completions/min_terminated_length": 463.0,
      "entropy": 0.30514436960220337,
      "epoch": 3.2003205128205128,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 60.97207260131836,
      "learning_rate": 1e-06,
      "loss": -0.0098,
      "num_tokens": 1212900051.0,
      "reward": 0.32839280366897583,
      "reward_std": 0.04186505824327469,
      "rewards/progression_diversity/mean": -0.005448305979371071,
      "rewards/progression_diversity/std": 0.039716459810733795,
      "rewards/symbolic_reward_accuracy/mean": 0.173828125,
      "rewards/symbolic_reward_accuracy/std": 0.3793322443962097,
      "rewards/symbolic_reward_partial_score/mean": 0.7497720718383789,
      "rewards/symbolic_reward_partial_score/std": 0.1919063776731491,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0453128814697266,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 13.722127914428711,
      "step": 1997
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.29429739713668823,
      "epoch": 3.201923076923077,
      "grad_norm": 349.4608459472656,
      "learning_rate": 1e-06,
      "loss": 0.0811,
      "step": 1998
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.30618786811828613,
      "epoch": 3.203525641025641,
      "grad_norm": 0.018446296453475952,
      "learning_rate": 1e-06,
      "loss": 0.0303,
      "step": 1999
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.29249128699302673,
      "epoch": 3.2051282051282053,
      "grad_norm": 0.015572361648082733,
      "learning_rate": 1e-06,
      "loss": 0.0433,
      "step": 2000
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2926.0,
      "completions/mean_length": 1567.892578125,
      "completions/mean_terminated_length": 1362.5208740234375,
      "completions/min_length": 600.0,
      "completions/min_terminated_length": 600.0,
      "entropy": 0.2950544059276581,
      "epoch": 3.206730769230769,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 891.14892578125,
      "learning_rate": 1e-06,
      "loss": 0.0646,
      "num_tokens": 1214607692.0,
      "reward": 0.26205992698669434,
      "reward_std": 0.03215021640062332,
      "rewards/progression_diversity/mean": -0.003969744313508272,
      "rewards/progression_diversity/std": 0.03686157613992691,
      "rewards/symbolic_reward_accuracy/mean": 0.087890625,
      "rewards/symbolic_reward_accuracy/std": 0.2834126651287079,
      "rewards/symbolic_reward_partial_score/mean": 0.6991862058639526,
      "rewards/symbolic_reward_partial_score/std": 0.1865617036819458,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0514659881591797,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 10.364349365234375,
      "step": 2001
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.29572050273418427,
      "epoch": 3.2083333333333335,
      "grad_norm": 0.008014540188014507,
      "learning_rate": 1e-06,
      "loss": 0.0259,
      "step": 2002
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2993382215499878,
      "epoch": 3.2099358974358974,
      "grad_norm": 0.021720819175243378,
      "learning_rate": 1e-06,
      "loss": -0.0056,
      "step": 2003
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.3016369044780731,
      "epoch": 3.2115384615384617,
      "grad_norm": 0.0189630426466465,
      "learning_rate": 1e-06,
      "loss": 0.0159,
      "step": 2004
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.021484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3574.0,
      "completions/mean_length": 1687.447265625,
      "completions/mean_terminated_length": 1364.7684326171875,
      "completions/min_length": 483.0,
      "completions/min_terminated_length": 483.0,
      "entropy": 0.29340440034866333,
      "epoch": 3.2131410256410255,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 509.89044189453125,
      "learning_rate": 1e-06,
      "loss": 0.0367,
      "num_tokens": 1216400945.0,
      "reward": 0.3805909752845764,
      "reward_std": 0.03878272697329521,
      "rewards/progression_diversity/mean": -0.007796908728778362,
      "rewards/progression_diversity/std": 0.05300579220056534,
      "rewards/symbolic_reward_accuracy/mean": 0.259765625,
      "rewards/symbolic_reward_accuracy/std": 0.4389347732067108,
      "rewards/symbolic_reward_partial_score/mean": 0.7513183355331421,
      "rewards/symbolic_reward_partial_score/std": 0.19548480212688446,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0402953624725342,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 16.1550235748291,
      "step": 2005
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.30069899559020996,
      "epoch": 3.21474358974359,
      "grad_norm": 0.02268948033452034,
      "learning_rate": 1e-06,
      "loss": 0.0185,
      "step": 2006
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3007173389196396,
      "epoch": 3.2163461538461537,
      "grad_norm": 0.008483972400426865,
      "learning_rate": 1e-06,
      "loss": 0.0269,
      "step": 2007
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2845690995454788,
      "epoch": 3.217948717948718,
      "grad_norm": 0.010253841057419777,
      "learning_rate": 1e-06,
      "loss": 0.0552,
      "step": 2008
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2020.0,
      "completions/mean_length": 1595.154296875,
      "completions/mean_terminated_length": 1360.4107666015625,
      "completions/min_length": 422.0,
      "completions/min_terminated_length": 422.0,
      "entropy": 0.28091390430927277,
      "epoch": 3.219551282051282,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 675.5444946289062,
      "learning_rate": 1e-06,
      "loss": 0.0337,
      "num_tokens": 1218156384.0,
      "reward": 0.2970985472202301,
      "reward_std": 0.03322647139430046,
      "rewards/progression_diversity/mean": -0.005477225407958031,
      "rewards/progression_diversity/std": 0.044522009789943695,
      "rewards/symbolic_reward_accuracy/mean": 0.142578125,
      "rewards/symbolic_reward_accuracy/std": 0.3499840497970581,
      "rewards/symbolic_reward_partial_score/mean": 0.7066569328308105,
      "rewards/symbolic_reward_partial_score/std": 0.1901051253080368,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0481798648834229,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 11.399556159973145,
      "step": 2009
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.29631251096725464,
      "epoch": 3.2211538461538463,
      "grad_norm": 0.01047491654753685,
      "learning_rate": 1e-06,
      "loss": 0.004,
      "step": 2010
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.2931921035051346,
      "epoch": 3.22275641025641,
      "grad_norm": 0.03937433660030365,
      "learning_rate": 1e-06,
      "loss": 0.0033,
      "step": 2011
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.29075874388217926,
      "epoch": 3.2243589743589745,
      "grad_norm": 0.010890385136008263,
      "learning_rate": 1e-06,
      "loss": 0.017,
      "step": 2012
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2629.0,
      "completions/mean_length": 1558.37890625,
      "completions/mean_terminated_length": 1382.5810546875,
      "completions/min_length": 729.0,
      "completions/min_terminated_length": 729.0,
      "entropy": 0.282899871468544,
      "epoch": 3.2259615384615383,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 980.3523559570312,
      "learning_rate": 1e-06,
      "loss": 0.0471,
      "num_tokens": 1219817762.0,
      "reward": 0.3504934310913086,
      "reward_std": 0.030621502548456192,
      "rewards/progression_diversity/mean": -0.004368623252958059,
      "rewards/progression_diversity/std": 0.040239714086055756,
      "rewards/symbolic_reward_accuracy/mean": 0.228515625,
      "rewards/symbolic_reward_accuracy/std": 0.4202871024608612,
      "rewards/symbolic_reward_partial_score/mean": 0.71142578125,
      "rewards/symbolic_reward_partial_score/std": 0.2032359093427658,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0505871772766113,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 9.663378715515137,
      "step": 2013
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2933056503534317,
      "epoch": 3.2275641025641026,
      "grad_norm": 0.016422249376773834,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "step": 2014
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.2866353541612625,
      "epoch": 3.2291666666666665,
      "grad_norm": 0.020436976104974747,
      "learning_rate": 1e-06,
      "loss": 0.0241,
      "step": 2015
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.28925251960754395,
      "epoch": 3.230769230769231,
      "grad_norm": 0.018621278926730156,
      "learning_rate": 1e-06,
      "loss": 0.0175,
      "step": 2016
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2538.0,
      "completions/mean_length": 1511.5234375,
      "completions/mean_terminated_length": 1335.1700439453125,
      "completions/min_length": 736.0,
      "completions/min_terminated_length": 736.0,
      "entropy": 0.29188382625579834,
      "epoch": 3.2323717948717947,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.03130757808685303,
      "learning_rate": 1e-06,
      "loss": 0.0056,
      "num_tokens": 1221518830.0,
      "reward": 0.3500271439552307,
      "reward_std": 0.07550652325153351,
      "rewards/progression_diversity/mean": -0.0036349566653370857,
      "rewards/progression_diversity/std": 0.03353985399007797,
      "rewards/symbolic_reward_accuracy/mean": 0.23828125,
      "rewards/symbolic_reward_accuracy/std": 0.42644867300987244,
      "rewards/symbolic_reward_partial_score/mean": 0.6909668445587158,
      "rewards/symbolic_reward_partial_score/std": 0.22590909898281097,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0503188371658325,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 8.717741012573242,
      "step": 2017
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.2773764133453369,
      "epoch": 3.233974358974359,
      "grad_norm": 0.017145292833447456,
      "learning_rate": 1e-06,
      "loss": 0.0126,
      "step": 2018
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.27354469895362854,
      "epoch": 3.235576923076923,
      "grad_norm": 0.011831255629658699,
      "learning_rate": 1e-06,
      "loss": 0.0443,
      "step": 2019
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0390625,
      "clip_ratio/low_mean": 0.234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2817593514919281,
      "epoch": 3.2371794871794872,
      "grad_norm": 0.016237668693065643,
      "learning_rate": 1e-06,
      "loss": 0.0098,
      "step": 2020
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2421.0,
      "completions/mean_length": 1795.888671875,
      "completions/mean_terminated_length": 1325.304443359375,
      "completions/min_length": 592.0,
      "completions/min_terminated_length": 592.0,
      "entropy": 0.2711557298898697,
      "epoch": 3.238782051282051,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 163.37515258789062,
      "learning_rate": 1e-06,
      "loss": 0.0598,
      "num_tokens": 1223317701.0,
      "reward": 0.3071795105934143,
      "reward_std": 0.022627130150794983,
      "rewards/progression_diversity/mean": -0.010563986375927925,
      "rewards/progression_diversity/std": 0.05871390923857689,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.6570963263511658,
      "rewards/symbolic_reward_partial_score/std": 0.21646232903003693,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0290942192077637,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 19.123592376708984,
      "step": 2021
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2706785798072815,
      "epoch": 3.2403846153846154,
      "grad_norm": 0.007887561805546284,
      "learning_rate": 1e-06,
      "loss": 0.0574,
      "step": 2022
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2705296725034714,
      "epoch": 3.2419871794871793,
      "grad_norm": 0.016599016264081,
      "learning_rate": 1e-06,
      "loss": 0.0139,
      "step": 2023
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.26619046926498413,
      "epoch": 3.2435897435897436,
      "grad_norm": 0.014114703983068466,
      "learning_rate": 1e-06,
      "loss": 0.049,
      "step": 2024
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2063.0,
      "completions/mean_length": 1582.685546875,
      "completions/mean_terminated_length": 1317.850830078125,
      "completions/min_length": 603.0,
      "completions/min_terminated_length": 603.0,
      "entropy": 0.27589017152786255,
      "epoch": 3.2451923076923075,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 714.7753295898438,
      "learning_rate": 1e-06,
      "loss": 0.0405,
      "num_tokens": 1224991732.0,
      "reward": 0.32780733704566956,
      "reward_std": 0.02535293623805046,
      "rewards/progression_diversity/mean": -0.006377549842000008,
      "rewards/progression_diversity/std": 0.04781962186098099,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.7276692390441895,
      "rewards/symbolic_reward_partial_score/std": 0.2073507308959961,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.040276288986206,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 13.916572570800781,
      "step": 2025
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.278276264667511,
      "epoch": 3.246794871794872,
      "grad_norm": 641.1627807617188,
      "learning_rate": 1e-06,
      "loss": 0.081,
      "step": 2026
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.2818632125854492,
      "epoch": 3.248397435897436,
      "grad_norm": 0.015987196937203407,
      "learning_rate": 1e-06,
      "loss": 0.0139,
      "step": 2027
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2770925313234329,
      "epoch": 3.25,
      "grad_norm": 0.011531771160662174,
      "learning_rate": 1e-06,
      "loss": -0.0061,
      "step": 2028
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2199.0,
      "completions/mean_length": 1594.681640625,
      "completions/mean_terminated_length": 1330.0615234375,
      "completions/min_length": 639.0,
      "completions/min_terminated_length": 639.0,
      "entropy": 0.28490838408470154,
      "epoch": 3.251602564102564,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 630.0040283203125,
      "learning_rate": 1e-06,
      "loss": 0.0141,
      "num_tokens": 1226658273.0,
      "reward": 0.3503781259059906,
      "reward_std": 0.02471214532852173,
      "rewards/progression_diversity/mean": -0.0061342837288975716,
      "rewards/progression_diversity/std": 0.0460776686668396,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.7469075322151184,
      "rewards/symbolic_reward_partial_score/std": 0.19170016050338745,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.040297508239746,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 14.458892822265625,
      "step": 2029
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.28325770795345306,
      "epoch": 3.253205128205128,
      "grad_norm": 0.02777559868991375,
      "learning_rate": 1e-06,
      "loss": 0.0386,
      "step": 2030
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.2924428731203079,
      "epoch": 3.2548076923076925,
      "grad_norm": 0.005939020775258541,
      "learning_rate": 1e-06,
      "loss": -0.0084,
      "step": 2031
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.26891082525253296,
      "epoch": 3.2564102564102564,
      "grad_norm": 0.010626004077494144,
      "learning_rate": 1e-06,
      "loss": 0.1295,
      "step": 2032
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2179.0,
      "completions/mean_length": 1786.677734375,
      "completions/mean_terminated_length": 1315.79638671875,
      "completions/min_length": 671.0,
      "completions/min_terminated_length": 671.0,
      "entropy": 0.29806482791900635,
      "epoch": 3.2580128205128207,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.012230968102812767,
      "learning_rate": 1e-06,
      "loss": -0.0145,
      "num_tokens": 1228425196.0,
      "reward": 0.32576221227645874,
      "reward_std": 0.05162687227129936,
      "rewards/progression_diversity/mean": -0.010695156641304493,
      "rewards/progression_diversity/std": 0.05995798856019974,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.717089831829071,
      "rewards/symbolic_reward_partial_score/std": 0.22520095109939575,
      "rewards/tag_count_reward/mean": -0.017578125,
      "rewards/tag_count_reward/std": 0.13154059648513794,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0235651731491089,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 23.860057830810547,
      "step": 2033
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2714930325746536,
      "epoch": 3.2596153846153846,
      "grad_norm": 0.009989427402615547,
      "learning_rate": 1e-06,
      "loss": 0.145,
      "step": 2034
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2667497843503952,
      "epoch": 3.261217948717949,
      "grad_norm": 0.01684493198990822,
      "learning_rate": 1e-06,
      "loss": 0.1274,
      "step": 2035
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.2843262106180191,
      "epoch": 3.2628205128205128,
      "grad_norm": 0.014857680536806583,
      "learning_rate": 1e-06,
      "loss": 0.0398,
      "step": 2036
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3595.0,
      "completions/mean_length": 1610.28125,
      "completions/mean_terminated_length": 1345.9403076171875,
      "completions/min_length": 710.0,
      "completions/min_terminated_length": 710.0,
      "entropy": 0.2858640253543854,
      "epoch": 3.264423076923077,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 336.9473571777344,
      "learning_rate": 1e-06,
      "loss": 0.0252,
      "num_tokens": 1230085228.0,
      "reward": 0.3623395264148712,
      "reward_std": 0.04071733355522156,
      "rewards/progression_diversity/mean": -0.005795993376523256,
      "rewards/progression_diversity/std": 0.04330339655280113,
      "rewards/symbolic_reward_accuracy/mean": 0.228515625,
      "rewards/symbolic_reward_accuracy/std": 0.4202871024608612,
      "rewards/symbolic_reward_partial_score/mean": 0.7535644173622131,
      "rewards/symbolic_reward_partial_score/std": 0.21194274723529816,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045485258102417,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 12.358026504516602,
      "step": 2037
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.2867206782102585,
      "epoch": 3.266025641025641,
      "grad_norm": 0.012359599582850933,
      "learning_rate": 1e-06,
      "loss": 0.0326,
      "step": 2038
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.279480904340744,
      "epoch": 3.2676282051282053,
      "grad_norm": 0.009706364013254642,
      "learning_rate": 1e-06,
      "loss": 0.0751,
      "step": 2039
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2926556468009949,
      "epoch": 3.269230769230769,
      "grad_norm": 0.013265096582472324,
      "learning_rate": 1e-06,
      "loss": 0.0111,
      "step": 2040
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.021484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3382.0,
      "completions/mean_length": 1684.5390625,
      "completions/mean_terminated_length": 1361.79638671875,
      "completions/min_length": 597.0,
      "completions/min_terminated_length": 597.0,
      "entropy": 0.27776144444942474,
      "epoch": 3.2708333333333335,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 754.3441162109375,
      "learning_rate": 1e-06,
      "loss": 0.0869,
      "num_tokens": 1231778640.0,
      "reward": 0.43440812826156616,
      "reward_std": 0.034479107707738876,
      "rewards/progression_diversity/mean": -0.00694271782413125,
      "rewards/progression_diversity/std": 0.04713256284594536,
      "rewards/symbolic_reward_accuracy/mean": 0.326171875,
      "rewards/symbolic_reward_accuracy/std": 0.4692695140838623,
      "rewards/symbolic_reward_partial_score/mean": 0.7978678941726685,
      "rewards/symbolic_reward_partial_score/std": 0.1925002634525299,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.036149501800537,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 17.656431198120117,
      "step": 2041
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.29258349537849426,
      "epoch": 3.2724358974358974,
      "grad_norm": 0.03969258442521095,
      "learning_rate": 1e-06,
      "loss": -0.0049,
      "step": 2042
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2861692011356354,
      "epoch": 3.2740384615384617,
      "grad_norm": 0.007797840982675552,
      "learning_rate": 1e-06,
      "loss": 0.0038,
      "step": 2043
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2861800342798233,
      "epoch": 3.2756410256410255,
      "grad_norm": 0.024233318865299225,
      "learning_rate": 1e-06,
      "loss": 0.0363,
      "step": 2044
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2312.0,
      "completions/mean_length": 1453.421875,
      "completions/mean_terminated_length": 1335.8582763671875,
      "completions/min_length": 547.0,
      "completions/min_terminated_length": 547.0,
      "entropy": 0.29494452476501465,
      "epoch": 3.27724358974359,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 475.1607971191406,
      "learning_rate": 1e-06,
      "loss": 0.0204,
      "num_tokens": 1233400632.0,
      "reward": 0.3379926085472107,
      "reward_std": 0.04110347479581833,
      "rewards/progression_diversity/mean": -0.00249581434763968,
      "rewards/progression_diversity/std": 0.027295051142573357,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.7068033814430237,
      "rewards/symbolic_reward_partial_score/std": 0.20671771466732025,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.05837082862854,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 6.645764350891113,
      "step": 2045
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.29623962938785553,
      "epoch": 3.2788461538461537,
      "grad_norm": 0.015234522521495819,
      "learning_rate": 1e-06,
      "loss": 0.0231,
      "step": 2046
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2966182827949524,
      "epoch": 3.280448717948718,
      "grad_norm": 0.028103666380047798,
      "learning_rate": 1e-06,
      "loss": 0.0202,
      "step": 2047
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.29752831161022186,
      "epoch": 3.282051282051282,
      "grad_norm": 0.016002511605620384,
      "learning_rate": 1e-06,
      "loss": 0.0205,
      "step": 2048
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2178.0,
      "completions/mean_length": 1659.55859375,
      "completions/mean_terminated_length": 1396.099365234375,
      "completions/min_length": 734.0,
      "completions/min_terminated_length": 734.0,
      "entropy": 0.3017704486846924,
      "epoch": 3.2836538461538463,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 0.022180970758199692,
      "learning_rate": 1e-06,
      "loss": 0.0183,
      "num_tokens": 1235048998.0,
      "reward": 0.34624069929122925,
      "reward_std": 0.03489365801215172,
      "rewards/progression_diversity/mean": -0.005324858706444502,
      "rewards/progression_diversity/std": 0.0404798686504364,
      "rewards/symbolic_reward_accuracy/mean": 0.20703125,
      "rewards/symbolic_reward_accuracy/std": 0.40557438135147095,
      "rewards/symbolic_reward_partial_score/mean": 0.7448079586029053,
      "rewards/symbolic_reward_partial_score/std": 0.2022588700056076,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0514497756958008,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 10.927322387695312,
      "step": 2049
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.09375,
      "entropy": 0.3079407513141632,
      "epoch": 3.28525641025641,
      "grad_norm": 1578.2557373046875,
      "learning_rate": 1e-06,
      "loss": 0.112,
      "step": 2050
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.29569441080093384,
      "epoch": 3.2868589743589745,
      "grad_norm": 0.012577169574797153,
      "learning_rate": 1e-06,
      "loss": 0.0704,
      "step": 2051
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.2986031025648117,
      "epoch": 3.2884615384615383,
      "grad_norm": 0.0136114452034235,
      "learning_rate": 1e-06,
      "loss": 0.0306,
      "step": 2052
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2134.0,
      "completions/mean_length": 1793.57421875,
      "completions/mean_terminated_length": 1443.404052734375,
      "completions/min_length": 413.0,
      "completions/min_terminated_length": 413.0,
      "entropy": 0.306133508682251,
      "epoch": 3.2900641025641026,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 395.223876953125,
      "learning_rate": 1e-06,
      "loss": 0.0256,
      "num_tokens": 1236814236.0,
      "reward": 0.4016110897064209,
      "reward_std": 0.042516425251960754,
      "rewards/progression_diversity/mean": -0.007838626392185688,
      "rewards/progression_diversity/std": 0.05088994279503822,
      "rewards/symbolic_reward_accuracy/mean": 0.267578125,
      "rewards/symbolic_reward_accuracy/std": 0.4431293308734894,
      "rewards/symbolic_reward_partial_score/mean": 0.8090169429779053,
      "rewards/symbolic_reward_partial_score/std": 0.18975462019443512,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0399129390716553,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 16.826345443725586,
      "step": 2053
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.304185152053833,
      "epoch": 3.2916666666666665,
      "grad_norm": 0.020139280706644058,
      "learning_rate": 1e-06,
      "loss": 0.0203,
      "step": 2054
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.28031550347805023,
      "epoch": 3.293269230769231,
      "grad_norm": 0.008938499726355076,
      "learning_rate": 1e-06,
      "loss": 0.2255,
      "step": 2055
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.29509237408638,
      "epoch": 3.2948717948717947,
      "grad_norm": 0.008660328574478626,
      "learning_rate": 1e-06,
      "loss": 0.0369,
      "step": 2056
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2419.0,
      "completions/mean_length": 1667.703125,
      "completions/mean_terminated_length": 1434.1112060546875,
      "completions/min_length": 891.0,
      "completions/min_terminated_length": 891.0,
      "entropy": 0.2912152111530304,
      "epoch": 3.296474358974359,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 549.008056640625,
      "learning_rate": 1e-06,
      "loss": 0.0571,
      "num_tokens": 1238580100.0,
      "reward": 0.28469109535217285,
      "reward_std": 0.039379484951496124,
      "rewards/progression_diversity/mean": -0.004526130389422178,
      "rewards/progression_diversity/std": 0.03515040501952171,
      "rewards/symbolic_reward_accuracy/mean": 0.119140625,
      "rewards/symbolic_reward_accuracy/std": 0.32427072525024414,
      "rewards/symbolic_reward_partial_score/mean": 0.7153971195220947,
      "rewards/symbolic_reward_partial_score/std": 0.1853335201740265,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0521401166915894,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 10.635208129882812,
      "step": 2057
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2999632656574249,
      "epoch": 3.298076923076923,
      "grad_norm": 0.015014038421213627,
      "learning_rate": 1e-06,
      "loss": 0.0142,
      "step": 2058
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.2982807457447052,
      "epoch": 3.2996794871794872,
      "grad_norm": 0.021116536110639572,
      "learning_rate": 1e-06,
      "loss": 0.0415,
      "step": 2059
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.30165646970272064,
      "epoch": 3.301282051282051,
      "grad_norm": 0.02505728229880333,
      "learning_rate": 1e-06,
      "loss": 0.0484,
      "step": 2060
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2724.0,
      "completions/mean_length": 1967.10546875,
      "completions/mean_terminated_length": 1441.7935791015625,
      "completions/min_length": 698.0,
      "completions/min_terminated_length": 698.0,
      "entropy": 0.28198811411857605,
      "epoch": 3.3028846153846154,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 2657.8173828125,
      "learning_rate": 1e-06,
      "loss": 0.1443,
      "num_tokens": 1240427130.0,
      "reward": 0.38248294591903687,
      "reward_std": 0.058625657111406326,
      "rewards/progression_diversity/mean": -0.010982503183186054,
      "rewards/progression_diversity/std": 0.05810944363474846,
      "rewards/symbolic_reward_accuracy/mean": 0.271484375,
      "rewards/symbolic_reward_accuracy/std": 0.44516023993492126,
      "rewards/symbolic_reward_partial_score/mean": 0.7408040761947632,
      "rewards/symbolic_reward_partial_score/std": 0.22931715846061707,
      "rewards/tag_count_reward/mean": -0.025390625,
      "rewards/tag_count_reward/std": 0.15746226906776428,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0257844924926758,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 24.58890151977539,
      "step": 2061
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.305512472987175,
      "epoch": 3.3044871794871793,
      "grad_norm": 0.010081891901791096,
      "learning_rate": 1e-06,
      "loss": 0.0085,
      "step": 2062
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.30105258524417877,
      "epoch": 3.3060897435897436,
      "grad_norm": 0.019923444837331772,
      "learning_rate": 1e-06,
      "loss": 0.0702,
      "step": 2063
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.2942555248737335,
      "epoch": 3.3076923076923075,
      "grad_norm": 0.01970398612320423,
      "learning_rate": 1e-06,
      "loss": 0.0693,
      "step": 2064
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2750.0,
      "completions/mean_length": 1714.302734375,
      "completions/mean_terminated_length": 1481.450439453125,
      "completions/min_length": 776.0,
      "completions/min_terminated_length": 776.0,
      "entropy": 0.3109530508518219,
      "epoch": 3.309294871794872,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.02412906475365162,
      "learning_rate": 1e-06,
      "loss": 0.0006,
      "num_tokens": 1242207125.0,
      "reward": 0.36652231216430664,
      "reward_std": 0.056298792362213135,
      "rewards/progression_diversity/mean": -0.005487251095473766,
      "rewards/progression_diversity/std": 0.04296133294701576,
      "rewards/symbolic_reward_accuracy/mean": 0.25,
      "rewards/symbolic_reward_accuracy/std": 0.43343618512153625,
      "rewards/symbolic_reward_partial_score/mean": 0.7264810800552368,
      "rewards/symbolic_reward_partial_score/std": 0.2101718783378601,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0529332160949707,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 11.437311172485352,
      "step": 2065
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.2926006019115448,
      "epoch": 3.310897435897436,
      "grad_norm": 0.01898755133152008,
      "learning_rate": 1e-06,
      "loss": 0.0776,
      "step": 2066
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.3063723295927048,
      "epoch": 3.3125,
      "grad_norm": 0.014987506903707981,
      "learning_rate": 1e-06,
      "loss": 0.0091,
      "step": 2067
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.29933975636959076,
      "epoch": 3.314102564102564,
      "grad_norm": 0.020037300884723663,
      "learning_rate": 1e-06,
      "loss": 0.0595,
      "step": 2068
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2365.0,
      "completions/mean_length": 1686.85546875,
      "completions/mean_terminated_length": 1453.5675048828125,
      "completions/min_length": 838.0,
      "completions/min_terminated_length": 838.0,
      "entropy": 0.3088431656360626,
      "epoch": 3.315705128205128,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 310.1370544433594,
      "learning_rate": 1e-06,
      "loss": 0.0196,
      "num_tokens": 1243928939.0,
      "reward": 0.328504741191864,
      "reward_std": 0.027106130495667458,
      "rewards/progression_diversity/mean": -0.005487233866006136,
      "rewards/progression_diversity/std": 0.04299917817115784,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.7319173216819763,
      "rewards/symbolic_reward_partial_score/std": 0.20580171048641205,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0549488067626953,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 10.485675811767578,
      "step": 2069
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.3059859573841095,
      "epoch": 3.3173076923076925,
      "grad_norm": 178.32826232910156,
      "learning_rate": 1e-06,
      "loss": 0.0184,
      "step": 2070
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.29551123082637787,
      "epoch": 3.3189102564102564,
      "grad_norm": 0.01883271336555481,
      "learning_rate": 1e-06,
      "loss": 0.0948,
      "step": 2071
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.30632631480693817,
      "epoch": 3.3205128205128207,
      "grad_norm": 0.015535010024905205,
      "learning_rate": 1e-06,
      "loss": 0.005,
      "step": 2072
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2583.0,
      "completions/mean_length": 1778.39453125,
      "completions/mean_terminated_length": 1487.4462890625,
      "completions/min_length": 668.0,
      "completions/min_terminated_length": 668.0,
      "entropy": 0.30395227670669556,
      "epoch": 3.3221153846153846,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 893.9067993164062,
      "learning_rate": 1e-06,
      "loss": 0.0203,
      "num_tokens": 1245691061.0,
      "reward": 0.36694616079330444,
      "reward_std": 0.03838976100087166,
      "rewards/progression_diversity/mean": -0.006556159816682339,
      "rewards/progression_diversity/std": 0.04625479876995087,
      "rewards/symbolic_reward_accuracy/mean": 0.24609375,
      "rewards/symbolic_reward_accuracy/std": 0.4311550557613373,
      "rewards/symbolic_reward_partial_score/mean": 0.7370442152023315,
      "rewards/symbolic_reward_partial_score/std": 0.21306481957435608,
      "rewards/tag_count_reward/mean": -0.017578125,
      "rewards/tag_count_reward/std": 0.13154059648513794,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0472426414489746,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 14.168490409851074,
      "step": 2073
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2978048026561737,
      "epoch": 3.323717948717949,
      "grad_norm": 10739.20703125,
      "learning_rate": 1e-06,
      "loss": 0.3728,
      "step": 2074
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.30055759847164154,
      "epoch": 3.3253205128205128,
      "grad_norm": 0.01299914252012968,
      "learning_rate": 1e-06,
      "loss": 0.0633,
      "step": 2075
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.30361442267894745,
      "epoch": 3.326923076923077,
      "grad_norm": 0.014193286187946796,
      "learning_rate": 1e-06,
      "loss": 0.0325,
      "step": 2076
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3344.0,
      "completions/mean_length": 1754.5703125,
      "completions/mean_terminated_length": 1463.1474609375,
      "completions/min_length": 561.0,
      "completions/min_terminated_length": 561.0,
      "entropy": 0.2972850501537323,
      "epoch": 3.328525641025641,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 378.9573974609375,
      "learning_rate": 1e-06,
      "loss": 0.0103,
      "num_tokens": 1247440089.0,
      "reward": 0.43345290422439575,
      "reward_std": 0.04101278632879257,
      "rewards/progression_diversity/mean": -0.005784530192613602,
      "rewards/progression_diversity/std": 0.04120200499892235,
      "rewards/symbolic_reward_accuracy/mean": 0.3359375,
      "rewards/symbolic_reward_accuracy/std": 0.4727790653705597,
      "rewards/symbolic_reward_partial_score/mean": 0.778369128704071,
      "rewards/symbolic_reward_partial_score/std": 0.21870289742946625,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.049149513244629,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 12.310261726379395,
      "step": 2077
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.29869405925273895,
      "epoch": 3.3301282051282053,
      "grad_norm": 0.006345598492771387,
      "learning_rate": 1e-06,
      "loss": 0.0603,
      "step": 2078
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2994239032268524,
      "epoch": 3.331730769230769,
      "grad_norm": 0.013875186443328857,
      "learning_rate": 1e-06,
      "loss": 0.0434,
      "step": 2079
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2941579520702362,
      "epoch": 3.3333333333333335,
      "grad_norm": 0.03364482894539833,
      "learning_rate": 1e-06,
      "loss": 0.0486,
      "step": 2080
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2217.0,
      "completions/mean_length": 1733.458984375,
      "completions/mean_terminated_length": 1471.322021484375,
      "completions/min_length": 824.0,
      "completions/min_terminated_length": 824.0,
      "entropy": 0.29007087647914886,
      "epoch": 3.3349358974358974,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 648.8250732421875,
      "learning_rate": 1e-06,
      "loss": 0.0486,
      "num_tokens": 1249261396.0,
      "reward": 0.32781529426574707,
      "reward_std": 0.039317332208156586,
      "rewards/progression_diversity/mean": -0.006068596616387367,
      "rewards/progression_diversity/std": 0.0454968586564064,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.7361490726470947,
      "rewards/symbolic_reward_partial_score/std": 0.19703876972198486,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.049651026725769,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 12.473758697509766,
      "step": 2081
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.29181039333343506,
      "epoch": 3.3365384615384617,
      "grad_norm": 1459.4117431640625,
      "learning_rate": 1e-06,
      "loss": 0.1145,
      "step": 2082
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.301235556602478,
      "epoch": 3.3381410256410255,
      "grad_norm": 0.025303874164819717,
      "learning_rate": 1e-06,
      "loss": -0.003,
      "step": 2083
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.29415562748908997,
      "epoch": 3.33974358974359,
      "grad_norm": 0.023084204643964767,
      "learning_rate": 1e-06,
      "loss": -0.0009,
      "step": 2084
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2563.0,
      "completions/mean_length": 1614.404296875,
      "completions/mean_terminated_length": 1439.270751953125,
      "completions/min_length": 747.0,
      "completions/min_terminated_length": 747.0,
      "entropy": 0.29135817289352417,
      "epoch": 3.3413461538461537,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 2127.327392578125,
      "learning_rate": 1e-06,
      "loss": 0.0704,
      "num_tokens": 1250959283.0,
      "reward": 0.36038675904273987,
      "reward_std": 0.021374184638261795,
      "rewards/progression_diversity/mean": -0.0038059065118432045,
      "rewards/progression_diversity/std": 0.03501441702246666,
      "rewards/symbolic_reward_accuracy/mean": 0.248046875,
      "rewards/symbolic_reward_accuracy/std": 0.4323015511035919,
      "rewards/symbolic_reward_partial_score/mean": 0.7085774540901184,
      "rewards/symbolic_reward_partial_score/std": 0.2277929186820984,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0553886890411377,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 9.732671737670898,
      "step": 2085
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.30615630745887756,
      "epoch": 3.342948717948718,
      "grad_norm": 0.018460825085639954,
      "learning_rate": 1e-06,
      "loss": 0.0185,
      "step": 2086
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.30728329718112946,
      "epoch": 3.344551282051282,
      "grad_norm": 0.009494633413851261,
      "learning_rate": 1e-06,
      "loss": 0.0159,
      "step": 2087
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.31128981709480286,
      "epoch": 3.3461538461538463,
      "grad_norm": 0.007844570092856884,
      "learning_rate": 1e-06,
      "loss": -0.0016,
      "step": 2088
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2454.0,
      "completions/mean_length": 1687.4375,
      "completions/mean_terminated_length": 1454.1588134765625,
      "completions/min_length": 707.0,
      "completions/min_terminated_length": 707.0,
      "entropy": 0.309854656457901,
      "epoch": 3.34775641025641,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.018283460289239883,
      "learning_rate": 1e-06,
      "loss": 0.0238,
      "num_tokens": 1252780163.0,
      "reward": 0.3683873414993286,
      "reward_std": 0.03664931654930115,
      "rewards/progression_diversity/mean": -0.0050177741795778275,
      "rewards/progression_diversity/std": 0.03996478021144867,
      "rewards/symbolic_reward_accuracy/mean": 0.25,
      "rewards/symbolic_reward_accuracy/std": 0.43343618512153625,
      "rewards/symbolic_reward_partial_score/mean": 0.7300781011581421,
      "rewards/symbolic_reward_partial_score/std": 0.2158641666173935,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0487573146820068,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 13.234875679016113,
      "step": 2089
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.2977482080459595,
      "epoch": 3.3493589743589745,
      "grad_norm": 3524.53466796875,
      "learning_rate": 1e-06,
      "loss": 0.2433,
      "step": 2090
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.30203960835933685,
      "epoch": 3.3509615384615383,
      "grad_norm": 0.010414248332381248,
      "learning_rate": 1e-06,
      "loss": 0.0241,
      "step": 2091
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.30492982268333435,
      "epoch": 3.3525641025641026,
      "grad_norm": 0.02597237005829811,
      "learning_rate": 1e-06,
      "loss": 0.0162,
      "step": 2092
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2489.0,
      "completions/mean_length": 1726.21484375,
      "completions/mean_terminated_length": 1463.9482421875,
      "completions/min_length": 914.0,
      "completions/min_terminated_length": 914.0,
      "entropy": 0.30012136697769165,
      "epoch": 3.3541666666666665,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 42.711727142333984,
      "learning_rate": 1e-06,
      "loss": 0.0384,
      "num_tokens": 1254551649.0,
      "reward": 0.33932000398635864,
      "reward_std": 0.04096568375825882,
      "rewards/progression_diversity/mean": -0.005503435619175434,
      "rewards/progression_diversity/std": 0.041275136172771454,
      "rewards/symbolic_reward_accuracy/mean": 0.203125,
      "rewards/symbolic_reward_accuracy/std": 0.4027182459831238,
      "rewards/symbolic_reward_partial_score/mean": 0.7276041507720947,
      "rewards/symbolic_reward_partial_score/std": 0.1983337253332138,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0483951568603516,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 13.683696746826172,
      "step": 2093
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3052857369184494,
      "epoch": 3.355769230769231,
      "grad_norm": 0.014036121778190136,
      "learning_rate": 1e-06,
      "loss": 0.0409,
      "step": 2094
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.304688423871994,
      "epoch": 3.3573717948717947,
      "grad_norm": 0.19088442623615265,
      "learning_rate": 1e-06,
      "loss": 0.0153,
      "step": 2095
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3058953136205673,
      "epoch": 3.358974358974359,
      "grad_norm": 0.011440307833254337,
      "learning_rate": 1e-06,
      "loss": 0.0268,
      "step": 2096
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2805.0,
      "completions/mean_length": 1805.677734375,
      "completions/mean_terminated_length": 1515.27294921875,
      "completions/min_length": 811.0,
      "completions/min_terminated_length": 811.0,
      "entropy": 0.3070167005062103,
      "epoch": 3.360576923076923,
      "frac_reward_zero_std": 0.65625,
      "grad_norm": 1267.1373291015625,
      "learning_rate": 1e-06,
      "loss": 0.0541,
      "num_tokens": 1256251692.0,
      "reward": 0.40323367714881897,
      "reward_std": 0.047273900359869,
      "rewards/progression_diversity/mean": -0.0057373507879674435,
      "rewards/progression_diversity/std": 0.041116341948509216,
      "rewards/symbolic_reward_accuracy/mean": 0.283203125,
      "rewards/symbolic_reward_accuracy/std": 0.4509948492050171,
      "rewards/symbolic_reward_partial_score/mean": 0.7844075560569763,
      "rewards/symbolic_reward_partial_score/std": 0.19797967374324799,
      "rewards/tag_count_reward/mean": -0.01953125,
      "rewards/tag_count_reward/std": 0.1385180652141571,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0512890815734863,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 12.443960189819336,
      "step": 2097
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.3034556806087494,
      "epoch": 3.3621794871794872,
      "grad_norm": 0.012835121713578701,
      "learning_rate": 1e-06,
      "loss": 0.0352,
      "step": 2098
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.3106335401535034,
      "epoch": 3.363782051282051,
      "grad_norm": 0.007598002441227436,
      "learning_rate": 1e-06,
      "loss": 0.0361,
      "step": 2099
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1171875,
      "entropy": 0.30908720195293427,
      "epoch": 3.3653846153846154,
      "grad_norm": 0.006704273633658886,
      "learning_rate": 1e-06,
      "loss": 0.0375,
      "step": 2100
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2538.0,
      "completions/mean_length": 1834.38671875,
      "completions/mean_terminated_length": 1485.196044921875,
      "completions/min_length": 867.0,
      "completions/min_terminated_length": 867.0,
      "entropy": 0.30240941047668457,
      "epoch": 3.3669871794871793,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 343.7887268066406,
      "learning_rate": 1e-06,
      "loss": 0.053,
      "num_tokens": 1258104818.0,
      "reward": 0.35596776008605957,
      "reward_std": 0.044219110161066055,
      "rewards/progression_diversity/mean": -0.008694911375641823,
      "rewards/progression_diversity/std": 0.0567040778696537,
      "rewards/symbolic_reward_accuracy/mean": 0.208984375,
      "rewards/symbolic_reward_accuracy/std": 0.40698084235191345,
      "rewards/symbolic_reward_partial_score/mean": 0.775390625,
      "rewards/symbolic_reward_partial_score/std": 0.20388461649417877,
      "rewards/tag_count_reward/mean": -0.01953125,
      "rewards/tag_count_reward/std": 0.1385180652141571,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0403454303741455,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 17.884170532226562,
      "step": 2101
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.30909933149814606,
      "epoch": 3.3685897435897436,
      "grad_norm": 0.010506085120141506,
      "learning_rate": 1e-06,
      "loss": 0.0457,
      "step": 2102
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.29592080414295197,
      "epoch": 3.3701923076923075,
      "grad_norm": 0.15130743384361267,
      "learning_rate": 1e-06,
      "loss": 0.0234,
      "step": 2103
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.29659251868724823,
      "epoch": 3.371794871794872,
      "grad_norm": 0.009021518751978874,
      "learning_rate": 1e-06,
      "loss": 0.0793,
      "step": 2104
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2437.0,
      "completions/mean_length": 1713.470703125,
      "completions/mean_terminated_length": 1510.1168212890625,
      "completions/min_length": 812.0,
      "completions/min_terminated_length": 812.0,
      "entropy": 0.3099695146083832,
      "epoch": 3.373397435897436,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 328.6873779296875,
      "learning_rate": 1e-06,
      "loss": 0.0443,
      "num_tokens": 1259801347.0,
      "reward": 0.3114762306213379,
      "reward_std": 0.029843464493751526,
      "rewards/progression_diversity/mean": -0.004722995683550835,
      "rewards/progression_diversity/std": 0.04021621122956276,
      "rewards/symbolic_reward_accuracy/mean": 0.16796875,
      "rewards/symbolic_reward_accuracy/std": 0.374204158782959,
      "rewards/symbolic_reward_partial_score/mean": 0.7057291269302368,
      "rewards/symbolic_reward_partial_score/std": 0.2080366462469101,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0551323890686035,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 11.337949752807617,
      "step": 2105
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.31999480724334717,
      "epoch": 3.375,
      "grad_norm": 0.016840625554323196,
      "learning_rate": 1e-06,
      "loss": 0.0714,
      "step": 2106
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.32205140590667725,
      "epoch": 3.376602564102564,
      "grad_norm": 0.014530826359987259,
      "learning_rate": 1e-06,
      "loss": -0.0033,
      "step": 2107
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3153335601091385,
      "epoch": 3.378205128205128,
      "grad_norm": 0.013645652681589127,
      "learning_rate": 1e-06,
      "loss": 0.0475,
      "step": 2108
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.025390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2217.0,
      "completions/mean_length": 1905.322265625,
      "completions/mean_terminated_length": 1528.122314453125,
      "completions/min_length": 677.0,
      "completions/min_terminated_length": 677.0,
      "entropy": 0.29995962977409363,
      "epoch": 3.3798076923076925,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1299.5802001953125,
      "learning_rate": 1e-06,
      "loss": 0.0862,
      "num_tokens": 1261670184.0,
      "reward": 0.3757146894931793,
      "reward_std": 0.06831994652748108,
      "rewards/progression_diversity/mean": -0.008122788742184639,
      "rewards/progression_diversity/std": 0.05085650831460953,
      "rewards/symbolic_reward_accuracy/mean": 0.267578125,
      "rewards/symbolic_reward_accuracy/std": 0.4431293308734894,
      "rewards/symbolic_reward_partial_score/mean": 0.7253092527389526,
      "rewards/symbolic_reward_partial_score/std": 0.23041704297065735,
      "rewards/tag_count_reward/mean": -0.0234375,
      "rewards/tag_count_reward/std": 0.15143637359142303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0357104539871216,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 20.614686965942383,
      "step": 2109
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.30187252163887024,
      "epoch": 3.3814102564102564,
      "grad_norm": 0.9350156188011169,
      "learning_rate": 1e-06,
      "loss": 0.033,
      "step": 2110
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.3068780303001404,
      "epoch": 3.3830128205128207,
      "grad_norm": 0.010901644825935364,
      "learning_rate": 1e-06,
      "loss": 0.0526,
      "step": 2111
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.29835619032382965,
      "epoch": 3.3846153846153846,
      "grad_norm": 16258.962890625,
      "learning_rate": 1e-06,
      "loss": 1.5092,
      "step": 2112
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2335.0,
      "completions/mean_length": 1839.962890625,
      "completions/mean_terminated_length": 1550.2410888671875,
      "completions/min_length": 633.0,
      "completions/min_terminated_length": 633.0,
      "entropy": 0.3106384575366974,
      "epoch": 3.386217948717949,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 982.4223022460938,
      "learning_rate": 1e-06,
      "loss": 0.0231,
      "num_tokens": 1263444485.0,
      "reward": 0.4431688189506531,
      "reward_std": 0.059487175196409225,
      "rewards/progression_diversity/mean": -0.005386716220527887,
      "rewards/progression_diversity/std": 0.03828057646751404,
      "rewards/symbolic_reward_accuracy/mean": 0.357421875,
      "rewards/symbolic_reward_accuracy/std": 0.4797092080116272,
      "rewards/symbolic_reward_partial_score/mean": 0.7671223878860474,
      "rewards/symbolic_reward_partial_score/std": 0.2305992692708969,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045996904373169,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 15.037046432495117,
      "step": 2113
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3080595135688782,
      "epoch": 3.3878205128205128,
      "grad_norm": 0.011529210954904556,
      "learning_rate": 1e-06,
      "loss": 0.0223,
      "step": 2114
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3010105937719345,
      "epoch": 3.389423076923077,
      "grad_norm": 2453.80712890625,
      "learning_rate": 1e-06,
      "loss": 0.1075,
      "step": 2115
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.29755890369415283,
      "epoch": 3.391025641025641,
      "grad_norm": 0.02169516310095787,
      "learning_rate": 1e-06,
      "loss": 0.0655,
      "step": 2116
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2517.0,
      "completions/mean_length": 1782.4921875,
      "completions/mean_terminated_length": 1550.7222900390625,
      "completions/min_length": 668.0,
      "completions/min_terminated_length": 668.0,
      "entropy": 0.3025133013725281,
      "epoch": 3.3926282051282053,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 774.7922973632812,
      "learning_rate": 1e-06,
      "loss": 0.0502,
      "num_tokens": 1265197169.0,
      "reward": 0.38823676109313965,
      "reward_std": 0.03497297316789627,
      "rewards/progression_diversity/mean": -0.0049418737180531025,
      "rewards/progression_diversity/std": 0.038801729679107666,
      "rewards/symbolic_reward_accuracy/mean": 0.26953125,
      "rewards/symbolic_reward_accuracy/std": 0.44415023922920227,
      "rewards/symbolic_reward_partial_score/mean": 0.7578287720680237,
      "rewards/symbolic_reward_partial_score/std": 0.20313479006290436,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0485944747924805,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 13.493077278137207,
      "step": 2117
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.31337393820285797,
      "epoch": 3.394230769230769,
      "grad_norm": 0.024296533316373825,
      "learning_rate": 1e-06,
      "loss": -0.0042,
      "step": 2118
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.30753542482852936,
      "epoch": 3.3958333333333335,
      "grad_norm": 0.02327045053243637,
      "learning_rate": 1e-06,
      "loss": -0.0024,
      "step": 2119
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.29232731461524963,
      "epoch": 3.3974358974358974,
      "grad_norm": 0.015478910878300667,
      "learning_rate": 1e-06,
      "loss": 0.0821,
      "step": 2120
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2356.0,
      "completions/mean_length": 1758.404296875,
      "completions/mean_terminated_length": 1526.2520751953125,
      "completions/min_length": 753.0,
      "completions/min_terminated_length": 753.0,
      "entropy": 0.2938903123140335,
      "epoch": 3.3990384615384617,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 500.75775146484375,
      "learning_rate": 1e-06,
      "loss": 0.0747,
      "num_tokens": 1267013136.0,
      "reward": 0.30194687843322754,
      "reward_std": 0.0503971166908741,
      "rewards/progression_diversity/mean": -0.005022912751883268,
      "rewards/progression_diversity/std": 0.04021916538476944,
      "rewards/symbolic_reward_accuracy/mean": 0.14453125,
      "rewards/symbolic_reward_accuracy/std": 0.35197147727012634,
      "rewards/symbolic_reward_partial_score/mean": 0.7201985716819763,
      "rewards/symbolic_reward_partial_score/std": 0.20605868101119995,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.051767349243164,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 11.300193786621094,
      "step": 2121
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.3104073405265808,
      "epoch": 3.4006410256410255,
      "grad_norm": 0.026015326380729675,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "step": 2122
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.30487897992134094,
      "epoch": 3.40224358974359,
      "grad_norm": 0.022471783682703972,
      "learning_rate": 1e-06,
      "loss": 0.0026,
      "step": 2123
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.299276664853096,
      "epoch": 3.4038461538461537,
      "grad_norm": 0.022479599341750145,
      "learning_rate": 1e-06,
      "loss": 0.0366,
      "step": 2124
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3687.0,
      "completions/mean_length": 1766.92578125,
      "completions/mean_terminated_length": 1564.3128662109375,
      "completions/min_length": 789.0,
      "completions/min_terminated_length": 789.0,
      "entropy": 0.30581988394260406,
      "epoch": 3.405448717948718,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.0528433695435524,
      "learning_rate": 1e-06,
      "loss": 0.037,
      "num_tokens": 1268731466.0,
      "reward": 0.3584764301776886,
      "reward_std": 0.056666046380996704,
      "rewards/progression_diversity/mean": -0.003432408208027482,
      "rewards/progression_diversity/std": 0.0318170040845871,
      "rewards/symbolic_reward_accuracy/mean": 0.234375,
      "rewards/symbolic_reward_accuracy/std": 0.42402184009552,
      "rewards/symbolic_reward_partial_score/mean": 0.7308430671691895,
      "rewards/symbolic_reward_partial_score/std": 0.21932339668273926,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0599393844604492,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 7.578122138977051,
      "step": 2125
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3097575902938843,
      "epoch": 3.407051282051282,
      "grad_norm": 0.023739317432045937,
      "learning_rate": 1e-06,
      "loss": 0.0415,
      "step": 2126
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.31125499308109283,
      "epoch": 3.4086538461538463,
      "grad_norm": 0.012953310273587704,
      "learning_rate": 1e-06,
      "loss": 0.0353,
      "step": 2127
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.31313419342041016,
      "epoch": 3.41025641025641,
      "grad_norm": 0.013461051508784294,
      "learning_rate": 1e-06,
      "loss": 0.0162,
      "step": 2128
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2428.0,
      "completions/mean_length": 1800.833984375,
      "completions/mean_terminated_length": 1539.9024658203125,
      "completions/min_length": 935.0,
      "completions/min_terminated_length": 935.0,
      "entropy": 0.3056839108467102,
      "epoch": 3.4118589743589745,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.035352922976017,
      "learning_rate": 1e-06,
      "loss": 0.0161,
      "num_tokens": 1270457349.0,
      "reward": 0.45632848143577576,
      "reward_std": 0.054405439645051956,
      "rewards/progression_diversity/mean": -0.005339231342077255,
      "rewards/progression_diversity/std": 0.04081280156970024,
      "rewards/symbolic_reward_accuracy/mean": 0.359375,
      "rewards/symbolic_reward_accuracy/std": 0.48028653860092163,
      "rewards/symbolic_reward_partial_score/mean": 0.8057780265808105,
      "rewards/symbolic_reward_partial_score/std": 0.20670460164546967,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.054563045501709,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 9.924196243286133,
      "step": 2129
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.29381316900253296,
      "epoch": 3.4134615384615383,
      "grad_norm": 11953.2880859375,
      "learning_rate": 1e-06,
      "loss": 0.3212,
      "step": 2130
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.2993968427181244,
      "epoch": 3.4150641025641026,
      "grad_norm": 0.016664206981658936,
      "learning_rate": 1e-06,
      "loss": 0.0118,
      "step": 2131
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0390625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.125,
      "entropy": 0.29864759743213654,
      "epoch": 3.4166666666666665,
      "grad_norm": 1193.7196044921875,
      "learning_rate": 1e-06,
      "loss": 0.0213,
      "step": 2132
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2587.0,
      "completions/mean_length": 2058.951171875,
      "completions/mean_terminated_length": 1506.8701171875,
      "completions/min_length": 912.0,
      "completions/min_terminated_length": 912.0,
      "entropy": 0.2732394188642502,
      "epoch": 3.418269230769231,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 527.1871337890625,
      "learning_rate": 1e-06,
      "loss": 0.0862,
      "num_tokens": 1272399164.0,
      "reward": 0.33772745728492737,
      "reward_std": 0.049203261733055115,
      "rewards/progression_diversity/mean": -0.010945793241262436,
      "rewards/progression_diversity/std": 0.0561109222471714,
      "rewards/symbolic_reward_accuracy/mean": 0.21484375,
      "rewards/symbolic_reward_accuracy/std": 0.4111155867576599,
      "rewards/symbolic_reward_partial_score/mean": 0.7062011957168579,
      "rewards/symbolic_reward_partial_score/std": 0.2347198873758316,
      "rewards/tag_count_reward/mean": -0.029296875,
      "rewards/tag_count_reward/std": 0.16880230605602264,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0237281322479248,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 24.222614288330078,
      "step": 2133
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2859821915626526,
      "epoch": 3.4198717948717947,
      "grad_norm": 0.02516080252826214,
      "learning_rate": 1e-06,
      "loss": -0.0053,
      "step": 2134
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2708042562007904,
      "epoch": 3.421474358974359,
      "grad_norm": 0.01243208535015583,
      "learning_rate": 1e-06,
      "loss": 0.077,
      "step": 2135
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.26599569618701935,
      "epoch": 3.423076923076923,
      "grad_norm": 0.008642677217721939,
      "learning_rate": 1e-06,
      "loss": 0.1548,
      "step": 2136
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2676.0,
      "completions/mean_length": 1775.25,
      "completions/mean_terminated_length": 1484.2391357421875,
      "completions/min_length": 803.0,
      "completions/min_terminated_length": 803.0,
      "entropy": 0.2837539613246918,
      "epoch": 3.4246794871794872,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 216.27127075195312,
      "learning_rate": 1e-06,
      "loss": 0.0381,
      "num_tokens": 1274171948.0,
      "reward": 0.28258827328681946,
      "reward_std": 0.03927141800522804,
      "rewards/progression_diversity/mean": -0.00484531931579113,
      "rewards/progression_diversity/std": 0.036202192306518555,
      "rewards/symbolic_reward_accuracy/mean": 0.12109375,
      "rewards/symbolic_reward_accuracy/std": 0.3265552520751953,
      "rewards/symbolic_reward_partial_score/mean": 0.7051432132720947,
      "rewards/symbolic_reward_partial_score/std": 0.19589273631572723,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0403019189834595,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 15.552392959594727,
      "step": 2137
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.27445077896118164,
      "epoch": 3.426282051282051,
      "grad_norm": 0.011864651925861835,
      "learning_rate": 1e-06,
      "loss": 0.093,
      "step": 2138
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2975247800350189,
      "epoch": 3.4278846153846154,
      "grad_norm": 0.012054034508764744,
      "learning_rate": 1e-06,
      "loss": -0.0002,
      "step": 2139
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.2823929637670517,
      "epoch": 3.4294871794871793,
      "grad_norm": 0.015474159270524979,
      "learning_rate": 1e-06,
      "loss": 0.0769,
      "step": 2140
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2296.0,
      "completions/mean_length": 1778.083984375,
      "completions/mean_terminated_length": 1516.7454833984375,
      "completions/min_length": 873.0,
      "completions/min_terminated_length": 873.0,
      "entropy": 0.28506043553352356,
      "epoch": 3.4310897435897436,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.018360108137130737,
      "learning_rate": 1e-06,
      "loss": -0.007,
      "num_tokens": 1275971479.0,
      "reward": 0.3357412815093994,
      "reward_std": 0.02756696194410324,
      "rewards/progression_diversity/mean": -0.004975297022610903,
      "rewards/progression_diversity/std": 0.03728037327528,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.7488607168197632,
      "rewards/symbolic_reward_partial_score/std": 0.1986795961856842,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0461030006408691,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 12.876171112060547,
      "step": 2141
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2805647701025009,
      "epoch": 3.4326923076923075,
      "grad_norm": 0.014698946848511696,
      "learning_rate": 1e-06,
      "loss": 0.0504,
      "step": 2142
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.286907359957695,
      "epoch": 3.434294871794872,
      "grad_norm": 0.020853351801633835,
      "learning_rate": 1e-06,
      "loss": 0.0685,
      "step": 2143
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.28968897461891174,
      "epoch": 3.435897435897436,
      "grad_norm": 0.010319734923541546,
      "learning_rate": 1e-06,
      "loss": 0.0058,
      "step": 2144
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2501.0,
      "completions/mean_length": 1784.212890625,
      "completions/mean_terminated_length": 1522.9840087890625,
      "completions/min_length": 875.0,
      "completions/min_terminated_length": 875.0,
      "entropy": 0.3004123270511627,
      "epoch": 3.4375,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 258.38421630859375,
      "learning_rate": 1e-06,
      "loss": 0.0078,
      "num_tokens": 1277659044.0,
      "reward": 0.4109523296356201,
      "reward_std": 0.08106425404548645,
      "rewards/progression_diversity/mean": -0.005354299675673246,
      "rewards/progression_diversity/std": 0.04184706509113312,
      "rewards/symbolic_reward_accuracy/mean": 0.306640625,
      "rewards/symbolic_reward_accuracy/std": 0.4615498185157776,
      "rewards/symbolic_reward_partial_score/mean": 0.7619466781616211,
      "rewards/symbolic_reward_partial_score/std": 0.23045581579208374,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0512118339538574,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 11.705215454101562,
      "step": 2145
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.29704779386520386,
      "epoch": 3.439102564102564,
      "grad_norm": 0.023789308965206146,
      "learning_rate": 1e-06,
      "loss": 0.0442,
      "step": 2146
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.29680609703063965,
      "epoch": 3.440705128205128,
      "grad_norm": 0.022177523002028465,
      "learning_rate": 1e-06,
      "loss": 0.038,
      "step": 2147
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.29306888580322266,
      "epoch": 3.4423076923076925,
      "grad_norm": 0.009412667714059353,
      "learning_rate": 1e-06,
      "loss": 0.058,
      "step": 2148
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2385.0,
      "completions/mean_length": 1976.712890625,
      "completions/mean_terminated_length": 1571.688720703125,
      "completions/min_length": 751.0,
      "completions/min_terminated_length": 751.0,
      "entropy": 0.2784651964902878,
      "epoch": 3.4439102564102564,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 62.723731994628906,
      "learning_rate": 1e-06,
      "loss": 0.0358,
      "num_tokens": 1279531361.0,
      "reward": 0.4271858334541321,
      "reward_std": 0.03608609363436699,
      "rewards/progression_diversity/mean": -0.008470406755805016,
      "rewards/progression_diversity/std": 0.05123213678598404,
      "rewards/symbolic_reward_accuracy/mean": 0.3359375,
      "rewards/symbolic_reward_accuracy/std": 0.4727790653705597,
      "rewards/symbolic_reward_partial_score/mean": 0.7549641728401184,
      "rewards/symbolic_reward_partial_score/std": 0.22292669117450714,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0334545373916626,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 19.96367835998535,
      "step": 2149
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.28350482881069183,
      "epoch": 3.4455128205128207,
      "grad_norm": 0.010959633626043797,
      "learning_rate": 1e-06,
      "loss": 0.07,
      "step": 2150
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2834584712982178,
      "epoch": 3.4471153846153846,
      "grad_norm": 0.014038468711078167,
      "learning_rate": 1e-06,
      "loss": 0.0019,
      "step": 2151
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.27503225207328796,
      "epoch": 3.448717948717949,
      "grad_norm": 0.03383004292845726,
      "learning_rate": 1e-06,
      "loss": 0.0429,
      "step": 2152
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2406.0,
      "completions/mean_length": 2020.60546875,
      "completions/mean_terminated_length": 1557.2701416015625,
      "completions/min_length": 758.0,
      "completions/min_terminated_length": 758.0,
      "entropy": 0.2737235426902771,
      "epoch": 3.4503205128205128,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 0.028786683455109596,
      "learning_rate": 1e-06,
      "loss": 0.0172,
      "num_tokens": 1281467799.0,
      "reward": 0.34394705295562744,
      "reward_std": 0.05187678337097168,
      "rewards/progression_diversity/mean": -0.009591124951839447,
      "rewards/progression_diversity/std": 0.05538209527730942,
      "rewards/symbolic_reward_accuracy/mean": 0.201171875,
      "rewards/symbolic_reward_accuracy/std": 0.4012683033943176,
      "rewards/symbolic_reward_partial_score/mean": 0.7490234375,
      "rewards/symbolic_reward_partial_score/std": 0.20375196635723114,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0315946340560913,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 20.09771728515625,
      "step": 2153
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2721341699361801,
      "epoch": 3.451923076923077,
      "grad_norm": 0.00952248927205801,
      "learning_rate": 1e-06,
      "loss": 0.1071,
      "step": 2154
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.27051886916160583,
      "epoch": 3.453525641025641,
      "grad_norm": 0.015221396461129189,
      "learning_rate": 1e-06,
      "loss": 0.0418,
      "step": 2155
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.26560334861278534,
      "epoch": 3.4551282051282053,
      "grad_norm": 0.010249440558254719,
      "learning_rate": 1e-06,
      "loss": 0.0637,
      "step": 2156
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2303.0,
      "completions/mean_length": 2162.498046875,
      "completions/mean_terminated_length": 1584.38818359375,
      "completions/min_length": 866.0,
      "completions/min_terminated_length": 866.0,
      "entropy": 0.266851544380188,
      "epoch": 3.456730769230769,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 609.803955078125,
      "learning_rate": 1e-06,
      "loss": 0.072,
      "num_tokens": 1283476326.0,
      "reward": 0.39281851053237915,
      "reward_std": 0.035197146236896515,
      "rewards/progression_diversity/mean": -0.013561811298131943,
      "rewards/progression_diversity/std": 0.06690501421689987,
      "rewards/symbolic_reward_accuracy/mean": 0.275390625,
      "rewards/symbolic_reward_accuracy/std": 0.44714778661727905,
      "rewards/symbolic_reward_partial_score/mean": 0.7610189318656921,
      "rewards/symbolic_reward_partial_score/std": 0.19838300347328186,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0253196954727173,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 22.992713928222656,
      "step": 2157
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.269494891166687,
      "epoch": 3.4583333333333335,
      "grad_norm": 0.024904318153858185,
      "learning_rate": 1e-06,
      "loss": 0.0599,
      "step": 2158
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.25990206003189087,
      "epoch": 3.4599358974358974,
      "grad_norm": 0.01417806651443243,
      "learning_rate": 1e-06,
      "loss": 0.0332,
      "step": 2159
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2642335593700409,
      "epoch": 3.4615384615384617,
      "grad_norm": 0.00956464372575283,
      "learning_rate": 1e-06,
      "loss": 0.0442,
      "step": 2160
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.04296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2432.0,
      "completions/mean_length": 2211.890625,
      "completions/mean_terminated_length": 1575.591796875,
      "completions/min_length": 823.0,
      "completions/min_terminated_length": 823.0,
      "entropy": 0.2689553052186966,
      "epoch": 3.4631410256410255,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 659.0357666015625,
      "learning_rate": 1e-06,
      "loss": 0.0529,
      "num_tokens": 1285388046.0,
      "reward": 0.28991633653640747,
      "reward_std": 0.031822673976421356,
      "rewards/progression_diversity/mean": -0.012760378420352936,
      "rewards/progression_diversity/std": 0.0614657998085022,
      "rewards/symbolic_reward_accuracy/mean": 0.12109375,
      "rewards/symbolic_reward_accuracy/std": 0.3265552520751953,
      "rewards/symbolic_reward_partial_score/mean": 0.7272297739982605,
      "rewards/symbolic_reward_partial_score/std": 0.18316437304019928,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.018622875213623,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 27.381484985351562,
      "step": 2161
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.26476864516735077,
      "epoch": 3.46474358974359,
      "grad_norm": 0.013940541073679924,
      "learning_rate": 1e-06,
      "loss": 0.1169,
      "step": 2162
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2731598913669586,
      "epoch": 3.4663461538461537,
      "grad_norm": 77.1856918334961,
      "learning_rate": 1e-06,
      "loss": 0.0693,
      "step": 2163
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2830631732940674,
      "epoch": 3.467948717948718,
      "grad_norm": 0.026339799165725708,
      "learning_rate": 1e-06,
      "loss": 0.0181,
      "step": 2164
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2373.0,
      "completions/mean_length": 2145.0546875,
      "completions/mean_terminated_length": 1566.2357177734375,
      "completions/min_length": 856.0,
      "completions/min_terminated_length": 856.0,
      "entropy": 0.27289319038391113,
      "epoch": 3.469551282051282,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 247.32151794433594,
      "learning_rate": 1e-06,
      "loss": 0.0502,
      "num_tokens": 1287364458.0,
      "reward": 0.4520633816719055,
      "reward_std": 0.041082873940467834,
      "rewards/progression_diversity/mean": -0.013879441656172276,
      "rewards/progression_diversity/std": 0.06810557842254639,
      "rewards/symbolic_reward_accuracy/mean": 0.3671875,
      "rewards/symbolic_reward_accuracy/std": 0.48250964283943176,
      "rewards/symbolic_reward_partial_score/mean": 0.7729655504226685,
      "rewards/symbolic_reward_partial_score/std": 0.21098491549491882,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.033386468887329,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 19.480430603027344,
      "step": 2165
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2645433694124222,
      "epoch": 3.4711538461538463,
      "grad_norm": 0.010487830266356468,
      "learning_rate": 1e-06,
      "loss": 0.081,
      "step": 2166
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2667945772409439,
      "epoch": 3.47275641025641,
      "grad_norm": 0.012948272749781609,
      "learning_rate": 1e-06,
      "loss": 0.0463,
      "step": 2167
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2764432579278946,
      "epoch": 3.4743589743589745,
      "grad_norm": 0.016345176845788956,
      "learning_rate": 1e-06,
      "loss": 0.0114,
      "step": 2168
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2373.0,
      "completions/mean_length": 2468.5,
      "completions/mean_terminated_length": 1540.800048828125,
      "completions/min_length": 993.0,
      "completions/min_terminated_length": 993.0,
      "entropy": 0.2626621425151825,
      "epoch": 3.4759615384615383,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 77.64374542236328,
      "learning_rate": 1e-06,
      "loss": 0.0235,
      "num_tokens": 1289573546.0,
      "reward": 0.23555369675159454,
      "reward_std": 0.02879556640982628,
      "rewards/progression_diversity/mean": -0.02080235444009304,
      "rewards/progression_diversity/std": 0.08193078637123108,
      "rewards/symbolic_reward_accuracy/mean": 0.05859375,
      "rewards/symbolic_reward_accuracy/std": 0.23509246110916138,
      "rewards/symbolic_reward_partial_score/mean": 0.6706380844116211,
      "rewards/symbolic_reward_partial_score/std": 0.17417360842227936,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0140269994735718,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 29.442203521728516,
      "step": 2169
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.25798384845256805,
      "epoch": 3.4775641025641026,
      "grad_norm": 9.645312309265137,
      "learning_rate": 1e-06,
      "loss": 0.0937,
      "step": 2170
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.25725966691970825,
      "epoch": 3.4791666666666665,
      "grad_norm": 0.025352254509925842,
      "learning_rate": 1e-06,
      "loss": 0.0417,
      "step": 2171
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.26060421764850616,
      "epoch": 3.480769230769231,
      "grad_norm": 0.012905375100672245,
      "learning_rate": 1e-06,
      "loss": 0.0643,
      "step": 2172
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2228.0,
      "completions/mean_length": 2100.515625,
      "completions/mean_terminated_length": 1550.0364990234375,
      "completions/min_length": 825.0,
      "completions/min_terminated_length": 825.0,
      "entropy": 0.2711750864982605,
      "epoch": 3.4823717948717947,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 3850.08984375,
      "learning_rate": 1e-06,
      "loss": 0.0719,
      "num_tokens": 1291521378.0,
      "reward": 0.2990831136703491,
      "reward_std": 0.02100459486246109,
      "rewards/progression_diversity/mean": -0.012100995518267155,
      "rewards/progression_diversity/std": 0.06185305863618851,
      "rewards/symbolic_reward_accuracy/mean": 0.123046875,
      "rewards/symbolic_reward_accuracy/std": 0.32881227135658264,
      "rewards/symbolic_reward_partial_score/mean": 0.7525553107261658,
      "rewards/symbolic_reward_partial_score/std": 0.1686214655637741,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.03285813331604,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 19.567764282226562,
      "step": 2173
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.26994919776916504,
      "epoch": 3.483974358974359,
      "grad_norm": 0.02370203286409378,
      "learning_rate": 1e-06,
      "loss": 0.039,
      "step": 2174
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.28110212087631226,
      "epoch": 3.485576923076923,
      "grad_norm": 0.02028624340891838,
      "learning_rate": 1e-06,
      "loss": 0.0025,
      "step": 2175
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2572477161884308,
      "epoch": 3.4871794871794872,
      "grad_norm": 0.007402379531413317,
      "learning_rate": 1e-06,
      "loss": 0.0899,
      "step": 2176
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2413.0,
      "completions/mean_length": 1678.212890625,
      "completions/mean_terminated_length": 1533.1854248046875,
      "completions/min_length": 979.0,
      "completions/min_terminated_length": 979.0,
      "entropy": 0.30332140624523163,
      "epoch": 3.488782051282051,
      "frac_reward_zero_std": 0.65625,
      "grad_norm": 0.013128817081451416,
      "learning_rate": 1e-06,
      "loss": 0.0003,
      "num_tokens": 1293154063.0,
      "reward": 0.4060882329940796,
      "reward_std": 0.014026455581188202,
      "rewards/progression_diversity/mean": -0.0029925985727459192,
      "rewards/progression_diversity/std": 0.02976119890809059,
      "rewards/symbolic_reward_accuracy/mean": 0.310546875,
      "rewards/symbolic_reward_accuracy/std": 0.46317005157470703,
      "rewards/symbolic_reward_partial_score/mean": 0.7332844734191895,
      "rewards/symbolic_reward_partial_score/std": 0.22508755326271057,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0598652362823486,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 6.740238666534424,
      "step": 2177
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1171875,
      "entropy": 0.2928498536348343,
      "epoch": 3.4903846153846154,
      "grad_norm": 1097.95703125,
      "learning_rate": 1e-06,
      "loss": 0.0812,
      "step": 2178
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.29989422857761383,
      "epoch": 3.4919871794871793,
      "grad_norm": 0.009113781154155731,
      "learning_rate": 1e-06,
      "loss": 0.0097,
      "step": 2179
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.29661744832992554,
      "epoch": 3.4935897435897436,
      "grad_norm": 0.041792068630456924,
      "learning_rate": 1e-06,
      "loss": 0.0276,
      "step": 2180
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2569.0,
      "completions/mean_length": 1740.89453125,
      "completions/mean_terminated_length": 1567.260986328125,
      "completions/min_length": 994.0,
      "completions/min_terminated_length": 994.0,
      "entropy": 0.28554295003414154,
      "epoch": 3.4951923076923075,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 380.84259033203125,
      "learning_rate": 1e-06,
      "loss": 0.0355,
      "num_tokens": 1294866057.0,
      "reward": 0.30031508207321167,
      "reward_std": 0.020207438617944717,
      "rewards/progression_diversity/mean": -0.004136052913963795,
      "rewards/progression_diversity/std": 0.03724845126271248,
      "rewards/symbolic_reward_accuracy/mean": 0.154296875,
      "rewards/symbolic_reward_accuracy/std": 0.36158639192581177,
      "rewards/symbolic_reward_partial_score/mean": 0.6932454109191895,
      "rewards/symbolic_reward_partial_score/std": 0.20394784212112427,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0567407608032227,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 7.55679988861084,
      "step": 2181
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.2933173179626465,
      "epoch": 3.496794871794872,
      "grad_norm": 0.02578102797269821,
      "learning_rate": 1e-06,
      "loss": 0.0043,
      "step": 2182
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2895757555961609,
      "epoch": 3.498397435897436,
      "grad_norm": 0.016278348863124847,
      "learning_rate": 1e-06,
      "loss": -0.004,
      "step": 2183
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.2855415791273117,
      "epoch": 3.5,
      "grad_norm": 0.013905326835811138,
      "learning_rate": 1e-06,
      "loss": 0.0135,
      "step": 2184
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2640.0,
      "completions/mean_length": 1967.490234375,
      "completions/mean_terminated_length": 1621.4940185546875,
      "completions/min_length": 1021.0,
      "completions/min_terminated_length": 1021.0,
      "entropy": 0.2863430380821228,
      "epoch": 3.501602564102564,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 103.70210266113281,
      "learning_rate": 1e-06,
      "loss": 0.0187,
      "num_tokens": 1296736836.0,
      "reward": 0.31307274103164673,
      "reward_std": 0.031172282993793488,
      "rewards/progression_diversity/mean": -0.0066928681917488575,
      "rewards/progression_diversity/std": 0.04427814856171608,
      "rewards/symbolic_reward_accuracy/mean": 0.162109375,
      "rewards/symbolic_reward_accuracy/std": 0.3689115643501282,
      "rewards/symbolic_reward_partial_score/mean": 0.7208821177482605,
      "rewards/symbolic_reward_partial_score/std": 0.19237685203552246,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0420305728912354,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 15.591737747192383,
      "step": 2185
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.2772316485643387,
      "epoch": 3.503205128205128,
      "grad_norm": 268.4742736816406,
      "learning_rate": 1e-06,
      "loss": 0.0571,
      "step": 2186
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.2852557599544525,
      "epoch": 3.5048076923076925,
      "grad_norm": 0.01805300824344158,
      "learning_rate": 1e-06,
      "loss": 0.0119,
      "step": 2187
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.288587287068367,
      "epoch": 3.5064102564102564,
      "grad_norm": 0.017689945176243782,
      "learning_rate": 1e-06,
      "loss": 0.0324,
      "step": 2188
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2846.0,
      "completions/mean_length": 1944.544921875,
      "completions/mean_terminated_length": 1686.184814453125,
      "completions/min_length": 1121.0,
      "completions/min_terminated_length": 1121.0,
      "entropy": 0.27069567143917084,
      "epoch": 3.5080128205128203,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 196.1943359375,
      "learning_rate": 1e-06,
      "loss": 0.0453,
      "num_tokens": 1298707867.0,
      "reward": 0.40786558389663696,
      "reward_std": 0.02084019035100937,
      "rewards/progression_diversity/mean": -0.004458786454051733,
      "rewards/progression_diversity/std": 0.03238670900464058,
      "rewards/symbolic_reward_accuracy/mean": 0.30859375,
      "rewards/symbolic_reward_accuracy/std": 0.4623647928237915,
      "rewards/symbolic_reward_partial_score/mean": 0.7438150644302368,
      "rewards/symbolic_reward_partial_score/std": 0.21741990745067596,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.052014708518982,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 9.943605422973633,
      "step": 2189
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.27081412076950073,
      "epoch": 3.5096153846153846,
      "grad_norm": 0.013844393193721771,
      "learning_rate": 1e-06,
      "loss": 0.3136,
      "step": 2190
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.2878330200910568,
      "epoch": 3.511217948717949,
      "grad_norm": 0.013577915728092194,
      "learning_rate": 1e-06,
      "loss": -0.0015,
      "step": 2191
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.27837416529655457,
      "epoch": 3.5128205128205128,
      "grad_norm": 0.009212753735482693,
      "learning_rate": 1e-06,
      "loss": 0.0124,
      "step": 2192
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3301.0,
      "completions/mean_length": 2025.873046875,
      "completions/mean_terminated_length": 1681.278076171875,
      "completions/min_length": 1087.0,
      "completions/min_terminated_length": 1087.0,
      "entropy": 0.277935653924942,
      "epoch": 3.5144230769230766,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 408.2825927734375,
      "learning_rate": 1e-06,
      "loss": 0.011,
      "num_tokens": 1300680634.0,
      "reward": 0.3115173280239105,
      "reward_std": 0.014852182939648628,
      "rewards/progression_diversity/mean": -0.005984636954963207,
      "rewards/progression_diversity/std": 0.039667367935180664,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.7267415523529053,
      "rewards/symbolic_reward_partial_score/std": 0.1806911826133728,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0477591753005981,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 12.26329231262207,
      "step": 2193
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.2869657427072525,
      "epoch": 3.516025641025641,
      "grad_norm": 0.009542958810925484,
      "learning_rate": 1e-06,
      "loss": 5.8881,
      "step": 2194
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2763095796108246,
      "epoch": 3.5176282051282053,
      "grad_norm": 68490.9453125,
      "learning_rate": 1e-06,
      "loss": 14.7655,
      "step": 2195
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.28788307309150696,
      "epoch": 3.519230769230769,
      "grad_norm": 0.022392556071281433,
      "learning_rate": 1e-06,
      "loss": 3.6531,
      "step": 2196
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2589.0,
      "completions/mean_length": 1910.25390625,
      "completions/mean_terminated_length": 1680.511962890625,
      "completions/min_length": 1155.0,
      "completions/min_terminated_length": 1155.0,
      "entropy": 0.2990610897541046,
      "epoch": 3.5208333333333335,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 134.34463500976562,
      "learning_rate": 1e-06,
      "loss": 0.0237,
      "num_tokens": 1302494364.0,
      "reward": 0.38631004095077515,
      "reward_std": 0.027447409927845,
      "rewards/progression_diversity/mean": -0.004252912942320108,
      "rewards/progression_diversity/std": 0.033583469688892365,
      "rewards/symbolic_reward_accuracy/mean": 0.275390625,
      "rewards/symbolic_reward_accuracy/std": 0.44714778661727905,
      "rewards/symbolic_reward_partial_score/mean": 0.7383626699447632,
      "rewards/symbolic_reward_partial_score/std": 0.19246533513069153,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0586574077606201,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 7.78310489654541,
      "step": 2197
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.30396056175231934,
      "epoch": 3.5224358974358974,
      "grad_norm": 0.016528572887182236,
      "learning_rate": 1e-06,
      "loss": -0.0008,
      "step": 2198
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.308067262172699,
      "epoch": 3.5240384615384617,
      "grad_norm": 0.01760723628103733,
      "learning_rate": 1e-06,
      "loss": -0.0059,
      "step": 2199
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.2869449555873871,
      "epoch": 3.5256410256410255,
      "grad_norm": 835913.125,
      "learning_rate": 1e-06,
      "loss": 89.1969,
      "step": 2200
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2777.0,
      "completions/mean_length": 1783.34375,
      "completions/mean_terminated_length": 1668.3779296875,
      "completions/min_length": 1084.0,
      "completions/min_terminated_length": 1084.0,
      "entropy": 0.3045912981033325,
      "epoch": 3.52724358974359,
      "frac_reward_zero_std": 0.65625,
      "grad_norm": 0.0315437950193882,
      "learning_rate": 1e-06,
      "loss": 0.0041,
      "num_tokens": 1304211468.0,
      "reward": 0.38467174768447876,
      "reward_std": 0.01725323125720024,
      "rewards/progression_diversity/mean": -0.002066644374281168,
      "rewards/progression_diversity/std": 0.023847082629799843,
      "rewards/symbolic_reward_accuracy/mean": 0.251953125,
      "rewards/symbolic_reward_accuracy/std": 0.43455907702445984,
      "rewards/symbolic_reward_partial_score/mean": 0.779052734375,
      "rewards/symbolic_reward_partial_score/std": 0.19922928512096405,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0659763813018799,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 4.61309814453125,
      "step": 2201
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.29725050926208496,
      "epoch": 3.5288461538461537,
      "grad_norm": 1866.8873291015625,
      "learning_rate": 1e-06,
      "loss": 0.1574,
      "step": 2202
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.2994164824485779,
      "epoch": 3.530448717948718,
      "grad_norm": 0.00927357655018568,
      "learning_rate": 1e-06,
      "loss": -0.0,
      "step": 2203
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1171875,
      "entropy": 0.30384716391563416,
      "epoch": 3.532051282051282,
      "grad_norm": 0.011953097768127918,
      "learning_rate": 1e-06,
      "loss": -0.0006,
      "step": 2204
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3051.0,
      "completions/mean_length": 1808.693359375,
      "completions/mean_terminated_length": 1635.8636474609375,
      "completions/min_length": 1000.0,
      "completions/min_terminated_length": 1000.0,
      "entropy": 0.2852201759815216,
      "epoch": 3.5336538461538463,
      "frac_reward_zero_std": 0.625,
      "grad_norm": 128.33778381347656,
      "learning_rate": 1e-06,
      "loss": 0.0167,
      "num_tokens": 1306004719.0,
      "reward": 0.3015816807746887,
      "reward_std": 0.009979461319744587,
      "rewards/progression_diversity/mean": -0.001987605821341276,
      "rewards/progression_diversity/std": 0.02013721875846386,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.7566406726837158,
      "rewards/symbolic_reward_partial_score/std": 0.17199444770812988,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0611894130706787,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 5.849099636077881,
      "step": 2205
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.2893652319908142,
      "epoch": 3.53525641025641,
      "grad_norm": 0.021172404289245605,
      "learning_rate": 1e-06,
      "loss": 0.0304,
      "step": 2206
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.30061855912208557,
      "epoch": 3.5368589743589745,
      "grad_norm": 0.007474346086382866,
      "learning_rate": 1e-06,
      "loss": -0.0081,
      "step": 2207
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.28687790036201477,
      "epoch": 3.5384615384615383,
      "grad_norm": 0.008998925797641277,
      "learning_rate": 1e-06,
      "loss": 0.063,
      "step": 2208
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2967.0,
      "completions/mean_length": 1647.38671875,
      "completions/mean_terminated_length": 1589.59619140625,
      "completions/min_length": 1078.0,
      "completions/min_terminated_length": 1078.0,
      "entropy": 0.2984195798635483,
      "epoch": 3.5400641025641026,
      "frac_reward_zero_std": 0.6875,
      "grad_norm": 0.022925743833184242,
      "learning_rate": 1e-06,
      "loss": 0.0077,
      "num_tokens": 1307736341.0,
      "reward": 0.3372430205345154,
      "reward_std": 0.02868266962468624,
      "rewards/progression_diversity/mean": -0.0008010210585780442,
      "rewards/progression_diversity/std": 0.012307359836995602,
      "rewards/symbolic_reward_accuracy/mean": 0.193359375,
      "rewards/symbolic_reward_accuracy/std": 0.39531853795051575,
      "rewards/symbolic_reward_partial_score/mean": 0.7381021976470947,
      "rewards/symbolic_reward_partial_score/std": 0.18334569036960602,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0685679912567139,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 2.327254056930542,
      "step": 2209
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.292475089430809,
      "epoch": 3.5416666666666665,
      "grad_norm": 0.010455112904310226,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "step": 2210
    },
    {
      "clip_ratio/high_max": 0.25,
      "clip_ratio/high_mean": 0.0234375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.078125,
      "entropy": 0.2838650196790695,
      "epoch": 3.543269230769231,
      "grad_norm": 0.0170154869556427,
      "learning_rate": 1e-06,
      "loss": 0.0223,
      "step": 2211
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.2912348359823227,
      "epoch": 3.5448717948717947,
      "grad_norm": 0.014277713373303413,
      "learning_rate": 1e-06,
      "loss": -0.001,
      "step": 2212
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2683.0,
      "completions/mean_length": 1732.833984375,
      "completions/mean_terminated_length": 1588.34521484375,
      "completions/min_length": 1037.0,
      "completions/min_terminated_length": 1037.0,
      "entropy": 0.29485444724559784,
      "epoch": 3.546474358974359,
      "frac_reward_zero_std": 0.625,
      "grad_norm": 0.012255196459591389,
      "learning_rate": 1e-06,
      "loss": 0.031,
      "num_tokens": 1309492480.0,
      "reward": 0.40414494276046753,
      "reward_std": 0.017114289104938507,
      "rewards/progression_diversity/mean": -0.002500717993825674,
      "rewards/progression_diversity/std": 0.025870388373732567,
      "rewards/symbolic_reward_accuracy/mean": 0.27734375,
      "rewards/symbolic_reward_accuracy/std": 0.4481254518032074,
      "rewards/symbolic_reward_partial_score/mean": 0.7931965589523315,
      "rewards/symbolic_reward_partial_score/std": 0.17990581691265106,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0640480518341064,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 4.959805011749268,
      "step": 2213
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.2949848175048828,
      "epoch": 3.5480769230769234,
      "grad_norm": 0.023037033155560493,
      "learning_rate": 1e-06,
      "loss": 0.0215,
      "step": 2214
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.2907959371805191,
      "epoch": 3.5496794871794872,
      "grad_norm": 0.017563870176672935,
      "learning_rate": 1e-06,
      "loss": 0.0405,
      "step": 2215
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1171875,
      "entropy": 0.29562048614025116,
      "epoch": 3.551282051282051,
      "grad_norm": 0.008448651060461998,
      "learning_rate": 1e-06,
      "loss": -0.0048,
      "step": 2216
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2515.0,
      "completions/mean_length": 1673.904296875,
      "completions/mean_terminated_length": 1528.8343505859375,
      "completions/min_length": 946.0,
      "completions/min_terminated_length": 946.0,
      "entropy": 0.2969505935907364,
      "epoch": 3.5528846153846154,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 0.011285161599516869,
      "learning_rate": 1e-06,
      "loss": -0.0043,
      "num_tokens": 1311276607.0,
      "reward": 0.37819382548332214,
      "reward_std": 0.02540595829486847,
      "rewards/progression_diversity/mean": -0.0028848343063145876,
      "rewards/progression_diversity/std": 0.028827061876654625,
      "rewards/symbolic_reward_accuracy/mean": 0.2421875,
      "rewards/symbolic_reward_accuracy/std": 0.42882615327835083,
      "rewards/symbolic_reward_partial_score/mean": 0.7763671875,
      "rewards/symbolic_reward_partial_score/std": 0.18114922940731049,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0641391277313232,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 5.106628894805908,
      "step": 2217
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.29557299613952637,
      "epoch": 3.5544871794871797,
      "grad_norm": 3402.72705078125,
      "learning_rate": 1e-06,
      "loss": 0.0937,
      "step": 2218
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.2919696867465973,
      "epoch": 3.5560897435897436,
      "grad_norm": 0.012350378558039665,
      "learning_rate": 1e-06,
      "loss": -0.0099,
      "step": 2219
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.28622542321681976,
      "epoch": 3.5576923076923075,
      "grad_norm": 0.01743849739432335,
      "learning_rate": 1e-06,
      "loss": 0.0084,
      "step": 2220
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2304.0,
      "completions/mean_length": 1677.265625,
      "completions/mean_terminated_length": 1532.228759765625,
      "completions/min_length": 989.0,
      "completions/min_terminated_length": 989.0,
      "entropy": 0.3065932095050812,
      "epoch": 3.559294871794872,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.021933453157544136,
      "learning_rate": 1e-06,
      "loss": -0.0006,
      "num_tokens": 1312952407.0,
      "reward": 0.42246508598327637,
      "reward_std": 0.04646175354719162,
      "rewards/progression_diversity/mean": -0.0020300918258726597,
      "rewards/progression_diversity/std": 0.02212829887866974,
      "rewards/symbolic_reward_accuracy/mean": 0.314453125,
      "rewards/symbolic_reward_accuracy/std": 0.4647517800331116,
      "rewards/symbolic_reward_partial_score/mean": 0.780029296875,
      "rewards/symbolic_reward_partial_score/std": 0.20408369600772858,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0639897584915161,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 5.478954792022705,
      "step": 2221
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.3041534572839737,
      "epoch": 3.560897435897436,
      "grad_norm": 0.019840234890580177,
      "learning_rate": 1e-06,
      "loss": 0.0237,
      "step": 2222
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.299197256565094,
      "epoch": 3.5625,
      "grad_norm": 0.015518907457590103,
      "learning_rate": 1e-06,
      "loss": 0.012,
      "step": 2223
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.29240258038043976,
      "epoch": 3.564102564102564,
      "grad_norm": 0.017431948333978653,
      "learning_rate": 1e-06,
      "loss": 0.069,
      "step": 2224
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2300.0,
      "completions/mean_length": 1671.21484375,
      "completions/mean_terminated_length": 1555.3660888671875,
      "completions/min_length": 881.0,
      "completions/min_terminated_length": 881.0,
      "entropy": 0.2979786843061447,
      "epoch": 3.565705128205128,
      "frac_reward_zero_std": 0.6875,
      "grad_norm": 362.736572265625,
      "learning_rate": 1e-06,
      "loss": 0.0251,
      "num_tokens": 1314672565.0,
      "reward": 0.35204920172691345,
      "reward_std": 0.017848532646894455,
      "rewards/progression_diversity/mean": -0.002111406996846199,
      "rewards/progression_diversity/std": 0.024554969742894173,
      "rewards/symbolic_reward_accuracy/mean": 0.21484375,
      "rewards/symbolic_reward_accuracy/std": 0.4111155867576599,
      "rewards/symbolic_reward_partial_score/mean": 0.7445312738418579,
      "rewards/symbolic_reward_partial_score/std": 0.18352799117565155,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0668435096740723,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 4.025793075561523,
      "step": 2225
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.29960186779499054,
      "epoch": 3.5673076923076925,
      "grad_norm": 0.014206199906766415,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "step": 2226
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.29641352593898773,
      "epoch": 3.5689102564102564,
      "grad_norm": 0.013473715633153915,
      "learning_rate": 1e-06,
      "loss": 0.0365,
      "step": 2227
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0859375,
      "entropy": 0.29800280928611755,
      "epoch": 3.5705128205128203,
      "grad_norm": 0.01775544323027134,
      "learning_rate": 1e-06,
      "loss": -0.0016,
      "step": 2228
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2511.0,
      "completions/mean_length": 1592.27734375,
      "completions/mean_terminated_length": 1534.2706298828125,
      "completions/min_length": 928.0,
      "completions/min_terminated_length": 928.0,
      "entropy": 0.2977886497974396,
      "epoch": 3.5721153846153846,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 1049.2088623046875,
      "learning_rate": 1e-06,
      "loss": 0.0501,
      "num_tokens": 1316468131.0,
      "reward": 0.3352065682411194,
      "reward_std": 0.040253110229969025,
      "rewards/progression_diversity/mean": -0.000830255274195224,
      "rewards/progression_diversity/std": 0.015343528240919113,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.7475911378860474,
      "rewards/symbolic_reward_partial_score/std": 0.184952974319458,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0706126689910889,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 2.3357362747192383,
      "step": 2229
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.2995491325855255,
      "epoch": 3.573717948717949,
      "grad_norm": 0.029800059273838997,
      "learning_rate": 1e-06,
      "loss": -0.0004,
      "step": 2230
    },
    {
      "clip_ratio/high_max": 0.25,
      "clip_ratio/high_mean": 0.03125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0859375,
      "entropy": 0.30069583654403687,
      "epoch": 3.5753205128205128,
      "grad_norm": 0.02240557037293911,
      "learning_rate": 1e-06,
      "loss": -0.0074,
      "step": 2231
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.29591603577136993,
      "epoch": 3.5769230769230766,
      "grad_norm": 0.020445875823497772,
      "learning_rate": 1e-06,
      "loss": 0.003,
      "step": 2232
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2225.0,
      "completions/max_terminated_length": 2225.0,
      "completions/mean_length": 1476.59765625,
      "completions/mean_terminated_length": 1476.59765625,
      "completions/min_length": 823.0,
      "completions/min_terminated_length": 823.0,
      "entropy": 0.3097764253616333,
      "epoch": 3.578525641025641,
      "frac_reward_zero_std": 0.90625,
      "grad_norm": 0.009613445959985256,
      "learning_rate": 1e-06,
      "loss": -0.0014,
      "num_tokens": 1318013077.0,
      "reward": 0.2948632836341858,
      "reward_std": 0.003713551675900817,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.7328775525093079,
      "rewards/symbolic_reward_partial_score/std": 0.16866321861743927,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0766642093658447,
      "sampling/importance_sampling_ratio/min": 5.8294663176639006e-05,
      "sampling/sampling_logp_difference/max": 9.75,
      "sampling/sampling_logp_difference/mean": 0.14255455136299133,
      "step": 2233
    },
    {
      "clip_ratio/high_max": 0.25,
      "clip_ratio/high_mean": 0.015625,
      "clip_ratio/low_mean": 0.0078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0234375,
      "entropy": 0.31367750465869904,
      "epoch": 3.5801282051282053,
      "grad_norm": 0.011548890732228756,
      "learning_rate": 1e-06,
      "loss": -0.0032,
      "step": 2234
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0234375,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.046875,
      "entropy": 0.3107854425907135,
      "epoch": 3.581730769230769,
      "grad_norm": 0.005681135691702366,
      "learning_rate": 1e-06,
      "loss": 0.0013,
      "step": 2235
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0234375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0703125,
      "entropy": 0.313510924577713,
      "epoch": 3.5833333333333335,
      "grad_norm": 0.0011630782391875982,
      "learning_rate": 1e-06,
      "loss": 0.0037,
      "step": 2236
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2139.0,
      "completions/max_terminated_length": 2139.0,
      "completions/mean_length": 1488.98828125,
      "completions/mean_terminated_length": 1488.98828125,
      "completions/min_length": 811.0,
      "completions/min_terminated_length": 811.0,
      "entropy": 0.3215675801038742,
      "epoch": 3.5849358974358974,
      "frac_reward_zero_std": 0.84375,
      "grad_norm": 0.008507279679179192,
      "learning_rate": 1e-06,
      "loss": -0.0003,
      "num_tokens": 1319585743.0,
      "reward": 0.3924804925918579,
      "reward_std": 0.01995375007390976,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.259765625,
      "rewards/symbolic_reward_accuracy/std": 0.4389347732067108,
      "rewards/symbolic_reward_partial_score/mean": 0.7887369990348816,
      "rewards/symbolic_reward_partial_score/std": 0.16583316028118134,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.078096866607666,
      "sampling/importance_sampling_ratio/min": 0.0001314402325078845,
      "sampling/sampling_logp_difference/max": 8.936958312988281,
      "sampling/sampling_logp_difference/mean": 0.14448994398117065,
      "step": 2237
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0859375,
      "entropy": 0.31759844720363617,
      "epoch": 3.5865384615384617,
      "grad_norm": 0.005916442256420851,
      "learning_rate": 1e-06,
      "loss": 0.0009,
      "step": 2238
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0234375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0625,
      "entropy": 0.31728270649909973,
      "epoch": 3.5881410256410255,
      "grad_norm": 0.01380041241645813,
      "learning_rate": 1e-06,
      "loss": -0.0004,
      "step": 2239
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0703125,
      "entropy": 0.3198378384113312,
      "epoch": 3.58974358974359,
      "grad_norm": 0.011921794153749943,
      "learning_rate": 1e-06,
      "loss": -0.0002,
      "step": 2240
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2938.0,
      "completions/mean_length": 1711.978515625,
      "completions/mean_terminated_length": 1508.60400390625,
      "completions/min_length": 959.0,
      "completions/min_terminated_length": 959.0,
      "entropy": 0.29430273175239563,
      "epoch": 3.5913461538461537,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 607.2608032226562,
      "learning_rate": 1e-06,
      "loss": 0.0939,
      "num_tokens": 1321497412.0,
      "reward": 0.3224715292453766,
      "reward_std": 0.018792927265167236,
      "rewards/progression_diversity/mean": -0.0033359499648213387,
      "rewards/progression_diversity/std": 0.029082374647259712,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.7052246332168579,
      "rewards/symbolic_reward_partial_score/std": 0.20231178402900696,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0549734830856323,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 11.582987785339355,
      "step": 2241
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.3087223768234253,
      "epoch": 3.592948717948718,
      "grad_norm": 0.012669588439166546,
      "learning_rate": 1e-06,
      "loss": -0.0031,
      "step": 2242
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3084537237882614,
      "epoch": 3.594551282051282,
      "grad_norm": 0.02636776864528656,
      "learning_rate": 1e-06,
      "loss": 0.0134,
      "step": 2243
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3076443672180176,
      "epoch": 3.5961538461538463,
      "grad_norm": 94.45758056640625,
      "learning_rate": 1e-06,
      "loss": 0.0224,
      "step": 2244
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3366.0,
      "completions/mean_length": 1752.544921875,
      "completions/mean_terminated_length": 1549.732666015625,
      "completions/min_length": 749.0,
      "completions/min_terminated_length": 749.0,
      "entropy": 0.31993138790130615,
      "epoch": 3.59775641025641,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 107.91487121582031,
      "learning_rate": 1e-06,
      "loss": 0.0304,
      "num_tokens": 1323181931.0,
      "reward": 0.44030338525772095,
      "reward_std": 0.029550693929195404,
      "rewards/progression_diversity/mean": -0.00335341296158731,
      "rewards/progression_diversity/std": 0.03010004572570324,
      "rewards/symbolic_reward_accuracy/mean": 0.3359375,
      "rewards/symbolic_reward_accuracy/std": 0.4727790653705597,
      "rewards/symbolic_reward_partial_score/mean": 0.7978678345680237,
      "rewards/symbolic_reward_partial_score/std": 0.19323518872261047,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.059898018836975,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 10.545293807983398,
      "step": 2245
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.3280244767665863,
      "epoch": 3.5993589743589745,
      "grad_norm": 0.02449607476592064,
      "learning_rate": 1e-06,
      "loss": 0.0009,
      "step": 2246
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.32079842686653137,
      "epoch": 3.6009615384615383,
      "grad_norm": 0.018285367637872696,
      "learning_rate": 1e-06,
      "loss": 0.059,
      "step": 2247
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.32340890169143677,
      "epoch": 3.6025641025641026,
      "grad_norm": 0.00783622357994318,
      "learning_rate": 1e-06,
      "loss": 0.0208,
      "step": 2248
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2539.0,
      "completions/mean_length": 1838.962890625,
      "completions/mean_terminated_length": 1578.713623046875,
      "completions/min_length": 827.0,
      "completions/min_terminated_length": 827.0,
      "entropy": 0.3316734880208969,
      "epoch": 3.6041666666666665,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 0.011173716746270657,
      "learning_rate": 1e-06,
      "loss": 0.0217,
      "num_tokens": 1324944408.0,
      "reward": 0.37416577339172363,
      "reward_std": 0.039710745215415955,
      "rewards/progression_diversity/mean": -0.004324798006564379,
      "rewards/progression_diversity/std": 0.033225029706954956,
      "rewards/symbolic_reward_accuracy/mean": 0.240234375,
      "rewards/symbolic_reward_accuracy/std": 0.4276435375213623,
      "rewards/symbolic_reward_partial_score/mean": 0.7694987058639526,
      "rewards/symbolic_reward_partial_score/std": 0.2098463922739029,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0569475889205933,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 12.69996452331543,
      "step": 2249
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.31876716017723083,
      "epoch": 3.605769230769231,
      "grad_norm": 0.010406982153654099,
      "learning_rate": 1e-06,
      "loss": 0.0613,
      "step": 2250
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.322706013917923,
      "epoch": 3.6073717948717947,
      "grad_norm": 8.13254451751709,
      "learning_rate": 1e-06,
      "loss": 0.0253,
      "step": 2251
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.109375,
      "entropy": 0.32810820639133453,
      "epoch": 3.608974358974359,
      "grad_norm": 0.506855845451355,
      "learning_rate": 1e-06,
      "loss": 0.0166,
      "step": 2252
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2847.0,
      "completions/mean_length": 1849.2734375,
      "completions/mean_terminated_length": 1589.2086181640625,
      "completions/min_length": 946.0,
      "completions/min_terminated_length": 946.0,
      "entropy": 0.3225103169679642,
      "epoch": 3.6105769230769234,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.011867878027260303,
      "learning_rate": 1e-06,
      "loss": -0.0036,
      "num_tokens": 1326873332.0,
      "reward": 0.24351388216018677,
      "reward_std": 0.017110900953412056,
      "rewards/progression_diversity/mean": -0.004082350060343742,
      "rewards/progression_diversity/std": 0.03328991308808327,
      "rewards/symbolic_reward_accuracy/mean": 0.0625,
      "rewards/symbolic_reward_accuracy/std": 0.2422981858253479,
      "rewards/symbolic_reward_partial_score/mean": 0.689453125,
      "rewards/symbolic_reward_partial_score/std": 0.1663842499256134,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0547860860824585,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 12.677425384521484,
      "step": 2253
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.31891578435897827,
      "epoch": 3.6121794871794872,
      "grad_norm": 250185.734375,
      "learning_rate": 1e-06,
      "loss": 26.996,
      "step": 2254
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.32373523712158203,
      "epoch": 3.613782051282051,
      "grad_norm": 46074924.0,
      "learning_rate": 1e-06,
      "loss": 2780.4724,
      "step": 2255
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.32337436079978943,
      "epoch": 3.6153846153846154,
      "grad_norm": 0.01320252288132906,
      "learning_rate": 1e-06,
      "loss": 2.3287,
      "step": 2256
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2528.0,
      "completions/mean_length": 1806.265625,
      "completions/mean_terminated_length": 1604.1981201171875,
      "completions/min_length": 986.0,
      "completions/min_terminated_length": 986.0,
      "entropy": 0.3251884877681732,
      "epoch": 3.6169871794871797,
      "frac_reward_zero_std": 0.6875,
      "grad_norm": 670.6405029296875,
      "learning_rate": 1e-06,
      "loss": 0.0407,
      "num_tokens": 1328659420.0,
      "reward": 0.4107644557952881,
      "reward_std": 0.033720072358846664,
      "rewards/progression_diversity/mean": -0.003146503819152713,
      "rewards/progression_diversity/std": 0.02718130685389042,
      "rewards/symbolic_reward_accuracy/mean": 0.28125,
      "rewards/symbolic_reward_accuracy/std": 0.45004892349243164,
      "rewards/symbolic_reward_partial_score/mean": 0.8081217408180237,
      "rewards/symbolic_reward_partial_score/std": 0.1699947714805603,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.063145637512207,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 9.102170944213867,
      "step": 2257
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.3352958559989929,
      "epoch": 3.6185897435897436,
      "grad_norm": 0.013571917079389095,
      "learning_rate": 1e-06,
      "loss": -0.009,
      "step": 2258
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.3343977779150009,
      "epoch": 3.6201923076923075,
      "grad_norm": 0.007707657292485237,
      "learning_rate": 1e-06,
      "loss": 0.0254,
      "step": 2259
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.3308611214160919,
      "epoch": 3.621794871794872,
      "grad_norm": 0.008745530620217323,
      "learning_rate": 1e-06,
      "loss": 0.0577,
      "step": 2260
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2865.0,
      "completions/mean_length": 1843.185546875,
      "completions/mean_terminated_length": 1612.3790283203125,
      "completions/min_length": 877.0,
      "completions/min_terminated_length": 877.0,
      "entropy": 0.32520605623722076,
      "epoch": 3.623397435897436,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 302.1920471191406,
      "learning_rate": 1e-06,
      "loss": 0.0245,
      "num_tokens": 1330441019.0,
      "reward": 0.3576836585998535,
      "reward_std": 0.02488100342452526,
      "rewards/progression_diversity/mean": -0.00360656064003706,
      "rewards/progression_diversity/std": 0.02981680817902088,
      "rewards/symbolic_reward_accuracy/mean": 0.216796875,
      "rewards/symbolic_reward_accuracy/std": 0.4124660789966583,
      "rewards/symbolic_reward_partial_score/mean": 0.7601073980331421,
      "rewards/symbolic_reward_partial_score/std": 0.19392609596252441,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0611423254013062,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 10.63503646850586,
      "step": 2261
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3336937874555588,
      "epoch": 3.625,
      "grad_norm": 0.02153988927602768,
      "learning_rate": 1e-06,
      "loss": 0.0082,
      "step": 2262
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.33279934525489807,
      "epoch": 3.626602564102564,
      "grad_norm": 0.03544102981686592,
      "learning_rate": 1e-06,
      "loss": -0.0076,
      "step": 2263
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.32690323889255524,
      "epoch": 3.628205128205128,
      "grad_norm": 0.018338177353143692,
      "learning_rate": 1e-06,
      "loss": 0.0575,
      "step": 2264
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3015.0,
      "completions/mean_length": 1828.125,
      "completions/mean_terminated_length": 1626.3604736328125,
      "completions/min_length": 1000.0,
      "completions/min_terminated_length": 1000.0,
      "entropy": 0.32338719069957733,
      "epoch": 3.6298076923076925,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 316.7826232910156,
      "learning_rate": 1e-06,
      "loss": 0.0849,
      "num_tokens": 1332244379.0,
      "reward": 0.2881593108177185,
      "reward_std": 0.0218312069773674,
      "rewards/progression_diversity/mean": -0.0034066529478877783,
      "rewards/progression_diversity/std": 0.03087565302848816,
      "rewards/symbolic_reward_accuracy/mean": 0.125,
      "rewards/symbolic_reward_accuracy/std": 0.3310423493385315,
      "rewards/symbolic_reward_partial_score/mean": 0.7152017951011658,
      "rewards/symbolic_reward_partial_score/std": 0.19993343949317932,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0661569833755493,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 8.5504789352417,
      "step": 2265
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.33843283355236053,
      "epoch": 3.6314102564102564,
      "grad_norm": 0.01766320690512657,
      "learning_rate": 1e-06,
      "loss": 0.0185,
      "step": 2266
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.332566499710083,
      "epoch": 3.6330128205128203,
      "grad_norm": 0.01683247648179531,
      "learning_rate": 1e-06,
      "loss": 0.3716,
      "step": 2267
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.334817498922348,
      "epoch": 3.6346153846153846,
      "grad_norm": 0.011439353227615356,
      "learning_rate": 1e-06,
      "loss": 0.0074,
      "step": 2268
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3406.0,
      "completions/mean_length": 1817.625,
      "completions/mean_terminated_length": 1644.9012451171875,
      "completions/min_length": 772.0,
      "completions/min_terminated_length": 772.0,
      "entropy": 0.3378443419933319,
      "epoch": 3.636217948717949,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 1454.8203125,
      "learning_rate": 1e-06,
      "loss": 0.0475,
      "num_tokens": 1334038011.0,
      "reward": 0.3225540518760681,
      "reward_std": 0.03539721295237541,
      "rewards/progression_diversity/mean": -0.0028966807294636965,
      "rewards/progression_diversity/std": 0.027032606303691864,
      "rewards/symbolic_reward_accuracy/mean": 0.181640625,
      "rewards/symbolic_reward_accuracy/std": 0.38592514395713806,
      "rewards/symbolic_reward_partial_score/mean": 0.7139486074447632,
      "rewards/symbolic_reward_partial_score/std": 0.20663875341415405,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0668741464614868,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 8.729013442993164,
      "step": 2269
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.32664966583251953,
      "epoch": 3.6378205128205128,
      "grad_norm": 0.023835686966776848,
      "learning_rate": 1e-06,
      "loss": 0.0819,
      "step": 2270
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3433900773525238,
      "epoch": 3.6394230769230766,
      "grad_norm": 0.015054881572723389,
      "learning_rate": 1e-06,
      "loss": -0.0076,
      "step": 2271
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.3428369015455246,
      "epoch": 3.641025641025641,
      "grad_norm": 0.014643248170614243,
      "learning_rate": 1e-06,
      "loss": -0.0061,
      "step": 2272
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2808.0,
      "completions/mean_length": 2124.025390625,
      "completions/mean_terminated_length": 1664.026123046875,
      "completions/min_length": 885.0,
      "completions/min_terminated_length": 885.0,
      "entropy": 0.32794930040836334,
      "epoch": 3.6426282051282053,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 337.78216552734375,
      "learning_rate": 1e-06,
      "loss": 0.064,
      "num_tokens": 1335932760.0,
      "reward": 0.40950584411621094,
      "reward_std": 0.062236420810222626,
      "rewards/progression_diversity/mean": -0.006936722435057163,
      "rewards/progression_diversity/std": 0.04011997580528259,
      "rewards/symbolic_reward_accuracy/mean": 0.302734375,
      "rewards/symbolic_reward_accuracy/std": 0.45989060401916504,
      "rewards/symbolic_reward_partial_score/mean": 0.7682454586029053,
      "rewards/symbolic_reward_partial_score/std": 0.2247115969657898,
      "rewards/tag_count_reward/mean": -0.025390625,
      "rewards/tag_count_reward/std": 0.15746226906776428,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0450644493103027,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 19.491121292114258,
      "step": 2273
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.3140611946582794,
      "epoch": 3.644230769230769,
      "grad_norm": 11.444162368774414,
      "learning_rate": 1e-06,
      "loss": 0.0813,
      "step": 2274
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3365636169910431,
      "epoch": 3.6458333333333335,
      "grad_norm": 0.028920264914631844,
      "learning_rate": 1e-06,
      "loss": 0.0278,
      "step": 2275
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3288937658071518,
      "epoch": 3.6474358974358974,
      "grad_norm": 0.01110448595136404,
      "learning_rate": 1e-06,
      "loss": 0.0625,
      "step": 2276
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2847.0,
      "completions/mean_length": 2268.736328125,
      "completions/mean_terminated_length": 1694.945068359375,
      "completions/min_length": 956.0,
      "completions/min_terminated_length": 956.0,
      "entropy": 0.3222566395998001,
      "epoch": 3.6490384615384617,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 1264.3094482421875,
      "learning_rate": 1e-06,
      "loss": 0.043,
      "num_tokens": 1337987217.0,
      "reward": 0.5130249261856079,
      "reward_std": 0.08471733331680298,
      "rewards/progression_diversity/mean": -0.01001398079097271,
      "rewards/progression_diversity/std": 0.0511719211935997,
      "rewards/symbolic_reward_accuracy/mean": 0.4453125,
      "rewards/symbolic_reward_accuracy/std": 0.49748632311820984,
      "rewards/symbolic_reward_partial_score/mean": 0.8282551765441895,
      "rewards/symbolic_reward_partial_score/std": 0.22973352670669556,
      "rewards/tag_count_reward/mean": -0.025390625,
      "rewards/tag_count_reward/std": 0.15746226906776428,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0400309562683105,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 21.84160614013672,
      "step": 2277
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3258931040763855,
      "epoch": 3.6506410256410255,
      "grad_norm": 0.021348005160689354,
      "learning_rate": 1e-06,
      "loss": 0.0619,
      "step": 2278
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.31560730934143066,
      "epoch": 3.65224358974359,
      "grad_norm": 0.010525341145694256,
      "learning_rate": 1e-06,
      "loss": 0.0952,
      "step": 2279
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3183016777038574,
      "epoch": 3.6538461538461537,
      "grad_norm": 0.01716870814561844,
      "learning_rate": 1e-06,
      "loss": 0.0683,
      "step": 2280
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.021484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2697.0,
      "completions/mean_length": 1967.486328125,
      "completions/mean_terminated_length": 1650.9560546875,
      "completions/min_length": 1010.0,
      "completions/min_terminated_length": 1010.0,
      "entropy": 0.3132941722869873,
      "epoch": 3.655448717948718,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1842.542236328125,
      "learning_rate": 1e-06,
      "loss": 0.1164,
      "num_tokens": 1339971770.0,
      "reward": 0.2741461396217346,
      "reward_std": 0.0371134914457798,
      "rewards/progression_diversity/mean": -0.006771244574338198,
      "rewards/progression_diversity/std": 0.048090219497680664,
      "rewards/symbolic_reward_accuracy/mean": 0.123046875,
      "rewards/symbolic_reward_accuracy/std": 0.32881227135658264,
      "rewards/symbolic_reward_partial_score/mean": 0.6731607913970947,
      "rewards/symbolic_reward_partial_score/std": 0.20429718494415283,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0507025718688965,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 15.552701950073242,
      "step": 2281
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.3280630111694336,
      "epoch": 3.657051282051282,
      "grad_norm": 0.013712570071220398,
      "learning_rate": 1e-06,
      "loss": 0.0099,
      "step": 2282
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.3313450366258621,
      "epoch": 3.6586538461538463,
      "grad_norm": 0.0131281279027462,
      "learning_rate": 1e-06,
      "loss": 0.0101,
      "step": 2283
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3225076198577881,
      "epoch": 3.66025641025641,
      "grad_norm": 0.02563142031431198,
      "learning_rate": 1e-06,
      "loss": 0.0724,
      "step": 2284
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3059.0,
      "completions/mean_length": 1943.8203125,
      "completions/mean_terminated_length": 1656.1673583984375,
      "completions/min_length": 718.0,
      "completions/min_terminated_length": 718.0,
      "entropy": 0.31988005340099335,
      "epoch": 3.6618589743589745,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 1075.149658203125,
      "learning_rate": 1e-06,
      "loss": 0.0841,
      "num_tokens": 1341844046.0,
      "reward": 0.3518640398979187,
      "reward_std": 0.06046595424413681,
      "rewards/progression_diversity/mean": -0.005980104673653841,
      "rewards/progression_diversity/std": 0.04314341023564339,
      "rewards/symbolic_reward_accuracy/mean": 0.23046875,
      "rewards/symbolic_reward_accuracy/std": 0.42154473066329956,
      "rewards/symbolic_reward_partial_score/mean": 0.7140950560569763,
      "rewards/symbolic_reward_partial_score/std": 0.21391168236732483,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0547668933868408,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 14.504049301147461,
      "step": 2285
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.33179762959480286,
      "epoch": 3.6634615384615383,
      "grad_norm": 0.028521860018372536,
      "learning_rate": 1e-06,
      "loss": 0.0208,
      "step": 2286
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.324516236782074,
      "epoch": 3.6650641025641026,
      "grad_norm": 0.011499549262225628,
      "learning_rate": 1e-06,
      "loss": 0.0164,
      "step": 2287
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3264227360486984,
      "epoch": 3.6666666666666665,
      "grad_norm": 1.3879690170288086,
      "learning_rate": 1e-06,
      "loss": 0.0003,
      "step": 2288
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2883.0,
      "completions/mean_length": 1695.943359375,
      "completions/mean_terminated_length": 1609.373291015625,
      "completions/min_length": 943.0,
      "completions/min_terminated_length": 943.0,
      "entropy": 0.336402028799057,
      "epoch": 3.668269230769231,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 696.1746826171875,
      "learning_rate": 1e-06,
      "loss": 0.0194,
      "num_tokens": 1343498033.0,
      "reward": 0.3187481164932251,
      "reward_std": 0.019451720640063286,
      "rewards/progression_diversity/mean": -0.0016582165844738483,
      "rewards/progression_diversity/std": 0.02263886295258999,
      "rewards/symbolic_reward_accuracy/mean": 0.150390625,
      "rewards/symbolic_reward_accuracy/std": 0.35780346393585205,
      "rewards/symbolic_reward_partial_score/mean": 0.7617676258087158,
      "rewards/symbolic_reward_partial_score/std": 0.14879107475280762,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.07216477394104,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 5.351583480834961,
      "step": 2289
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.334991991519928,
      "epoch": 3.6698717948717947,
      "grad_norm": 0.016783706843852997,
      "learning_rate": 1e-06,
      "loss": 0.0124,
      "step": 2290
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.3304772675037384,
      "epoch": 3.671474358974359,
      "grad_norm": 0.022302983328700066,
      "learning_rate": 1e-06,
      "loss": 0.0245,
      "step": 2291
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3341434746980667,
      "epoch": 3.6730769230769234,
      "grad_norm": 0.012253678403794765,
      "learning_rate": 1e-06,
      "loss": 0.0019,
      "step": 2292
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3012.0,
      "completions/mean_length": 1679.7109375,
      "completions/mean_terminated_length": 1593.0452880859375,
      "completions/min_length": 671.0,
      "completions/min_terminated_length": 671.0,
      "entropy": 0.32631611824035645,
      "epoch": 3.6746794871794872,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 701.1774291992188,
      "learning_rate": 1e-06,
      "loss": 0.0256,
      "num_tokens": 1345151549.0,
      "reward": 0.3470765948295593,
      "reward_std": 0.03532809019088745,
      "rewards/progression_diversity/mean": -0.0018145160283893347,
      "rewards/progression_diversity/std": 0.023674938827753067,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.737060546875,
      "rewards/symbolic_reward_partial_score/std": 0.2012006640434265,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0714975595474243,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 5.101583480834961,
      "step": 2293
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.33157244324684143,
      "epoch": 3.676282051282051,
      "grad_norm": 0.024722039699554443,
      "learning_rate": 1e-06,
      "loss": 0.0206,
      "step": 2294
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.3273026645183563,
      "epoch": 3.6778846153846154,
      "grad_norm": 0.015624015592038631,
      "learning_rate": 1e-06,
      "loss": 0.0245,
      "step": 2295
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.33001944422721863,
      "epoch": 3.6794871794871797,
      "grad_norm": 0.014608138240873814,
      "learning_rate": 1e-06,
      "loss": 0.0007,
      "step": 2296
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2430.0,
      "completions/mean_length": 1686.3046875,
      "completions/mean_terminated_length": 1541.3570556640625,
      "completions/min_length": 771.0,
      "completions/min_terminated_length": 771.0,
      "entropy": 0.33564648032188416,
      "epoch": 3.6810897435897436,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 0.027066614478826523,
      "learning_rate": 1e-06,
      "loss": 0.0002,
      "num_tokens": 1346826425.0,
      "reward": 0.43666860461235046,
      "reward_std": 0.019992386922240257,
      "rewards/progression_diversity/mean": -0.0030630123801529408,
      "rewards/progression_diversity/std": 0.031911518424749374,
      "rewards/symbolic_reward_accuracy/mean": 0.341796875,
      "rewards/symbolic_reward_accuracy/std": 0.4747757613658905,
      "rewards/symbolic_reward_partial_score/mean": 0.7746745347976685,
      "rewards/symbolic_reward_partial_score/std": 0.22795653343200684,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0644441843032837,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 8.418769836425781,
      "step": 2297
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.3229387104511261,
      "epoch": 3.6826923076923075,
      "grad_norm": 9548.986328125,
      "learning_rate": 1e-06,
      "loss": 0.9497,
      "step": 2298
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.31784258782863617,
      "epoch": 3.684294871794872,
      "grad_norm": 0.017409533262252808,
      "learning_rate": 1e-06,
      "loss": 0.0463,
      "step": 2299
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.32778775691986084,
      "epoch": 3.685897435897436,
      "grad_norm": 0.01562940888106823,
      "learning_rate": 1e-06,
      "loss": -0.0083,
      "step": 2300
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2580.0,
      "completions/mean_length": 1658.37109375,
      "completions/mean_terminated_length": 1542.4212646484375,
      "completions/min_length": 980.0,
      "completions/min_terminated_length": 980.0,
      "entropy": 0.3309648334980011,
      "epoch": 3.6875,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.03380941227078438,
      "learning_rate": 1e-06,
      "loss": -0.0012,
      "num_tokens": 1348522775.0,
      "reward": 0.25678491592407227,
      "reward_std": 0.02014780044555664,
      "rewards/progression_diversity/mean": -0.001687450218014419,
      "rewards/progression_diversity/std": 0.02322215586900711,
      "rewards/symbolic_reward_accuracy/mean": 0.087890625,
      "rewards/symbolic_reward_accuracy/std": 0.2834126651287079,
      "rewards/symbolic_reward_partial_score/mean": 0.6808756589889526,
      "rewards/symbolic_reward_partial_score/std": 0.177845299243927,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0702204704284668,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 5.331143379211426,
      "step": 2301
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.32819345593452454,
      "epoch": 3.689102564102564,
      "grad_norm": 297679.28125,
      "learning_rate": 1e-06,
      "loss": 8.1288,
      "step": 2302
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.3345276117324829,
      "epoch": 3.690705128205128,
      "grad_norm": 0.007232366129755974,
      "learning_rate": 1e-06,
      "loss": 0.0004,
      "step": 2303
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.32529085874557495,
      "epoch": 3.6923076923076925,
      "grad_norm": 31849.216796875,
      "learning_rate": 1e-06,
      "loss": 4.3377,
      "step": 2304
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2630.0,
      "completions/mean_length": 1738.244140625,
      "completions/mean_terminated_length": 1564.5791015625,
      "completions/min_length": 800.0,
      "completions/min_terminated_length": 800.0,
      "entropy": 0.3245372474193573,
      "epoch": 3.6939102564102564,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 0.024981984868645668,
      "learning_rate": 1e-06,
      "loss": 0.0394,
      "num_tokens": 1350288916.0,
      "reward": 0.2519104480743408,
      "reward_std": 0.015460798516869545,
      "rewards/progression_diversity/mean": -0.002804999705404043,
      "rewards/progression_diversity/std": 0.027330690994858742,
      "rewards/symbolic_reward_accuracy/mean": 0.0625,
      "rewards/symbolic_reward_accuracy/std": 0.2422981858253479,
      "rewards/symbolic_reward_partial_score/mean": 0.7173991203308105,
      "rewards/symbolic_reward_partial_score/std": 0.16349852085113525,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.06309175491333,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 9.228836059570312,
      "step": 2305
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.32626084983348846,
      "epoch": 3.6955128205128203,
      "grad_norm": 1247.531494140625,
      "learning_rate": 1e-06,
      "loss": 0.0508,
      "step": 2306
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.3255147635936737,
      "epoch": 3.6971153846153846,
      "grad_norm": 0.009068300016224384,
      "learning_rate": 1e-06,
      "loss": 0.5594,
      "step": 2307
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.125,
      "entropy": 0.32918526232242584,
      "epoch": 3.698717948717949,
      "grad_norm": 0.017367210239171982,
      "learning_rate": 1e-06,
      "loss": -0.0038,
      "step": 2308
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3004.0,
      "completions/mean_length": 1774.068359375,
      "completions/mean_terminated_length": 1600.828125,
      "completions/min_length": 883.0,
      "completions/min_terminated_length": 883.0,
      "entropy": 0.3189825713634491,
      "epoch": 3.7003205128205128,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.019421013072133064,
      "learning_rate": 1e-06,
      "loss": 0.0116,
      "num_tokens": 1352175175.0,
      "reward": 0.33472028374671936,
      "reward_std": 0.026918543502688408,
      "rewards/progression_diversity/mean": -0.0030723384115844965,
      "rewards/progression_diversity/std": 0.029028048738837242,
      "rewards/symbolic_reward_accuracy/mean": 0.208984375,
      "rewards/symbolic_reward_accuracy/std": 0.40698084235191345,
      "rewards/symbolic_reward_partial_score/mean": 0.6978678703308105,
      "rewards/symbolic_reward_partial_score/std": 0.19654740393161774,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0647794008255005,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 8.113250732421875,
      "step": 2309
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.32003454864025116,
      "epoch": 3.7019230769230766,
      "grad_norm": 0.031067878007888794,
      "learning_rate": 1e-06,
      "loss": 0.4607,
      "step": 2310
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.32604101300239563,
      "epoch": 3.703525641025641,
      "grad_norm": 7065.14501953125,
      "learning_rate": 1e-06,
      "loss": 0.5045,
      "step": 2311
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.32465484738349915,
      "epoch": 3.7051282051282053,
      "grad_norm": 0.008808803744614124,
      "learning_rate": 1e-06,
      "loss": 0.004,
      "step": 2312
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2823.0,
      "completions/mean_length": 1703.431640625,
      "completions/mean_terminated_length": 1645.86083984375,
      "completions/min_length": 877.0,
      "completions/min_terminated_length": 877.0,
      "entropy": 0.33033812046051025,
      "epoch": 3.706730769230769,
      "frac_reward_zero_std": 0.65625,
      "grad_norm": 0.029636235907673836,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "num_tokens": 1353921444.0,
      "reward": 0.3808506727218628,
      "reward_std": 0.02052554301917553,
      "rewards/progression_diversity/mean": -0.0008730281260795891,
      "rewards/progression_diversity/std": 0.01423015259206295,
      "rewards/symbolic_reward_accuracy/mean": 0.263671875,
      "rewards/symbolic_reward_accuracy/std": 0.4410543739795685,
      "rewards/symbolic_reward_partial_score/mean": 0.7421875,
      "rewards/symbolic_reward_partial_score/std": 0.20971913635730743,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0755610466003418,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 2.919158458709717,
      "step": 2313
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.3263317346572876,
      "epoch": 3.7083333333333335,
      "grad_norm": 0.009872586466372013,
      "learning_rate": 1e-06,
      "loss": 0.049,
      "step": 2314
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1171875,
      "entropy": 0.3279000520706177,
      "epoch": 3.7099358974358974,
      "grad_norm": 0.025471488013863564,
      "learning_rate": 1e-06,
      "loss": 0.0166,
      "step": 2315
    },
    {
      "clip_ratio/high_max": 0.25,
      "clip_ratio/high_mean": 0.0234375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.125,
      "entropy": 0.3354235142469406,
      "epoch": 3.7115384615384617,
      "grad_norm": 0.012011319398880005,
      "learning_rate": 1e-06,
      "loss": 0.0027,
      "step": 2316
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3046.0,
      "completions/mean_length": 2033.783203125,
      "completions/mean_terminated_length": 1747.92236328125,
      "completions/min_length": 1049.0,
      "completions/min_terminated_length": 1049.0,
      "entropy": 0.31582309305667877,
      "epoch": 3.7131410256410255,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 145.75550842285156,
      "learning_rate": 1e-06,
      "loss": 0.0422,
      "num_tokens": 1355863077.0,
      "reward": 0.37086308002471924,
      "reward_std": 0.03677962347865105,
      "rewards/progression_diversity/mean": -0.004513155668973923,
      "rewards/progression_diversity/std": 0.03423220291733742,
      "rewards/symbolic_reward_accuracy/mean": 0.244140625,
      "rewards/symbolic_reward_accuracy/std": 0.42999663949012756,
      "rewards/symbolic_reward_partial_score/mean": 0.7519856691360474,
      "rewards/symbolic_reward_partial_score/std": 0.19709475338459015,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0543160438537598,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 12.873621940612793,
      "step": 2317
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.3110759109258652,
      "epoch": 3.71474358974359,
      "grad_norm": 2783.682861328125,
      "learning_rate": 1e-06,
      "loss": 1.9094,
      "step": 2318
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.31409376859664917,
      "epoch": 3.7163461538461537,
      "grad_norm": 5945.8662109375,
      "learning_rate": 1e-06,
      "loss": 0.319,
      "step": 2319
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.32960157096385956,
      "epoch": 3.717948717948718,
      "grad_norm": 0.012837960384786129,
      "learning_rate": 1e-06,
      "loss": -0.0077,
      "step": 2320
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3660.0,
      "completions/mean_length": 2013.263671875,
      "completions/mean_terminated_length": 1814.0654296875,
      "completions/min_length": 973.0,
      "completions/min_terminated_length": 973.0,
      "entropy": 0.32310742139816284,
      "epoch": 3.719551282051282,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 137.2488250732422,
      "learning_rate": 1e-06,
      "loss": 0.0302,
      "num_tokens": 1357745916.0,
      "reward": 0.28083521127700806,
      "reward_std": 0.030796613544225693,
      "rewards/progression_diversity/mean": -0.002905955072492361,
      "rewards/progression_diversity/std": 0.02647358365356922,
      "rewards/symbolic_reward_accuracy/mean": 0.111328125,
      "rewards/symbolic_reward_accuracy/std": 0.31484565138816833,
      "rewards/symbolic_reward_partial_score/mean": 0.7168131470680237,
      "rewards/symbolic_reward_partial_score/std": 0.1985962688922882,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0641801357269287,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 7.660578727722168,
      "step": 2321
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.31349869072437286,
      "epoch": 3.7211538461538463,
      "grad_norm": 15555.6708984375,
      "learning_rate": 1e-06,
      "loss": 6.2007,
      "step": 2322
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.140625,
      "entropy": 0.3223974406719208,
      "epoch": 3.72275641025641,
      "grad_norm": 968024.4375,
      "learning_rate": 1e-06,
      "loss": 23.6673,
      "step": 2323
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.31636857986450195,
      "epoch": 3.7243589743589745,
      "grad_norm": 0.015157344751060009,
      "learning_rate": 1e-06,
      "loss": 0.0284,
      "step": 2324
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3458.0,
      "completions/mean_length": 2236.482421875,
      "completions/mean_terminated_length": 1838.760986328125,
      "completions/min_length": 874.0,
      "completions/min_terminated_length": 874.0,
      "entropy": 0.3094794303178787,
      "epoch": 3.7259615384615383,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 0.025273853912949562,
      "learning_rate": 1e-06,
      "loss": 0.0773,
      "num_tokens": 1359728803.0,
      "reward": 0.36245208978652954,
      "reward_std": 0.04594516381621361,
      "rewards/progression_diversity/mean": -0.005767707247287035,
      "rewards/progression_diversity/std": 0.036357343196868896,
      "rewards/symbolic_reward_accuracy/mean": 0.2421875,
      "rewards/symbolic_reward_accuracy/std": 0.42882615327835083,
      "rewards/symbolic_reward_partial_score/mean": 0.7291991710662842,
      "rewards/symbolic_reward_partial_score/std": 0.22089123725891113,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.050485372543335,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 14.873160362243652,
      "step": 2325
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.31531573832035065,
      "epoch": 3.7275641025641026,
      "grad_norm": 0.019093314185738564,
      "learning_rate": 1e-06,
      "loss": 0.0435,
      "step": 2326
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.3113931119441986,
      "epoch": 3.7291666666666665,
      "grad_norm": 0.015845902264118195,
      "learning_rate": 1e-06,
      "loss": 0.0502,
      "step": 2327
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.3193650245666504,
      "epoch": 3.730769230769231,
      "grad_norm": 0.1452263742685318,
      "learning_rate": 1e-06,
      "loss": 0.0405,
      "step": 2328
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3291.0,
      "completions/mean_length": 2277.3984375,
      "completions/mean_terminated_length": 1880.8272705078125,
      "completions/min_length": 1066.0,
      "completions/min_terminated_length": 1066.0,
      "entropy": 0.32257337868213654,
      "epoch": 3.7323717948717947,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 67.19734954833984,
      "learning_rate": 1e-06,
      "loss": -0.0058,
      "num_tokens": 1361753231.0,
      "reward": 0.4234715700149536,
      "reward_std": 0.052917372435331345,
      "rewards/progression_diversity/mean": -0.00587063655257225,
      "rewards/progression_diversity/std": 0.03647465631365776,
      "rewards/symbolic_reward_accuracy/mean": 0.328125,
      "rewards/symbolic_reward_accuracy/std": 0.4699897766113281,
      "rewards/symbolic_reward_partial_score/mean": 0.7568196654319763,
      "rewards/symbolic_reward_partial_score/std": 0.2349434494972229,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.053420066833496,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 13.831912994384766,
      "step": 2329
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.3165089935064316,
      "epoch": 3.733974358974359,
      "grad_norm": 1270778.375,
      "learning_rate": 1e-06,
      "loss": 109.9686,
      "step": 2330
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.31606537103652954,
      "epoch": 3.7355769230769234,
      "grad_norm": 241028.328125,
      "learning_rate": 1e-06,
      "loss": 29.3297,
      "step": 2331
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.31982485949993134,
      "epoch": 3.7371794871794872,
      "grad_norm": 5490.81640625,
      "learning_rate": 1e-06,
      "loss": 1.0961,
      "step": 2332
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.052734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3391.0,
      "completions/mean_length": 2758.09375,
      "completions/mean_terminated_length": 1999.5382080078125,
      "completions/min_length": 779.0,
      "completions/min_terminated_length": 779.0,
      "entropy": 0.3020293414592743,
      "epoch": 3.738782051282051,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 2158.609375,
      "learning_rate": 1e-06,
      "loss": 0.0567,
      "num_tokens": 1364147151.0,
      "reward": 0.28579181432724,
      "reward_std": 0.057038046419620514,
      "rewards/progression_diversity/mean": -0.01163883414119482,
      "rewards/progression_diversity/std": 0.052136827260255814,
      "rewards/symbolic_reward_accuracy/mean": 0.140625,
      "rewards/symbolic_reward_accuracy/std": 0.3479743003845215,
      "rewards/symbolic_reward_partial_score/mean": 0.6769856810569763,
      "rewards/symbolic_reward_partial_score/std": 0.21585947275161743,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0365989208221436,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 21.644445419311523,
      "step": 2333
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.29430215060710907,
      "epoch": 3.7403846153846154,
      "grad_norm": 2930266.0,
      "learning_rate": 1e-06,
      "loss": 34.3929,
      "step": 2334
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3125629723072052,
      "epoch": 3.7419871794871797,
      "grad_norm": 533435.4375,
      "learning_rate": 1e-06,
      "loss": 14.3577,
      "step": 2335
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.30381515622138977,
      "epoch": 3.7435897435897436,
      "grad_norm": 0.9284766912460327,
      "learning_rate": 1e-06,
      "loss": 0.0779,
      "step": 2336
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.06640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3698.0,
      "completions/mean_length": 2933.576171875,
      "completions/mean_terminated_length": 1976.8514404296875,
      "completions/min_length": 802.0,
      "completions/min_terminated_length": 802.0,
      "entropy": 0.28462617099285126,
      "epoch": 3.7451923076923075,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 4198.36865234375,
      "learning_rate": 1e-06,
      "loss": 0.1291,
      "num_tokens": 1366628678.0,
      "reward": 0.24286873638629913,
      "reward_std": 0.05714884027838707,
      "rewards/progression_diversity/mean": -0.014884944073855877,
      "rewards/progression_diversity/std": 0.05795244127511978,
      "rewards/symbolic_reward_accuracy/mean": 0.080078125,
      "rewards/symbolic_reward_accuracy/std": 0.271679550409317,
      "rewards/symbolic_reward_partial_score/mean": 0.6570637822151184,
      "rewards/symbolic_reward_partial_score/std": 0.2119654417037964,
      "rewards/tag_count_reward/mean": -0.021484375,
      "rewards/tag_count_reward/std": 0.14513419568538666,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0245647430419922,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 27.924781799316406,
      "step": 2337
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.3178318291902542,
      "epoch": 3.746794871794872,
      "grad_norm": 0.02091590128839016,
      "learning_rate": 1e-06,
      "loss": 0.039,
      "step": 2338
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.29258735477924347,
      "epoch": 3.748397435897436,
      "grad_norm": 0.03195264935493469,
      "learning_rate": 1e-06,
      "loss": 0.0798,
      "step": 2339
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.3096400648355484,
      "epoch": 3.75,
      "grad_norm": 0.027371717616915703,
      "learning_rate": 1e-06,
      "loss": 0.0889,
      "step": 2340
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.060546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4023.0,
      "completions/mean_length": 2866.087890625,
      "completions/mean_terminated_length": 1994.87109375,
      "completions/min_length": 936.0,
      "completions/min_terminated_length": 936.0,
      "entropy": 0.3162682503461838,
      "epoch": 3.751602564102564,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 2558.345458984375,
      "learning_rate": 1e-06,
      "loss": 0.0613,
      "num_tokens": 1368975267.0,
      "reward": 0.2645840644836426,
      "reward_std": 0.05693788453936577,
      "rewards/progression_diversity/mean": -0.013275885954499245,
      "rewards/progression_diversity/std": 0.05504615232348442,
      "rewards/symbolic_reward_accuracy/mean": 0.1171875,
      "rewards/symbolic_reward_accuracy/std": 0.32195815443992615,
      "rewards/symbolic_reward_partial_score/mean": 0.6545247435569763,
      "rewards/symbolic_reward_partial_score/std": 0.22332006692886353,
      "rewards/tag_count_reward/mean": -0.01953125,
      "rewards/tag_count_reward/std": 0.1385180652141571,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.033745288848877,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 23.267847061157227,
      "step": 2341
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.3221198171377182,
      "epoch": 3.753205128205128,
      "grad_norm": 0.016970986500382423,
      "learning_rate": 1e-06,
      "loss": 0.0128,
      "step": 2342
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2914904057979584,
      "epoch": 3.7548076923076925,
      "grad_norm": 0.016527332365512848,
      "learning_rate": 1e-06,
      "loss": 0.1016,
      "step": 2343
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.29611217975616455,
      "epoch": 3.7564102564102564,
      "grad_norm": 0.015034251846373081,
      "learning_rate": 1e-06,
      "loss": 0.1444,
      "step": 2344
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.068359375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3699.0,
      "completions/mean_length": 2942.474609375,
      "completions/mean_terminated_length": 1956.1990966796875,
      "completions/min_length": 891.0,
      "completions/min_terminated_length": 891.0,
      "entropy": 0.324889600276947,
      "epoch": 3.7580128205128203,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 564.7570190429688,
      "learning_rate": 1e-06,
      "loss": -0.0078,
      "num_tokens": 1371389206.0,
      "reward": 0.4041377902030945,
      "reward_std": 0.0859639048576355,
      "rewards/progression_diversity/mean": -0.01590690203011036,
      "rewards/progression_diversity/std": 0.06111828610301018,
      "rewards/symbolic_reward_accuracy/mean": 0.314453125,
      "rewards/symbolic_reward_accuracy/std": 0.4647517800331116,
      "rewards/symbolic_reward_partial_score/mean": 0.7259114384651184,
      "rewards/symbolic_reward_partial_score/std": 0.2665213644504547,
      "rewards/tag_count_reward/mean": -0.021484375,
      "rewards/tag_count_reward/std": 0.14513419568538666,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0233384370803833,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 29.020061492919922,
      "step": 2345
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.2806452810764313,
      "epoch": 3.7596153846153846,
      "grad_norm": 0.017962336540222168,
      "learning_rate": 1e-06,
      "loss": 0.1727,
      "step": 2346
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.359375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.3103862851858139,
      "epoch": 3.761217948717949,
      "grad_norm": 0.013579954393208027,
      "learning_rate": 1e-06,
      "loss": 0.0811,
      "step": 2347
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.305005207657814,
      "epoch": 3.7628205128205128,
      "grad_norm": 0.01808071881532669,
      "learning_rate": 1e-06,
      "loss": 0.1148,
      "step": 2348
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.048828125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3296.0,
      "completions/mean_length": 2663.65234375,
      "completions/mean_terminated_length": 1959.322509765625,
      "completions/min_length": 626.0,
      "completions/min_terminated_length": 626.0,
      "entropy": 0.33493170142173767,
      "epoch": 3.7644230769230766,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 88.44670867919922,
      "learning_rate": 1e-06,
      "loss": -0.0023,
      "num_tokens": 1373608292.0,
      "reward": 0.3504989743232727,
      "reward_std": 0.0687357485294342,
      "rewards/progression_diversity/mean": -0.013580359518527985,
      "rewards/progression_diversity/std": 0.062451787292957306,
      "rewards/symbolic_reward_accuracy/mean": 0.22265625,
      "rewards/symbolic_reward_accuracy/std": 0.41643625497817993,
      "rewards/symbolic_reward_partial_score/mean": 0.7293294668197632,
      "rewards/symbolic_reward_partial_score/std": 0.23006510734558105,
      "rewards/tag_count_reward/mean": -0.017578125,
      "rewards/tag_count_reward/std": 0.13154059648513794,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0389269590377808,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 22.056781768798828,
      "step": 2349
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.3003799021244049,
      "epoch": 3.766025641025641,
      "grad_norm": 0.014394021593034267,
      "learning_rate": 1e-06,
      "loss": 0.1444,
      "step": 2350
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.3186776787042618,
      "epoch": 3.7676282051282053,
      "grad_norm": 0.02413056790828705,
      "learning_rate": 1e-06,
      "loss": 0.0553,
      "step": 2351
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.32591530680656433,
      "epoch": 3.769230769230769,
      "grad_norm": 0.01364830881357193,
      "learning_rate": 1e-06,
      "loss": 0.0332,
      "step": 2352
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3292.0,
      "completions/mean_length": 3137.02734375,
      "completions/mean_terminated_length": 1953.2552490234375,
      "completions/min_length": 601.0,
      "completions/min_terminated_length": 601.0,
      "entropy": 0.30336521565914154,
      "epoch": 3.7708333333333335,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 689.3433227539062,
      "learning_rate": 1e-06,
      "loss": 0.0575,
      "num_tokens": 1376180162.0,
      "reward": 0.30668240785598755,
      "reward_std": 0.07010701298713684,
      "rewards/progression_diversity/mean": -0.021703477948904037,
      "rewards/progression_diversity/std": 0.07450754940509796,
      "rewards/symbolic_reward_accuracy/mean": 0.171875,
      "rewards/symbolic_reward_accuracy/std": 0.3776407241821289,
      "rewards/symbolic_reward_partial_score/mean": 0.6844563484191895,
      "rewards/symbolic_reward_partial_score/std": 0.23659244179725647,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0194364786148071,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 32.71318054199219,
      "step": 2353
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.3008395582437515,
      "epoch": 3.7724358974358974,
      "grad_norm": 0.02489347755908966,
      "learning_rate": 1e-06,
      "loss": 0.0919,
      "step": 2354
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.3014819622039795,
      "epoch": 3.7740384615384617,
      "grad_norm": 0.018208302557468414,
      "learning_rate": 1e-06,
      "loss": 0.1138,
      "step": 2355
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.30451202392578125,
      "epoch": 3.7756410256410255,
      "grad_norm": 0.03824426978826523,
      "learning_rate": 1e-06,
      "loss": 0.049,
      "step": 2356
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3241.0,
      "completions/mean_length": 2671.005859375,
      "completions/mean_terminated_length": 1937.3887939453125,
      "completions/min_length": 908.0,
      "completions/min_terminated_length": 908.0,
      "entropy": 0.32287219166755676,
      "epoch": 3.77724358974359,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 901.6248779296875,
      "learning_rate": 1e-06,
      "loss": 0.0303,
      "num_tokens": 1378516597.0,
      "reward": 0.2821524143218994,
      "reward_std": 0.06520838290452957,
      "rewards/progression_diversity/mean": -0.015226154588162899,
      "rewards/progression_diversity/std": 0.06770697236061096,
      "rewards/symbolic_reward_accuracy/mean": 0.126953125,
      "rewards/symbolic_reward_accuracy/std": 0.33324605226516724,
      "rewards/symbolic_reward_partial_score/mean": 0.6903645992279053,
      "rewards/symbolic_reward_partial_score/std": 0.22543098032474518,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0395439863204956,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 22.674461364746094,
      "step": 2357
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.30759990215301514,
      "epoch": 3.7788461538461537,
      "grad_norm": 0.5585769414901733,
      "learning_rate": 1e-06,
      "loss": 0.0984,
      "step": 2358
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.3166644871234894,
      "epoch": 3.780448717948718,
      "grad_norm": 0.1928190290927887,
      "learning_rate": 1e-06,
      "loss": 0.0079,
      "step": 2359
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3207753002643585,
      "epoch": 3.782051282051282,
      "grad_norm": 0.02237272635102272,
      "learning_rate": 1e-06,
      "loss": 0.0525,
      "step": 2360
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.04296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3211.0,
      "completions/mean_length": 2510.060546875,
      "completions/mean_terminated_length": 1887.14892578125,
      "completions/min_length": 916.0,
      "completions/min_terminated_length": 916.0,
      "entropy": 0.3191475421190262,
      "epoch": 3.7836538461538463,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 950.1478271484375,
      "learning_rate": 1e-06,
      "loss": 0.0732,
      "num_tokens": 1380652708.0,
      "reward": 0.36586660146713257,
      "reward_std": 0.09463842213153839,
      "rewards/progression_diversity/mean": -0.011485631577670574,
      "rewards/progression_diversity/std": 0.05633028969168663,
      "rewards/symbolic_reward_accuracy/mean": 0.2421875,
      "rewards/symbolic_reward_accuracy/std": 0.42882615327835083,
      "rewards/symbolic_reward_partial_score/mean": 0.7420735359191895,
      "rewards/symbolic_reward_partial_score/std": 0.243971049785614,
      "rewards/tag_count_reward/mean": -0.01953125,
      "rewards/tag_count_reward/std": 0.1385180652141571,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.038116455078125,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 23.570289611816406,
      "step": 2361
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.33112041652202606,
      "epoch": 3.78525641025641,
      "grad_norm": 14.519754409790039,
      "learning_rate": 1e-06,
      "loss": 0.061,
      "step": 2362
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.33028391003608704,
      "epoch": 3.7868589743589745,
      "grad_norm": 0.024247992783784866,
      "learning_rate": 1e-06,
      "loss": 0.0352,
      "step": 2363
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.3137480616569519,
      "epoch": 3.7884615384615383,
      "grad_norm": 0.031128326430916786,
      "learning_rate": 1e-06,
      "loss": 0.0846,
      "step": 2364
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.060546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2966.0,
      "completions/mean_length": 2746.02734375,
      "completions/mean_terminated_length": 1867.07275390625,
      "completions/min_length": 746.0,
      "completions/min_terminated_length": 746.0,
      "entropy": 0.30356088280677795,
      "epoch": 3.7900641025641026,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 1132.28857421875,
      "learning_rate": 1e-06,
      "loss": 0.1076,
      "num_tokens": 1382941074.0,
      "reward": 0.30738329887390137,
      "reward_std": 0.057456813752651215,
      "rewards/progression_diversity/mean": -0.017531972378492355,
      "rewards/progression_diversity/std": 0.07124343514442444,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.6723307371139526,
      "rewards/symbolic_reward_partial_score/std": 0.23962588608264923,
      "rewards/tag_count_reward/mean": -0.01953125,
      "rewards/tag_count_reward/std": 0.1385180652141571,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0223664045333862,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 30.863096237182617,
      "step": 2365
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.306650310754776,
      "epoch": 3.7916666666666665,
      "grad_norm": 0.026560494676232338,
      "learning_rate": 1e-06,
      "loss": 0.1009,
      "step": 2366
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.313573881983757,
      "epoch": 3.793269230769231,
      "grad_norm": 0.01731313206255436,
      "learning_rate": 1e-06,
      "loss": 0.1042,
      "step": 2367
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.32431718707084656,
      "epoch": 3.7948717948717947,
      "grad_norm": 0.012698481790721416,
      "learning_rate": 1e-06,
      "loss": 0.0471,
      "step": 2368
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.06640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3138.0,
      "completions/mean_length": 2853.966796875,
      "completions/mean_terminated_length": 1891.5794677734375,
      "completions/min_length": 693.0,
      "completions/min_terminated_length": 693.0,
      "entropy": 0.31773290038108826,
      "epoch": 3.796474358974359,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 901.328369140625,
      "learning_rate": 1e-06,
      "loss": 0.0236,
      "num_tokens": 1385256705.0,
      "reward": 0.2686350345611572,
      "reward_std": 0.06755457073450089,
      "rewards/progression_diversity/mean": -0.018334418535232544,
      "rewards/progression_diversity/std": 0.07072144001722336,
      "rewards/symbolic_reward_accuracy/mean": 0.09765625,
      "rewards/symbolic_reward_accuracy/std": 0.29713961482048035,
      "rewards/symbolic_reward_partial_score/mean": 0.7098632454872131,
      "rewards/symbolic_reward_partial_score/std": 0.21099787950515747,
      "rewards/tag_count_reward/mean": -0.02734375,
      "rewards/tag_count_reward/std": 0.16324250400066376,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.017049789428711,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 34.733001708984375,
      "step": 2369
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.33099713921546936,
      "epoch": 3.7980769230769234,
      "grad_norm": 0.01458797324448824,
      "learning_rate": 1e-06,
      "loss": 0.0647,
      "step": 2370
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2994411885738373,
      "epoch": 3.7996794871794872,
      "grad_norm": 0.03737116977572441,
      "learning_rate": 1e-06,
      "loss": 0.1793,
      "step": 2371
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.32057875394821167,
      "epoch": 3.801282051282051,
      "grad_norm": 0.011851564049720764,
      "learning_rate": 1e-06,
      "loss": 0.0763,
      "step": 2372
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.04296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3275.0,
      "completions/mean_length": 2494.095703125,
      "completions/mean_terminated_length": 1870.46728515625,
      "completions/min_length": 597.0,
      "completions/min_terminated_length": 597.0,
      "entropy": 0.321204349398613,
      "epoch": 3.8028846153846154,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 464.8273620605469,
      "learning_rate": 1e-06,
      "loss": 0.0536,
      "num_tokens": 1387418002.0,
      "reward": 0.38406121730804443,
      "reward_std": 0.06720907241106033,
      "rewards/progression_diversity/mean": -0.011848336085677147,
      "rewards/progression_diversity/std": 0.05758029222488403,
      "rewards/symbolic_reward_accuracy/mean": 0.265625,
      "rewards/symbolic_reward_accuracy/std": 0.44209739565849304,
      "rewards/symbolic_reward_partial_score/mean": 0.7552083730697632,
      "rewards/symbolic_reward_partial_score/std": 0.21863536536693573,
      "rewards/tag_count_reward/mean": -0.017578125,
      "rewards/tag_count_reward/std": 0.13154059648513794,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.042336344718933,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 22.52627944946289,
      "step": 2373
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.33628255128860474,
      "epoch": 3.8044871794871797,
      "grad_norm": 16.08426856994629,
      "learning_rate": 1e-06,
      "loss": 0.047,
      "step": 2374
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.3366074562072754,
      "epoch": 3.8060897435897436,
      "grad_norm": 4.6837029457092285,
      "learning_rate": 1e-06,
      "loss": 0.0627,
      "step": 2375
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.32915179431438446,
      "epoch": 3.8076923076923075,
      "grad_norm": 0.2614407241344452,
      "learning_rate": 1e-06,
      "loss": 0.0403,
      "step": 2376
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3255.0,
      "completions/mean_length": 2107.373046875,
      "completions/mean_terminated_length": 1880.760009765625,
      "completions/min_length": 697.0,
      "completions/min_terminated_length": 697.0,
      "entropy": 0.33755332231521606,
      "epoch": 3.809294871794872,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 317.5009460449219,
      "learning_rate": 1e-06,
      "loss": 0.0604,
      "num_tokens": 1389344513.0,
      "reward": 0.2682853937149048,
      "reward_std": 0.022097572684288025,
      "rewards/progression_diversity/mean": -0.003493726020678878,
      "rewards/progression_diversity/std": 0.02860984578728676,
      "rewards/symbolic_reward_accuracy/mean": 0.091796875,
      "rewards/symbolic_reward_accuracy/std": 0.289021372795105,
      "rewards/symbolic_reward_partial_score/mean": 0.712109386920929,
      "rewards/symbolic_reward_partial_score/std": 0.19135794043540955,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.067301869392395,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 10.378652572631836,
      "step": 2377
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.3546573221683502,
      "epoch": 3.810897435897436,
      "grad_norm": 0.024478528648614883,
      "learning_rate": 1e-06,
      "loss": 0.2443,
      "step": 2378
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.36258837580680847,
      "epoch": 3.8125,
      "grad_norm": 0.020279182121157646,
      "learning_rate": 1e-06,
      "loss": 0.0001,
      "step": 2379
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.3494533598423004,
      "epoch": 3.814102564102564,
      "grad_norm": 0.021380124613642693,
      "learning_rate": 1e-06,
      "loss": 0.0156,
      "step": 2380
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3135.0,
      "completions/mean_length": 2129.833984375,
      "completions/mean_terminated_length": 1874.7891845703125,
      "completions/min_length": 1129.0,
      "completions/min_terminated_length": 1129.0,
      "entropy": 0.33585888147354126,
      "epoch": 3.815705128205128,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 1741.7381591796875,
      "learning_rate": 1e-06,
      "loss": 0.0675,
      "num_tokens": 1391386876.0,
      "reward": 0.23786211013793945,
      "reward_std": 0.01681671291589737,
      "rewards/progression_diversity/mean": -0.0043185316026210785,
      "rewards/progression_diversity/std": 0.03434579074382782,
      "rewards/symbolic_reward_accuracy/mean": 0.0625,
      "rewards/symbolic_reward_accuracy/std": 0.2422981858253479,
      "rewards/symbolic_reward_partial_score/mean": 0.6706217527389526,
      "rewards/symbolic_reward_partial_score/std": 0.17827239632606506,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0659494400024414,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 11.326656341552734,
      "step": 2381
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.35594816505908966,
      "epoch": 3.8173076923076925,
      "grad_norm": 0.012054548598825932,
      "learning_rate": 1e-06,
      "loss": 0.001,
      "step": 2382
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3579705059528351,
      "epoch": 3.8189102564102564,
      "grad_norm": 0.025561781600117683,
      "learning_rate": 1e-06,
      "loss": 0.0028,
      "step": 2383
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1328125,
      "entropy": 0.33812710642814636,
      "epoch": 3.8205128205128203,
      "grad_norm": 0.027250172570347786,
      "learning_rate": 1e-06,
      "loss": 0.0297,
      "step": 2384
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3329.0,
      "completions/mean_length": 2672.564453125,
      "completions/mean_terminated_length": 1819.1556396484375,
      "completions/min_length": 961.0,
      "completions/min_terminated_length": 961.0,
      "entropy": 0.3372623771429062,
      "epoch": 3.8221153846153846,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 106.94781494140625,
      "learning_rate": 1e-06,
      "loss": -0.0134,
      "num_tokens": 1393736781.0,
      "reward": 0.338218629360199,
      "reward_std": 0.033895812928676605,
      "rewards/progression_diversity/mean": -0.015542363747954369,
      "rewards/progression_diversity/std": 0.06333222985267639,
      "rewards/symbolic_reward_accuracy/mean": 0.21484375,
      "rewards/symbolic_reward_accuracy/std": 0.4111155867576599,
      "rewards/symbolic_reward_partial_score/mean": 0.7027831673622131,
      "rewards/symbolic_reward_partial_score/std": 0.22551730275154114,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0194793939590454,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 35.487525939941406,
      "step": 2385
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3197958320379257,
      "epoch": 3.823717948717949,
      "grad_norm": 3424.5888671875,
      "learning_rate": 1e-06,
      "loss": 0.4028,
      "step": 2386
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3210483342409134,
      "epoch": 3.8253205128205128,
      "grad_norm": 0.28248870372772217,
      "learning_rate": 1e-06,
      "loss": 0.031,
      "step": 2387
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.31268228590488434,
      "epoch": 3.8269230769230766,
      "grad_norm": 0.01564093679189682,
      "learning_rate": 1e-06,
      "loss": 0.1354,
      "step": 2388
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.072265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2804.0,
      "completions/mean_length": 2821.177734375,
      "completions/mean_terminated_length": 1764.7052001953125,
      "completions/min_length": 957.0,
      "completions/min_terminated_length": 957.0,
      "entropy": 0.3302348703145981,
      "epoch": 3.828525641025641,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 1121.2408447265625,
      "learning_rate": 1e-06,
      "loss": 0.0995,
      "num_tokens": 1395997880.0,
      "reward": 0.3832484483718872,
      "reward_std": 0.039947956800460815,
      "rewards/progression_diversity/mean": -0.017442844808101654,
      "rewards/progression_diversity/std": 0.06394115090370178,
      "rewards/symbolic_reward_accuracy/mean": 0.2734375,
      "rewards/symbolic_reward_accuracy/std": 0.4461594223976135,
      "rewards/symbolic_reward_partial_score/mean": 0.733154296875,
      "rewards/symbolic_reward_partial_score/std": 0.22632841765880585,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0048656463623047,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 43.669517517089844,
      "step": 2389
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.2924426198005676,
      "epoch": 3.8301282051282053,
      "grad_norm": 1060.7529296875,
      "learning_rate": 1e-06,
      "loss": 0.3549,
      "step": 2390
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.34059783816337585,
      "epoch": 3.831730769230769,
      "grad_norm": 1.79716956615448,
      "learning_rate": 1e-06,
      "loss": 0.0341,
      "step": 2391
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.3371441662311554,
      "epoch": 3.8333333333333335,
      "grad_norm": 15.651530265808105,
      "learning_rate": 1e-06,
      "loss": 0.0659,
      "step": 2392
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.021484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2952.0,
      "completions/mean_length": 2100.259765625,
      "completions/mean_terminated_length": 1786.6446533203125,
      "completions/min_length": 694.0,
      "completions/min_terminated_length": 694.0,
      "entropy": 0.3499729782342911,
      "epoch": 3.8349358974358974,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 0.016351254656910896,
      "learning_rate": 1e-06,
      "loss": 0.0287,
      "num_tokens": 1397980253.0,
      "reward": 0.4495462477207184,
      "reward_std": 0.04341159760951996,
      "rewards/progression_diversity/mean": -0.006314532831311226,
      "rewards/progression_diversity/std": 0.04331482574343681,
      "rewards/symbolic_reward_accuracy/mean": 0.361328125,
      "rewards/symbolic_reward_accuracy/std": 0.48085519671440125,
      "rewards/symbolic_reward_partial_score/mean": 0.779296875,
      "rewards/symbolic_reward_partial_score/std": 0.22735320031642914,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0571317672729492,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 16.10877227783203,
      "step": 2393
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.3411282151937485,
      "epoch": 3.8365384615384617,
      "grad_norm": 8039.83837890625,
      "learning_rate": 1e-06,
      "loss": 0.8501,
      "step": 2394
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.356269970536232,
      "epoch": 3.8381410256410255,
      "grad_norm": 289.86322021484375,
      "learning_rate": 1e-06,
      "loss": 0.0269,
      "step": 2395
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.3452070653438568,
      "epoch": 3.83974358974359,
      "grad_norm": 0.014756686054170132,
      "learning_rate": 1e-06,
      "loss": 0.0689,
      "step": 2396
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4212.0,
      "completions/mean_length": 2003.048828125,
      "completions/mean_terminated_length": 1774.7799072265625,
      "completions/min_length": 998.0,
      "completions/min_terminated_length": 998.0,
      "entropy": 0.3432788699865341,
      "epoch": 3.8413461538461537,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 1194.415771484375,
      "learning_rate": 1e-06,
      "loss": 0.0683,
      "num_tokens": 1399852966.0,
      "reward": 0.3248797655105591,
      "reward_std": 0.023387808352708817,
      "rewards/progression_diversity/mean": -0.004212523810565472,
      "rewards/progression_diversity/std": 0.034703828394412994,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.7139322757720947,
      "rewards/symbolic_reward_partial_score/std": 0.2121579349040985,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0637973546981812,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 12.788324356079102,
      "step": 2397
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.36405032873153687,
      "epoch": 3.842948717948718,
      "grad_norm": 0.022353997454047203,
      "learning_rate": 1e-06,
      "loss": 0.0133,
      "step": 2398
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.35727696120738983,
      "epoch": 3.844551282051282,
      "grad_norm": 1.3744477033615112,
      "learning_rate": 1e-06,
      "loss": -0.002,
      "step": 2399
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.35195624828338623,
      "epoch": 3.8461538461538463,
      "grad_norm": 0.02140852063894272,
      "learning_rate": 1e-06,
      "loss": 0.0567,
      "step": 2400
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.005859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3118.0,
      "completions/mean_length": 1852.724609375,
      "completions/mean_terminated_length": 1767.07861328125,
      "completions/min_length": 1004.0,
      "completions/min_terminated_length": 1004.0,
      "entropy": 0.36357755959033966,
      "epoch": 3.84775641025641,
      "frac_reward_zero_std": 0.65625,
      "grad_norm": 0.02421417459845543,
      "learning_rate": 1e-06,
      "loss": -0.0003,
      "num_tokens": 1401668985.0,
      "reward": 0.31430160999298096,
      "reward_std": 0.01744687370955944,
      "rewards/progression_diversity/mean": -0.001969876466318965,
      "rewards/progression_diversity/std": 0.026346473023295403,
      "rewards/symbolic_reward_accuracy/mean": 0.154296875,
      "rewards/symbolic_reward_accuracy/std": 0.36158639192581177,
      "rewards/symbolic_reward_partial_score/mean": 0.739794909954071,
      "rewards/symbolic_reward_partial_score/std": 0.19770273566246033,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0794093608856201,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 5.0978240966796875,
      "step": 2401
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.125,
      "entropy": 0.35801486670970917,
      "epoch": 3.8493589743589745,
      "grad_norm": 0.02062184363603592,
      "learning_rate": 1e-06,
      "loss": 36.5217,
      "step": 2402
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1015625,
      "entropy": 0.36254678666591644,
      "epoch": 3.8509615384615383,
      "grad_norm": 0.024333849549293518,
      "learning_rate": 1e-06,
      "loss": 216.8933,
      "step": 2403
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0859375,
      "entropy": 0.3597784638404846,
      "epoch": 3.8525641025641026,
      "grad_norm": 0.016081588342785835,
      "learning_rate": 1e-06,
      "loss": 2.3758,
      "step": 2404
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2842.0,
      "completions/mean_length": 1912.380859375,
      "completions/mean_terminated_length": 1798.43115234375,
      "completions/min_length": 1112.0,
      "completions/min_terminated_length": 1112.0,
      "entropy": 0.3635348379611969,
      "epoch": 3.8541666666666665,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 754.1912231445312,
      "learning_rate": 1e-06,
      "loss": 0.0195,
      "num_tokens": 1403545500.0,
      "reward": 0.31084516644477844,
      "reward_std": 0.025590339675545692,
      "rewards/progression_diversity/mean": -0.002397148869931698,
      "rewards/progression_diversity/std": 0.027250634506344795,
      "rewards/symbolic_reward_accuracy/mean": 0.17578125,
      "rewards/symbolic_reward_accuracy/std": 0.3810062110424042,
      "rewards/symbolic_reward_partial_score/mean": 0.6853190660476685,
      "rewards/symbolic_reward_partial_score/std": 0.19453242421150208,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0779601335525513,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 6.665762901306152,
      "step": 2405
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.3639901280403137,
      "epoch": 3.855769230769231,
      "grad_norm": 0.02506587654352188,
      "learning_rate": 1e-06,
      "loss": 0.0106,
      "step": 2406
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.36371636390686035,
      "epoch": 3.8573717948717947,
      "grad_norm": 0.019767044112086296,
      "learning_rate": 1e-06,
      "loss": 0.0172,
      "step": 2407
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0546875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.37076058983802795,
      "epoch": 3.858974358974359,
      "grad_norm": 0.021978937089443207,
      "learning_rate": 1e-06,
      "loss": -0.0016,
      "step": 2408
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.009765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3139.0,
      "completions/mean_length": 1938.568359375,
      "completions/mean_terminated_length": 1796.1085205078125,
      "completions/min_length": 1101.0,
      "completions/min_terminated_length": 1101.0,
      "entropy": 0.3613286018371582,
      "epoch": 3.8605769230769234,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 0.014481249265372753,
      "learning_rate": 1e-06,
      "loss": 0.0153,
      "num_tokens": 1405497455.0,
      "reward": 0.3004390001296997,
      "reward_std": 0.0335419736802578,
      "rewards/progression_diversity/mean": -0.0029758564196527004,
      "rewards/progression_diversity/std": 0.031783487647771835,
      "rewards/symbolic_reward_accuracy/mean": 0.14453125,
      "rewards/symbolic_reward_accuracy/std": 0.35197147727012634,
      "rewards/symbolic_reward_partial_score/mean": 0.7151041626930237,
      "rewards/symbolic_reward_partial_score/std": 0.19497309625148773,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.074684500694275,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 8.23452377319336,
      "step": 2409
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.3613293021917343,
      "epoch": 3.8621794871794872,
      "grad_norm": 0.0199386365711689,
      "learning_rate": 1e-06,
      "loss": 0.0316,
      "step": 2410
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.3616602122783661,
      "epoch": 3.863782051282051,
      "grad_norm": 0.5316094756126404,
      "learning_rate": 1e-06,
      "loss": 0.0207,
      "step": 2411
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.36158451437950134,
      "epoch": 3.8653846153846154,
      "grad_norm": 0.012657379731535912,
      "learning_rate": 1e-06,
      "loss": 0.0239,
      "step": 2412
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3447.0,
      "completions/mean_length": 1951.318359375,
      "completions/mean_terminated_length": 1837.6751708984375,
      "completions/min_length": 755.0,
      "completions/min_terminated_length": 755.0,
      "entropy": 0.3670198619365692,
      "epoch": 3.8669871794871797,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.029966576024889946,
      "learning_rate": 1e-06,
      "loss": 0.0243,
      "num_tokens": 1407303266.0,
      "reward": 0.29933494329452515,
      "reward_std": 0.02662837505340576,
      "rewards/progression_diversity/mean": -0.001564997830428183,
      "rewards/progression_diversity/std": 0.01767376810312271,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.693798840045929,
      "rewards/symbolic_reward_partial_score/std": 0.1757093071937561,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.079904317855835,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 6.317461967468262,
      "step": 2413
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.380515456199646,
      "epoch": 3.8685897435897436,
      "grad_norm": 0.014633157290518284,
      "learning_rate": 1e-06,
      "loss": -0.0096,
      "step": 2414
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.36171650886535645,
      "epoch": 3.8701923076923075,
      "grad_norm": 0.019229549914598465,
      "learning_rate": 1e-06,
      "loss": 0.0416,
      "step": 2415
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.3781423568725586,
      "epoch": 3.871794871794872,
      "grad_norm": 0.031127827242016792,
      "learning_rate": 1e-06,
      "loss": 0.0018,
      "step": 2416
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3458.0,
      "completions/mean_length": 2027.650390625,
      "completions/mean_terminated_length": 1914.6082763671875,
      "completions/min_length": 1041.0,
      "completions/min_terminated_length": 1041.0,
      "entropy": 0.37264350056648254,
      "epoch": 3.873397435897436,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 181.88681030273438,
      "learning_rate": 1e-06,
      "loss": 0.0241,
      "num_tokens": 1409189023.0,
      "reward": 0.33668702840805054,
      "reward_std": 0.035758331418037415,
      "rewards/progression_diversity/mean": -0.0017103657592087984,
      "rewards/progression_diversity/std": 0.020190343260765076,
      "rewards/symbolic_reward_accuracy/mean": 0.20703125,
      "rewards/symbolic_reward_accuracy/std": 0.40557438135147095,
      "rewards/symbolic_reward_partial_score/mean": 0.7089356184005737,
      "rewards/symbolic_reward_partial_score/std": 0.2059522420167923,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.079702377319336,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 5.716109275817871,
      "step": 2417
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3729843199253082,
      "epoch": 3.875,
      "grad_norm": 0.0172417052090168,
      "learning_rate": 1e-06,
      "loss": -0.0047,
      "step": 2418
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.3554026186466217,
      "epoch": 3.876602564102564,
      "grad_norm": 0.014652607962489128,
      "learning_rate": 1e-06,
      "loss": 0.0365,
      "step": 2419
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.36666855216026306,
      "epoch": 3.878205128205128,
      "grad_norm": 0.025280388072133064,
      "learning_rate": 1e-06,
      "loss": 0.0004,
      "step": 2420
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3165.0,
      "completions/mean_length": 2169.5234375,
      "completions/mean_terminated_length": 1972.4912109375,
      "completions/min_length": 1212.0,
      "completions/min_terminated_length": 1212.0,
      "entropy": 0.35489410161972046,
      "epoch": 3.8798076923076925,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 993.462158203125,
      "learning_rate": 1e-06,
      "loss": 0.0898,
      "num_tokens": 1411120619.0,
      "reward": 0.3209129571914673,
      "reward_std": 0.036054469645023346,
      "rewards/progression_diversity/mean": -0.0024563795886933804,
      "rewards/progression_diversity/std": 0.022509947419166565,
      "rewards/symbolic_reward_accuracy/mean": 0.16796875,
      "rewards/symbolic_reward_accuracy/std": 0.374204158782959,
      "rewards/symbolic_reward_partial_score/mean": 0.7371094226837158,
      "rewards/symbolic_reward_partial_score/std": 0.17446953058242798,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0719996690750122,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 9.719746589660645,
      "step": 2421
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.372598797082901,
      "epoch": 3.8814102564102564,
      "grad_norm": 0.00928138755261898,
      "learning_rate": 1e-06,
      "loss": -0.0118,
      "step": 2422
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.3691015988588333,
      "epoch": 3.8830128205128203,
      "grad_norm": 0.018029429018497467,
      "learning_rate": 1e-06,
      "loss": 0.0338,
      "step": 2423
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.36332499980926514,
      "epoch": 3.8846153846153846,
      "grad_norm": 0.013046249747276306,
      "learning_rate": 1e-06,
      "loss": 0.0302,
      "step": 2424
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3514.0,
      "completions/mean_length": 2295.619140625,
      "completions/mean_terminated_length": 1957.498046875,
      "completions/min_length": 1022.0,
      "completions/min_terminated_length": 1022.0,
      "entropy": 0.3619091063737869,
      "epoch": 3.886217948717949,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1141.9002685546875,
      "learning_rate": 1e-06,
      "loss": 0.0544,
      "num_tokens": 1413051752.0,
      "reward": 0.3810175061225891,
      "reward_std": 0.04954010993242264,
      "rewards/progression_diversity/mean": -0.00469579640775919,
      "rewards/progression_diversity/std": 0.03242946416139603,
      "rewards/symbolic_reward_accuracy/mean": 0.2421875,
      "rewards/symbolic_reward_accuracy/std": 0.42882615327835083,
      "rewards/symbolic_reward_partial_score/mean": 0.789746105670929,
      "rewards/symbolic_reward_partial_score/std": 0.18808871507644653,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0633162260055542,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 14.651718139648438,
      "step": 2425
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.03125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.36615803837776184,
      "epoch": 3.8878205128205128,
      "grad_norm": 833.3787231445312,
      "learning_rate": 1e-06,
      "loss": 0.0552,
      "step": 2426
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.37386661767959595,
      "epoch": 3.8894230769230766,
      "grad_norm": 0.020444680005311966,
      "learning_rate": 1e-06,
      "loss": 0.0178,
      "step": 2427
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.3629365861415863,
      "epoch": 3.891025641025641,
      "grad_norm": 0.02197657711803913,
      "learning_rate": 1e-06,
      "loss": 0.0594,
      "step": 2428
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.017578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3316.0,
      "completions/mean_length": 2237.09375,
      "completions/mean_terminated_length": 1983.9681396484375,
      "completions/min_length": 1003.0,
      "completions/min_terminated_length": 1003.0,
      "entropy": 0.36241379380226135,
      "epoch": 3.8926282051282053,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 351.0287780761719,
      "learning_rate": 1e-06,
      "loss": 0.0279,
      "num_tokens": 1415054008.0,
      "reward": 0.33530277013778687,
      "reward_std": 0.013850128278136253,
      "rewards/progression_diversity/mean": -0.0029261300805956125,
      "rewards/progression_diversity/std": 0.02486591413617134,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.7447265386581421,
      "rewards/symbolic_reward_partial_score/std": 0.18490070104599,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0711203813552856,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 10.22694206237793,
      "step": 2429
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.3631092756986618,
      "epoch": 3.894230769230769,
      "grad_norm": 0.015336292795836926,
      "learning_rate": 1e-06,
      "loss": 0.0474,
      "step": 2430
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.3701094090938568,
      "epoch": 3.8958333333333335,
      "grad_norm": 0.019072409719228745,
      "learning_rate": 1e-06,
      "loss": -0.0113,
      "step": 2431
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.36329878866672516,
      "epoch": 3.8974358974358974,
      "grad_norm": 0.009201987646520138,
      "learning_rate": 1e-06,
      "loss": 0.0567,
      "step": 2432
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.013671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3290.0,
      "completions/mean_length": 2161.703125,
      "completions/mean_terminated_length": 1964.5623779296875,
      "completions/min_length": 1002.0,
      "completions/min_terminated_length": 1002.0,
      "entropy": 0.36919447779655457,
      "epoch": 3.8990384615384617,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 651.1842041015625,
      "learning_rate": 1e-06,
      "loss": 0.0223,
      "num_tokens": 1417071456.0,
      "reward": 0.26072680950164795,
      "reward_std": 0.03266744688153267,
      "rewards/progression_diversity/mean": -0.0025163046084344387,
      "rewards/progression_diversity/std": 0.024175414815545082,
      "rewards/symbolic_reward_accuracy/mean": 0.0859375,
      "rewards/symbolic_reward_accuracy/std": 0.28054583072662354,
      "rewards/symbolic_reward_partial_score/mean": 0.6992512941360474,
      "rewards/symbolic_reward_partial_score/std": 0.1867949366569519,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0748156309127808,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 8.038581848144531,
      "step": 2433
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.3648574501276016,
      "epoch": 3.9006410256410255,
      "grad_norm": 0.026968425139784813,
      "learning_rate": 1e-06,
      "loss": 0.0404,
      "step": 2434
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.3710328936576843,
      "epoch": 3.90224358974359,
      "grad_norm": 0.022169405594468117,
      "learning_rate": 1e-06,
      "loss": 0.0208,
      "step": 2435
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3560585528612137,
      "epoch": 3.9038461538461537,
      "grad_norm": 0.012255042791366577,
      "learning_rate": 1e-06,
      "loss": 0.0366,
      "step": 2436
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3374.0,
      "completions/mean_length": 2268.74609375,
      "completions/mean_terminated_length": 1987.5657958984375,
      "completions/min_length": 932.0,
      "completions/min_terminated_length": 932.0,
      "entropy": 0.36314675211906433,
      "epoch": 3.905448717948718,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 26.744871139526367,
      "learning_rate": 1e-06,
      "loss": 0.0218,
      "num_tokens": 1419090318.0,
      "reward": 0.41112661361694336,
      "reward_std": 0.022687647491693497,
      "rewards/progression_diversity/mean": -0.0030626305378973484,
      "rewards/progression_diversity/std": 0.022572841495275497,
      "rewards/symbolic_reward_accuracy/mean": 0.306640625,
      "rewards/symbolic_reward_accuracy/std": 0.4615498185157776,
      "rewards/symbolic_reward_partial_score/mean": 0.7591959238052368,
      "rewards/symbolic_reward_partial_score/std": 0.19385521113872528,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0674618482589722,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 11.076675415039062,
      "step": 2437
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.3628970682621002,
      "epoch": 3.907051282051282,
      "grad_norm": 0.018419716507196426,
      "learning_rate": 1e-06,
      "loss": 0.0137,
      "step": 2438
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.34759269654750824,
      "epoch": 3.9086538461538463,
      "grad_norm": 0.008998063392937183,
      "learning_rate": 1e-06,
      "loss": 0.0594,
      "step": 2439
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.3605688661336899,
      "epoch": 3.91025641025641,
      "grad_norm": 0.010430578142404556,
      "learning_rate": 1e-06,
      "loss": 0.0363,
      "step": 2440
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3622.0,
      "completions/mean_length": 2239.3984375,
      "completions/mean_terminated_length": 2014.881103515625,
      "completions/min_length": 714.0,
      "completions/min_terminated_length": 714.0,
      "entropy": 0.3595893830060959,
      "epoch": 3.9118589743589745,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 0.020290104672312737,
      "learning_rate": 1e-06,
      "loss": 0.0345,
      "num_tokens": 1421054490.0,
      "reward": 0.3346139192581177,
      "reward_std": 0.03494077920913696,
      "rewards/progression_diversity/mean": -0.002476356690749526,
      "rewards/progression_diversity/std": 0.022448500618338585,
      "rewards/symbolic_reward_accuracy/mean": 0.193359375,
      "rewards/symbolic_reward_accuracy/std": 0.39531853795051575,
      "rewards/symbolic_reward_partial_score/mean": 0.7313476800918579,
      "rewards/symbolic_reward_partial_score/std": 0.19797399640083313,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0738534927368164,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 8.171432495117188,
      "step": 2441
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.3624930679798126,
      "epoch": 3.9134615384615383,
      "grad_norm": 0.01627454161643982,
      "learning_rate": 1e-06,
      "loss": 0.0318,
      "step": 2442
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.3630795478820801,
      "epoch": 3.9150641025641026,
      "grad_norm": 0.011845182627439499,
      "learning_rate": 1e-06,
      "loss": 0.0527,
      "step": 2443
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1640625,
      "entropy": 0.3635891079902649,
      "epoch": 3.9166666666666665,
      "grad_norm": 0.012927143834531307,
      "learning_rate": 1e-06,
      "loss": 0.0315,
      "step": 2444
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3619.0,
      "completions/mean_length": 2434.078125,
      "completions/mean_terminated_length": 2041.91162109375,
      "completions/min_length": 687.0,
      "completions/min_terminated_length": 687.0,
      "entropy": 0.3447739779949188,
      "epoch": 3.918269230769231,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 580.8510131835938,
      "learning_rate": 1e-06,
      "loss": 0.0723,
      "num_tokens": 1423102786.0,
      "reward": 0.39957115054130554,
      "reward_std": 0.04453998804092407,
      "rewards/progression_diversity/mean": -0.00382626592181623,
      "rewards/progression_diversity/std": 0.02524666115641594,
      "rewards/symbolic_reward_accuracy/mean": 0.283203125,
      "rewards/symbolic_reward_accuracy/std": 0.4509948492050171,
      "rewards/symbolic_reward_partial_score/mean": 0.7708333134651184,
      "rewards/symbolic_reward_partial_score/std": 0.20369426906108856,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0627055168151855,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 13.541450500488281,
      "step": 2445
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3629473149776459,
      "epoch": 3.9198717948717947,
      "grad_norm": 0.03049568459391594,
      "learning_rate": 1e-06,
      "loss": 0.041,
      "step": 2446
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.360695943236351,
      "epoch": 3.921474358974359,
      "grad_norm": 0.03725721687078476,
      "learning_rate": 1e-06,
      "loss": 0.0375,
      "step": 2447
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.36237798631191254,
      "epoch": 3.9230769230769234,
      "grad_norm": 0.01814689300954342,
      "learning_rate": 1e-06,
      "loss": 0.0465,
      "step": 2448
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.029296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4143.0,
      "completions/mean_length": 2448.4453125,
      "completions/mean_terminated_length": 2027.8551025390625,
      "completions/min_length": 1131.0,
      "completions/min_terminated_length": 1131.0,
      "entropy": 0.3400779664516449,
      "epoch": 3.9246794871794872,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 717.636962890625,
      "learning_rate": 1e-06,
      "loss": 0.1025,
      "num_tokens": 1425199558.0,
      "reward": 0.3190501034259796,
      "reward_std": 0.047453030943870544,
      "rewards/progression_diversity/mean": -0.0046582636423408985,
      "rewards/progression_diversity/std": 0.029829828068614006,
      "rewards/symbolic_reward_accuracy/mean": 0.173828125,
      "rewards/symbolic_reward_accuracy/std": 0.3793322443962097,
      "rewards/symbolic_reward_partial_score/mean": 0.7205566167831421,
      "rewards/symbolic_reward_partial_score/std": 0.21557864546775818,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0601270198822021,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 14.778732299804688,
      "step": 2449
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.36184197664260864,
      "epoch": 3.926282051282051,
      "grad_norm": 0.07136629521846771,
      "learning_rate": 1e-06,
      "loss": -0.0179,
      "step": 2450
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.36823903024196625,
      "epoch": 3.9278846153846154,
      "grad_norm": 0.012042498216032982,
      "learning_rate": 1e-06,
      "loss": 0.0099,
      "step": 2451
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.3471096456050873,
      "epoch": 3.9294871794871797,
      "grad_norm": 0.020903823897242546,
      "learning_rate": 1e-06,
      "loss": 0.1012,
      "step": 2452
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.025390625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4468.0,
      "completions/mean_length": 2371.39453125,
      "completions/mean_terminated_length": 2006.336669921875,
      "completions/min_length": 1198.0,
      "completions/min_terminated_length": 1198.0,
      "entropy": 0.3484654426574707,
      "epoch": 3.9310897435897436,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 2592.323974609375,
      "learning_rate": 1e-06,
      "loss": 0.0898,
      "num_tokens": 1427341920.0,
      "reward": 0.32160401344299316,
      "reward_std": 0.039857640862464905,
      "rewards/progression_diversity/mean": -0.004153084009885788,
      "rewards/progression_diversity/std": 0.028649339452385902,
      "rewards/symbolic_reward_accuracy/mean": 0.173828125,
      "rewards/symbolic_reward_accuracy/std": 0.3793322443962097,
      "rewards/symbolic_reward_partial_score/mean": 0.729052722454071,
      "rewards/symbolic_reward_partial_score/std": 0.21416015923023224,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0638084411621094,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 12.2067232131958,
      "step": 2453
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.347993940114975,
      "epoch": 3.9326923076923075,
      "grad_norm": 0.02119533158838749,
      "learning_rate": 1e-06,
      "loss": 0.0476,
      "step": 2454
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3478931188583374,
      "epoch": 3.934294871794872,
      "grad_norm": 0.014371760189533234,
      "learning_rate": 1e-06,
      "loss": 0.0499,
      "step": 2455
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.35639145970344543,
      "epoch": 3.935897435897436,
      "grad_norm": 0.015084910206496716,
      "learning_rate": 1e-06,
      "loss": 0.0251,
      "step": 2456
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3296.0,
      "completions/mean_length": 2630.390625,
      "completions/mean_terminated_length": 1953.9835205078125,
      "completions/min_length": 1124.0,
      "completions/min_terminated_length": 1124.0,
      "entropy": 0.34472502768039703,
      "epoch": 3.9375,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 823.965576171875,
      "learning_rate": 1e-06,
      "loss": 0.0218,
      "num_tokens": 1429514888.0,
      "reward": 0.386918306350708,
      "reward_std": 0.04596348851919174,
      "rewards/progression_diversity/mean": -0.009828666225075722,
      "rewards/progression_diversity/std": 0.04648015275597572,
      "rewards/symbolic_reward_accuracy/mean": 0.271484375,
      "rewards/symbolic_reward_accuracy/std": 0.44516023993492126,
      "rewards/symbolic_reward_partial_score/mean": 0.7503417730331421,
      "rewards/symbolic_reward_partial_score/std": 0.21974274516105652,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0469261407852173,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 20.604896545410156,
      "step": 2457
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.3289848268032074,
      "epoch": 3.939102564102564,
      "grad_norm": 0.01477957796305418,
      "learning_rate": 1e-06,
      "loss": 0.131,
      "step": 2458
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.3502272367477417,
      "epoch": 3.940705128205128,
      "grad_norm": 0.026974955573678017,
      "learning_rate": 1e-06,
      "loss": 0.0262,
      "step": 2459
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.34581419825553894,
      "epoch": 3.9423076923076925,
      "grad_norm": 0.015963274985551834,
      "learning_rate": 1e-06,
      "loss": 0.0406,
      "step": 2460
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3357.0,
      "completions/mean_length": 2388.841796875,
      "completions/mean_terminated_length": 1937.385009765625,
      "completions/min_length": 1023.0,
      "completions/min_terminated_length": 1023.0,
      "entropy": 0.35383833944797516,
      "epoch": 3.9439102564102564,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 648.1668090820312,
      "learning_rate": 1e-06,
      "loss": 0.0584,
      "num_tokens": 1431583207.0,
      "reward": 0.3101181387901306,
      "reward_std": 0.026755383238196373,
      "rewards/progression_diversity/mean": -0.00625104084610939,
      "rewards/progression_diversity/std": 0.03660469129681587,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.7318522334098816,
      "rewards/symbolic_reward_partial_score/std": 0.19400212168693542,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0573663711547852,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 16.051483154296875,
      "step": 2461
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.36202606558799744,
      "epoch": 3.9455128205128203,
      "grad_norm": 0.061913300305604935,
      "learning_rate": 1e-06,
      "loss": 0.015,
      "step": 2462
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.36147141456604004,
      "epoch": 3.9471153846153846,
      "grad_norm": 0.014589556492865086,
      "learning_rate": 1e-06,
      "loss": 0.0385,
      "step": 2463
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.3382275700569153,
      "epoch": 3.948717948717949,
      "grad_norm": 0.013087283819913864,
      "learning_rate": 1e-06,
      "loss": 0.0554,
      "step": 2464
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3255.0,
      "completions/mean_length": 2498.75,
      "completions/mean_terminated_length": 1963.61865234375,
      "completions/min_length": 982.0,
      "completions/min_terminated_length": 982.0,
      "entropy": 0.35105569660663605,
      "epoch": 3.9503205128205128,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 1137.3775634765625,
      "learning_rate": 1e-06,
      "loss": 0.0478,
      "num_tokens": 1433752439.0,
      "reward": 0.384078711271286,
      "reward_std": 0.03686286136507988,
      "rewards/progression_diversity/mean": -0.007657800801098347,
      "rewards/progression_diversity/std": 0.04040105640888214,
      "rewards/symbolic_reward_accuracy/mean": 0.275390625,
      "rewards/symbolic_reward_accuracy/std": 0.44714778661727905,
      "rewards/symbolic_reward_partial_score/mean": 0.7329915165901184,
      "rewards/symbolic_reward_partial_score/std": 0.22255964577198029,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0503382682800293,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 19.593889236450195,
      "step": 2465
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3548099845647812,
      "epoch": 3.9519230769230766,
      "grad_norm": 0.07064145803451538,
      "learning_rate": 1e-06,
      "loss": 0.0108,
      "step": 2466
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.3451448976993561,
      "epoch": 3.953525641025641,
      "grad_norm": 0.03678792715072632,
      "learning_rate": 1e-06,
      "loss": 0.0902,
      "step": 2467
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.34411998093128204,
      "epoch": 3.9551282051282053,
      "grad_norm": 0.026875555515289307,
      "learning_rate": 1e-06,
      "loss": 0.0816,
      "step": 2468
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3373.0,
      "completions/mean_length": 2192.21484375,
      "completions/mean_terminated_length": 1909.510009765625,
      "completions/min_length": 1090.0,
      "completions/min_terminated_length": 1090.0,
      "entropy": 0.3781355321407318,
      "epoch": 3.956730769230769,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.01978994905948639,
      "learning_rate": 1e-06,
      "loss": -0.0087,
      "num_tokens": 1435768037.0,
      "reward": 0.2731976807117462,
      "reward_std": 0.02287241816520691,
      "rewards/progression_diversity/mean": -0.004451290238648653,
      "rewards/progression_diversity/std": 0.0341835618019104,
      "rewards/symbolic_reward_accuracy/mean": 0.123046875,
      "rewards/symbolic_reward_accuracy/std": 0.32881227135658264,
      "rewards/symbolic_reward_partial_score/mean": 0.6666666865348816,
      "rewards/symbolic_reward_partial_score/std": 0.2050163298845291,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0635859966278076,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 13.839681625366211,
      "step": 2469
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3631334751844406,
      "epoch": 3.9583333333333335,
      "grad_norm": 4.625295639038086,
      "learning_rate": 1e-06,
      "loss": 0.0434,
      "step": 2470
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0390625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.171875,
      "entropy": 0.35326145589351654,
      "epoch": 3.9599358974358974,
      "grad_norm": 0.01084475964307785,
      "learning_rate": 1e-06,
      "loss": 0.0951,
      "step": 2471
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.3633986711502075,
      "epoch": 3.9615384615384617,
      "grad_norm": 0.013056226074695587,
      "learning_rate": 1e-06,
      "loss": 0.0322,
      "step": 2472
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.029296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3047.0,
      "completions/mean_length": 2316.728515625,
      "completions/mean_terminated_length": 1892.1629638671875,
      "completions/min_length": 1024.0,
      "completions/min_terminated_length": 1024.0,
      "entropy": 0.36079515516757965,
      "epoch": 3.9631410256410255,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 649.272216796875,
      "learning_rate": 1e-06,
      "loss": 0.0074,
      "num_tokens": 1437927290.0,
      "reward": 0.2825604975223541,
      "reward_std": 0.047691911458969116,
      "rewards/progression_diversity/mean": -0.007623001933097839,
      "rewards/progression_diversity/std": 0.04595175012946129,
      "rewards/symbolic_reward_accuracy/mean": 0.134765625,
      "rewards/symbolic_reward_accuracy/std": 0.3418070077896118,
      "rewards/symbolic_reward_partial_score/mean": 0.6758463382720947,
      "rewards/symbolic_reward_partial_score/std": 0.2019786536693573,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0513596534729004,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 19.716333389282227,
      "step": 2473
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3496604412794113,
      "epoch": 3.96474358974359,
      "grad_norm": 3.18587327003479,
      "learning_rate": 1e-06,
      "loss": 0.0646,
      "step": 2474
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.3459862917661667,
      "epoch": 3.9663461538461537,
      "grad_norm": 0.016324063763022423,
      "learning_rate": 1e-06,
      "loss": 0.0452,
      "step": 2475
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.35571056604385376,
      "epoch": 3.967948717948718,
      "grad_norm": 0.014685800299048424,
      "learning_rate": 1e-06,
      "loss": 0.0737,
      "step": 2476
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3300.0,
      "completions/mean_length": 2103.21484375,
      "completions/mean_terminated_length": 1876.5357666015625,
      "completions/min_length": 1079.0,
      "completions/min_terminated_length": 1079.0,
      "entropy": 0.36506491899490356,
      "epoch": 3.969551282051282,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.013999047689139843,
      "learning_rate": 1e-06,
      "loss": 0.0181,
      "num_tokens": 1439892248.0,
      "reward": 0.34133556485176086,
      "reward_std": 0.026946410536766052,
      "rewards/progression_diversity/mean": -0.0046281758695840836,
      "rewards/progression_diversity/std": 0.038301728665828705,
      "rewards/symbolic_reward_accuracy/mean": 0.189453125,
      "rewards/symbolic_reward_accuracy/std": 0.3922513723373413,
      "rewards/symbolic_reward_partial_score/mean": 0.7603353261947632,
      "rewards/symbolic_reward_partial_score/std": 0.17550498247146606,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0721997022628784,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 10.131376266479492,
      "step": 2477
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.36544330418109894,
      "epoch": 3.9711538461538463,
      "grad_norm": 0.02787620946764946,
      "learning_rate": 1e-06,
      "loss": 0.0459,
      "step": 2478
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.3669802248477936,
      "epoch": 3.97275641025641,
      "grad_norm": 0.015974465757608414,
      "learning_rate": 1e-06,
      "loss": 0.0146,
      "step": 2479
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.36941133439540863,
      "epoch": 3.9743589743589745,
      "grad_norm": 0.02147599868476391,
      "learning_rate": 1e-06,
      "loss": 0.0229,
      "step": 2480
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2934.0,
      "completions/mean_length": 3074.95703125,
      "completions/mean_terminated_length": 1885.63818359375,
      "completions/min_length": 1061.0,
      "completions/min_terminated_length": 1061.0,
      "entropy": 0.3382083773612976,
      "epoch": 3.9759615384615383,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 1267.6973876953125,
      "learning_rate": 1e-06,
      "loss": 0.0402,
      "num_tokens": 1442438978.0,
      "reward": 0.2858215570449829,
      "reward_std": 0.044375061988830566,
      "rewards/progression_diversity/mean": -0.023311004042625427,
      "rewards/progression_diversity/std": 0.08079320937395096,
      "rewards/symbolic_reward_accuracy/mean": 0.134765625,
      "rewards/symbolic_reward_accuracy/std": 0.3418070077896118,
      "rewards/symbolic_reward_partial_score/mean": 0.6885416507720947,
      "rewards/symbolic_reward_partial_score/std": 0.19891256093978882,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.005120038986206,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 43.875850677490234,
      "step": 2481
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.3479975014925003,
      "epoch": 3.9775641025641026,
      "grad_norm": 0.0137171084061265,
      "learning_rate": 1e-06,
      "loss": 0.042,
      "step": 2482
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.31158269941806793,
      "epoch": 3.9791666666666665,
      "grad_norm": 0.028567982837557793,
      "learning_rate": 1e-06,
      "loss": 0.159,
      "step": 2483
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.31820471584796906,
      "epoch": 3.980769230769231,
      "grad_norm": 0.03962567821145058,
      "learning_rate": 1e-06,
      "loss": 0.1121,
      "step": 2484
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.076171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3611.0,
      "completions/mean_length": 3046.513671875,
      "completions/mean_terminated_length": 1946.805419921875,
      "completions/min_length": 1105.0,
      "completions/min_terminated_length": 1105.0,
      "entropy": 0.33995670080184937,
      "epoch": 3.9823717948717947,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 649.2957763671875,
      "learning_rate": 1e-06,
      "loss": 0.0291,
      "num_tokens": 1444918329.0,
      "reward": 0.3851238489151001,
      "reward_std": 0.020528025925159454,
      "rewards/progression_diversity/mean": -0.020821530371904373,
      "rewards/progression_diversity/std": 0.0761089101433754,
      "rewards/symbolic_reward_accuracy/mean": 0.28125,
      "rewards/symbolic_reward_accuracy/std": 0.45004892349243164,
      "rewards/symbolic_reward_partial_score/mean": 0.7258462905883789,
      "rewards/symbolic_reward_partial_score/std": 0.23976966738700867,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0180208683013916,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 37.60797882080078,
      "step": 2485
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3226911425590515,
      "epoch": 3.983974358974359,
      "grad_norm": 3.0483925342559814,
      "learning_rate": 1e-06,
      "loss": 0.1089,
      "step": 2486
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.078125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1875,
      "entropy": 0.3385407626628876,
      "epoch": 3.9855769230769234,
      "grad_norm": 0.01842661201953888,
      "learning_rate": 1e-06,
      "loss": 0.0844,
      "step": 2487
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3237884044647217,
      "epoch": 3.9871794871794872,
      "grad_norm": 0.01716519333422184,
      "learning_rate": 1e-06,
      "loss": 0.1273,
      "step": 2488
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3178.0,
      "completions/mean_length": 3356.56640625,
      "completions/mean_terminated_length": 1883.89990234375,
      "completions/min_length": 1032.0,
      "completions/min_terminated_length": 1032.0,
      "entropy": 0.33329902589321136,
      "epoch": 3.988782051282051,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 1143.9063720703125,
      "learning_rate": 1e-06,
      "loss": 0.085,
      "num_tokens": 1447458059.0,
      "reward": 0.48112747073173523,
      "reward_std": 0.07644528150558472,
      "rewards/progression_diversity/mean": -0.028856143355369568,
      "rewards/progression_diversity/std": 0.08837366849184036,
      "rewards/symbolic_reward_accuracy/mean": 0.4140625,
      "rewards/symbolic_reward_accuracy/std": 0.49304109811782837,
      "rewards/symbolic_reward_partial_score/mean": 0.7818033695220947,
      "rewards/symbolic_reward_partial_score/std": 0.23644909262657166,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0008317232131958,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 46.746429443359375,
      "step": 2489
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.3214826136827469,
      "epoch": 3.9903846153846154,
      "grad_norm": 32.26913833618164,
      "learning_rate": 1e-06,
      "loss": 0.0057,
      "step": 2490
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.3154735267162323,
      "epoch": 3.9919871794871797,
      "grad_norm": 9.52782154083252,
      "learning_rate": 1e-06,
      "loss": 0.5166,
      "step": 2491
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.31883853673934937,
      "epoch": 3.9935897435897436,
      "grad_norm": 1062.50439453125,
      "learning_rate": 1e-06,
      "loss": 0.3673,
      "step": 2492
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3200.0,
      "completions/mean_length": 3408.443359375,
      "completions/mean_terminated_length": 1941.6412353515625,
      "completions/min_length": 1132.0,
      "completions/min_terminated_length": 1132.0,
      "entropy": 0.3397838622331619,
      "epoch": 3.9951923076923075,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 321.0043640136719,
      "learning_rate": 1e-06,
      "loss": 0.0395,
      "num_tokens": 1449967230.0,
      "reward": 0.4184061884880066,
      "reward_std": 0.0656353235244751,
      "rewards/progression_diversity/mean": -0.028522029519081116,
      "rewards/progression_diversity/std": 0.08747442811727524,
      "rewards/symbolic_reward_accuracy/mean": 0.3203125,
      "rewards/symbolic_reward_accuracy/std": 0.4670529365539551,
      "rewards/symbolic_reward_partial_score/mean": 0.7582682371139526,
      "rewards/symbolic_reward_partial_score/std": 0.22663599252700806,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0017540454864502,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 46.01112747192383,
      "step": 2493
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.3056804835796356,
      "epoch": 3.996794871794872,
      "grad_norm": 35.7710075378418,
      "learning_rate": 1e-06,
      "loss": 0.2021,
      "step": 2494
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.33890417218208313,
      "epoch": 3.998397435897436,
      "grad_norm": 0.29032865166664124,
      "learning_rate": 1e-06,
      "loss": 0.0841,
      "step": 2495
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.31833402812480927,
      "epoch": 4.0,
      "grad_norm": 0.019398123025894165,
      "learning_rate": 1e-06,
      "loss": 0.1324,
      "step": 2496
    },
    {
      "epoch": 4.0,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.0859375,
      "eval_completions/max_length": 16384.0,
      "eval_completions/max_terminated_length": 2986.0625,
      "eval_completions/mean_length": 3210.553466796875,
      "eval_completions/mean_terminated_length": 1972.495346069336,
      "eval_completions/min_length": 1199.8125,
      "eval_completions/min_terminated_length": 1199.8125,
      "eval_entropy": 0.3062564991414547,
      "eval_frac_reward_zero_std": 0.23828125,
      "eval_loss": 0.03558493033051491,
      "eval_num_tokens": 1449967230.0,
      "eval_reward": 0.25646816566586494,
      "eval_reward_std": 0.03971813467796892,
      "eval_rewards/progression_diversity/mean": -0.022923253040062264,
      "eval_rewards/progression_diversity/std": 0.07360347534995526,
      "eval_rewards/symbolic_reward_accuracy/mean": 0.093994140625,
      "eval_rewards/symbolic_reward_accuracy/std": 0.21256664022803307,
      "eval_rewards/symbolic_reward_partial_score/mean": 0.6708434987813234,
      "eval_rewards/symbolic_reward_partial_score/std": 0.1871126431506127,
      "eval_rewards/tag_count_reward/mean": -0.009521484375,
      "eval_rewards/tag_count_reward/std": 0.07448925846256316,
      "eval_runtime": 4359.773,
      "eval_samples_per_second": 0.057,
      "eval_sampling/importance_sampling_ratio/max": 2.0,
      "eval_sampling/importance_sampling_ratio/mean": 1.03790351934731,
      "eval_sampling/importance_sampling_ratio/min": 0.0,
      "eval_sampling/sampling_logp_difference/max": 782.375,
      "eval_sampling/sampling_logp_difference/mean": 27.384851962327957,
      "eval_steps_per_second": 0.0,
      "step": 2496
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3075.0,
      "completions/mean_length": 2979.962890625,
      "completions/mean_terminated_length": 1966.2122802734375,
      "completions/min_length": 1032.0,
      "completions/min_terminated_length": 1032.0,
      "entropy": 0.32689496874809265,
      "epoch": 4.001602564102564,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 373.3982849121094,
      "learning_rate": 1e-06,
      "loss": 0.0951,
      "num_tokens": 1452348555.0,
      "reward": 0.3907577395439148,
      "reward_std": 0.044927872717380524,
      "rewards/progression_diversity/mean": -0.018952488899230957,
      "rewards/progression_diversity/std": 0.0710325539112091,
      "rewards/symbolic_reward_accuracy/mean": 0.271484375,
      "rewards/symbolic_reward_accuracy/std": 0.44516023993492126,
      "rewards/symbolic_reward_partial_score/mean": 0.7653970718383789,
      "rewards/symbolic_reward_partial_score/std": 0.23204264044761658,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0241422653198242,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 34.15648651123047,
      "step": 2497
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0234375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.3391586244106293,
      "epoch": 4.003205128205129,
      "grad_norm": 0.013818570412695408,
      "learning_rate": 1e-06,
      "loss": 0.2369,
      "step": 2498
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.32884329557418823,
      "epoch": 4.0048076923076925,
      "grad_norm": 7747.47119140625,
      "learning_rate": 1e-06,
      "loss": 1.4664,
      "step": 2499
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3341674506664276,
      "epoch": 4.006410256410256,
      "grad_norm": 0.5125430822372437,
      "learning_rate": 1e-06,
      "loss": 0.2162,
      "step": 2500
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.095703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3270.0,
      "completions/mean_length": 3343.3125,
      "completions/mean_terminated_length": 1963.196533203125,
      "completions/min_length": 1081.0,
      "completions/min_terminated_length": 1081.0,
      "entropy": 0.3030368685722351,
      "epoch": 4.00801282051282,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 691.9620971679688,
      "learning_rate": 1e-06,
      "loss": 0.1049,
      "num_tokens": 1454962379.0,
      "reward": 0.34921398758888245,
      "reward_std": 0.05184464529156685,
      "rewards/progression_diversity/mean": -0.02537970244884491,
      "rewards/progression_diversity/std": 0.08060704171657562,
      "rewards/symbolic_reward_accuracy/mean": 0.244140625,
      "rewards/symbolic_reward_accuracy/std": 0.42999663949012756,
      "rewards/symbolic_reward_partial_score/mean": 0.6792155504226685,
      "rewards/symbolic_reward_partial_score/std": 0.23507006466388702,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0067884922027588,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 43.07496643066406,
      "step": 2501
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.3297167718410492,
      "epoch": 4.009615384615385,
      "grad_norm": 0.07673323899507523,
      "learning_rate": 1e-06,
      "loss": 0.0893,
      "step": 2502
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.3251052349805832,
      "epoch": 4.011217948717949,
      "grad_norm": 0.011674604378640652,
      "learning_rate": 1e-06,
      "loss": 0.1128,
      "step": 2503
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.3353799283504486,
      "epoch": 4.012820512820513,
      "grad_norm": 0.029069218784570694,
      "learning_rate": 1e-06,
      "loss": 0.0488,
      "step": 2504
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3116.0,
      "completions/mean_length": 3221.0078125,
      "completions/mean_terminated_length": 1983.461669921875,
      "completions/min_length": 1074.0,
      "completions/min_terminated_length": 1074.0,
      "entropy": 0.3239881694316864,
      "epoch": 4.014423076923077,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 496.55010986328125,
      "learning_rate": 1e-06,
      "loss": 0.0835,
      "num_tokens": 1457508255.0,
      "reward": 0.38050615787506104,
      "reward_std": 0.04678242653608322,
      "rewards/progression_diversity/mean": -0.02165273018181324,
      "rewards/progression_diversity/std": 0.07281038910150528,
      "rewards/symbolic_reward_accuracy/mean": 0.265625,
      "rewards/symbolic_reward_accuracy/std": 0.44209739565849304,
      "rewards/symbolic_reward_partial_score/mean": 0.740429699420929,
      "rewards/symbolic_reward_partial_score/std": 0.20741601288318634,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0144760608673096,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 38.09929656982422,
      "step": 2505
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.29394128918647766,
      "epoch": 4.016025641025641,
      "grad_norm": 0.03365239128470421,
      "learning_rate": 1e-06,
      "loss": 0.1895,
      "step": 2506
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.3248167932033539,
      "epoch": 4.017628205128205,
      "grad_norm": 0.03242243453860283,
      "learning_rate": 1e-06,
      "loss": 0.0367,
      "step": 2507
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.3440161794424057,
      "epoch": 4.019230769230769,
      "grad_norm": 0.023058878257870674,
      "learning_rate": 1e-06,
      "loss": 0.0453,
      "step": 2508
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.056640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3390.0,
      "completions/mean_length": 2804.220703125,
      "completions/mean_terminated_length": 1988.8717041015625,
      "completions/min_length": 1169.0,
      "completions/min_terminated_length": 1169.0,
      "entropy": 0.3260979652404785,
      "epoch": 4.020833333333333,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 859.6063842773438,
      "learning_rate": 1e-06,
      "loss": 0.0866,
      "num_tokens": 1459937296.0,
      "reward": 0.23110386729240417,
      "reward_std": 0.02382086217403412,
      "rewards/progression_diversity/mean": -0.015101805329322815,
      "rewards/progression_diversity/std": 0.06296249479055405,
      "rewards/symbolic_reward_accuracy/mean": 0.05859375,
      "rewards/symbolic_reward_accuracy/std": 0.23509246110916138,
      "rewards/symbolic_reward_partial_score/mean": 0.6582194566726685,
      "rewards/symbolic_reward_partial_score/std": 0.18645817041397095,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.035254716873169,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 27.520709991455078,
      "step": 2509
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.34093551337718964,
      "epoch": 4.022435897435898,
      "grad_norm": 2172.674560546875,
      "learning_rate": 1e-06,
      "loss": 0.2025,
      "step": 2510
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.3240719437599182,
      "epoch": 4.024038461538462,
      "grad_norm": 67.2519760131836,
      "learning_rate": 1e-06,
      "loss": 0.0682,
      "step": 2511
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.3448804020881653,
      "epoch": 4.0256410256410255,
      "grad_norm": 0.01443031057715416,
      "learning_rate": 1e-06,
      "loss": 0.0467,
      "step": 2512
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.033203125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3084.0,
      "completions/mean_length": 2421.150390625,
      "completions/mean_terminated_length": 1941.6182861328125,
      "completions/min_length": 1169.0,
      "completions/min_terminated_length": 1169.0,
      "entropy": 0.3457055538892746,
      "epoch": 4.027243589743589,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 373.1669616699219,
      "learning_rate": 1e-06,
      "loss": 0.0384,
      "num_tokens": 1462033373.0,
      "reward": 0.36610591411590576,
      "reward_std": 0.008678528480231762,
      "rewards/progression_diversity/mean": -0.008554144762456417,
      "rewards/progression_diversity/std": 0.04716269671916962,
      "rewards/symbolic_reward_accuracy/mean": 0.25,
      "rewards/symbolic_reward_accuracy/std": 0.43343618512153625,
      "rewards/symbolic_reward_partial_score/mean": 0.7212890386581421,
      "rewards/symbolic_reward_partial_score/std": 0.2121674120426178,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0559264421463013,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 17.93334197998047,
      "step": 2513
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.0859375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.35698336362838745,
      "epoch": 4.028846153846154,
      "grad_norm": 0.10000839084386826,
      "learning_rate": 1e-06,
      "loss": 0.022,
      "step": 2514
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1796875,
      "entropy": 0.3595133423805237,
      "epoch": 4.030448717948718,
      "grad_norm": 0.02287139743566513,
      "learning_rate": 1e-06,
      "loss": 0.0303,
      "step": 2515
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.34900110960006714,
      "epoch": 4.032051282051282,
      "grad_norm": 0.011496487073600292,
      "learning_rate": 1e-06,
      "loss": 0.0666,
      "step": 2516
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3091.0,
      "completions/mean_length": 2512.10546875,
      "completions/mean_terminated_length": 1977.48876953125,
      "completions/min_length": 1222.0,
      "completions/min_terminated_length": 1222.0,
      "entropy": 0.3486749231815338,
      "epoch": 4.033653846153846,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 842.0504150390625,
      "learning_rate": 1e-06,
      "loss": 0.0281,
      "num_tokens": 1464206211.0,
      "reward": 0.31601575016975403,
      "reward_std": 0.02459620125591755,
      "rewards/progression_diversity/mean": -0.00828959047794342,
      "rewards/progression_diversity/std": 0.04331068694591522,
      "rewards/symbolic_reward_accuracy/mean": 0.154296875,
      "rewards/symbolic_reward_accuracy/std": 0.36158639192581177,
      "rewards/symbolic_reward_partial_score/mean": 0.7470214366912842,
      "rewards/symbolic_reward_partial_score/std": 0.18681012094020844,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0497366189956665,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 21.2126522064209,
      "step": 2517
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3538134843111038,
      "epoch": 4.035256410256411,
      "grad_norm": 0.03899611532688141,
      "learning_rate": 1e-06,
      "loss": 0.0486,
      "step": 2518
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.3479982316493988,
      "epoch": 4.0368589743589745,
      "grad_norm": 0.013328113593161106,
      "learning_rate": 1e-06,
      "loss": 0.0789,
      "step": 2519
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.35109949111938477,
      "epoch": 4.038461538461538,
      "grad_norm": 0.015972906723618507,
      "learning_rate": 1e-06,
      "loss": 0.0823,
      "step": 2520
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3298.0,
      "completions/mean_length": 2378.287109375,
      "completions/mean_terminated_length": 1867.95751953125,
      "completions/min_length": 1063.0,
      "completions/min_terminated_length": 1063.0,
      "entropy": 0.35145680606365204,
      "epoch": 4.040064102564102,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 592.3827514648438,
      "learning_rate": 1e-06,
      "loss": 0.0507,
      "num_tokens": 1466305798.0,
      "reward": 0.3080458343029022,
      "reward_std": 0.013700053095817566,
      "rewards/progression_diversity/mean": -0.00889340415596962,
      "rewards/progression_diversity/std": 0.04732125252485275,
      "rewards/symbolic_reward_accuracy/mean": 0.154296875,
      "rewards/symbolic_reward_accuracy/std": 0.36158639192581177,
      "rewards/symbolic_reward_partial_score/mean": 0.7198241949081421,
      "rewards/symbolic_reward_partial_score/std": 0.19643016159534454,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0504951477050781,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 20.981121063232422,
      "step": 2521
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.15625,
      "entropy": 0.3594615310430527,
      "epoch": 4.041666666666667,
      "grad_norm": 0.03896607458591461,
      "learning_rate": 1e-06,
      "loss": 0.017,
      "step": 2522
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.34222954511642456,
      "epoch": 4.043269230769231,
      "grad_norm": 0.006857524160295725,
      "learning_rate": 1e-06,
      "loss": 0.0644,
      "step": 2523
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1484375,
      "entropy": 0.35084451735019684,
      "epoch": 4.044871794871795,
      "grad_norm": 0.009686310775578022,
      "learning_rate": 1e-06,
      "loss": 0.0258,
      "step": 2524
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.04296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2915.0,
      "completions/mean_length": 2450.3046875,
      "completions/mean_terminated_length": 1824.7100830078125,
      "completions/min_length": 1064.0,
      "completions/min_terminated_length": 1064.0,
      "entropy": 0.3546312600374222,
      "epoch": 4.046474358974359,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 310.8623962402344,
      "learning_rate": 1e-06,
      "loss": -0.0016,
      "num_tokens": 1468479282.0,
      "reward": 0.38561779260635376,
      "reward_std": 0.018528560176491737,
      "rewards/progression_diversity/mean": -0.010976451449096203,
      "rewards/progression_diversity/std": 0.05261799693107605,
      "rewards/symbolic_reward_accuracy/mean": 0.265625,
      "rewards/symbolic_reward_accuracy/std": 0.44209739565849304,
      "rewards/symbolic_reward_partial_score/mean": 0.7551594972610474,
      "rewards/symbolic_reward_partial_score/std": 0.20599855482578278,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0471153259277344,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 22.519685745239258,
      "step": 2525
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.3387318253517151,
      "epoch": 4.048076923076923,
      "grad_norm": 0.017549673095345497,
      "learning_rate": 1e-06,
      "loss": 0.1121,
      "step": 2526
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.3566637337207794,
      "epoch": 4.049679487179487,
      "grad_norm": 0.025295212864875793,
      "learning_rate": 1e-06,
      "loss": 0.0235,
      "step": 2527
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.0625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.109375,
      "entropy": 0.33428405225276947,
      "epoch": 4.051282051282051,
      "grad_norm": 0.019428903236985207,
      "learning_rate": 1e-06,
      "loss": 0.0258,
      "step": 2528
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.091796875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2715.0,
      "completions/mean_length": 3127.302734375,
      "completions/mean_terminated_length": 1787.3785400390625,
      "completions/min_length": 1011.0,
      "completions/min_terminated_length": 1011.0,
      "entropy": 0.3295024484395981,
      "epoch": 4.052884615384615,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 1348.5477294921875,
      "learning_rate": 1e-06,
      "loss": 0.057,
      "num_tokens": 1470886541.0,
      "reward": 0.28540563583374023,
      "reward_std": 0.019026556983590126,
      "rewards/progression_diversity/mean": -0.02584357187151909,
      "rewards/progression_diversity/std": 0.08347862958908081,
      "rewards/symbolic_reward_accuracy/mean": 0.119140625,
      "rewards/symbolic_reward_accuracy/std": 0.32427072525024414,
      "rewards/symbolic_reward_partial_score/mean": 0.7152343988418579,
      "rewards/symbolic_reward_partial_score/std": 0.19548086822032928,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.010728359222412,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 41.08893585205078,
      "step": 2529
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.30960477888584137,
      "epoch": 4.05448717948718,
      "grad_norm": 96.72750091552734,
      "learning_rate": 1e-06,
      "loss": 0.0919,
      "step": 2530
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.32376880943775177,
      "epoch": 4.056089743589744,
      "grad_norm": 0.019057314842939377,
      "learning_rate": 1e-06,
      "loss": 0.0911,
      "step": 2531
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.32927900552749634,
      "epoch": 4.0576923076923075,
      "grad_norm": 0.010551064275205135,
      "learning_rate": 1e-06,
      "loss": 0.1108,
      "step": 2532
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2729.0,
      "completions/mean_length": 2688.408203125,
      "completions/mean_terminated_length": 1775.368896484375,
      "completions/min_length": 935.0,
      "completions/min_terminated_length": 935.0,
      "entropy": 0.34093132615089417,
      "epoch": 4.059294871794871,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 454.11712646484375,
      "learning_rate": 1e-06,
      "loss": 0.0975,
      "num_tokens": 1473027214.0,
      "reward": 0.42107725143432617,
      "reward_std": 0.030925702303647995,
      "rewards/progression_diversity/mean": -0.01678628847002983,
      "rewards/progression_diversity/std": 0.06638278067111969,
      "rewards/symbolic_reward_accuracy/mean": 0.3046875,
      "rewards/symbolic_reward_accuracy/std": 0.4607250988483429,
      "rewards/symbolic_reward_partial_score/mean": 0.7967284917831421,
      "rewards/symbolic_reward_partial_score/std": 0.18162044882774353,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0306940078735352,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 30.967906951904297,
      "step": 2533
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.32887102663517,
      "epoch": 4.060897435897436,
      "grad_norm": 0.00896961148828268,
      "learning_rate": 1e-06,
      "loss": 0.1063,
      "step": 2534
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.3550725132226944,
      "epoch": 4.0625,
      "grad_norm": 0.014958408661186695,
      "learning_rate": 1e-06,
      "loss": 0.069,
      "step": 2535
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.34709444642066956,
      "epoch": 4.064102564102564,
      "grad_norm": 0.013090104795992374,
      "learning_rate": 1e-06,
      "loss": 0.0225,
      "step": 2536
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.087890625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2727.0,
      "completions/mean_length": 3024.65625,
      "completions/mean_terminated_length": 1737.353271484375,
      "completions/min_length": 1045.0,
      "completions/min_terminated_length": 1045.0,
      "entropy": 0.3089066445827484,
      "epoch": 4.065705128205129,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 508.8141174316406,
      "learning_rate": 1e-06,
      "loss": 0.0951,
      "num_tokens": 1475532462.0,
      "reward": 0.31843122839927673,
      "reward_std": 0.011976012028753757,
      "rewards/progression_diversity/mean": -0.025041041895747185,
      "rewards/progression_diversity/std": 0.08240573108196259,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.6918293833732605,
      "rewards/symbolic_reward_partial_score/std": 0.20478877425193787,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0157009363174438,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 37.54655075073242,
      "step": 2537
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.31688109040260315,
      "epoch": 4.0673076923076925,
      "grad_norm": 1613.686279296875,
      "learning_rate": 1e-06,
      "loss": 0.1161,
      "step": 2538
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.30626846849918365,
      "epoch": 4.068910256410256,
      "grad_norm": 0.013836408033967018,
      "learning_rate": 1e-06,
      "loss": 0.0784,
      "step": 2539
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.3128751218318939,
      "epoch": 4.07051282051282,
      "grad_norm": 0.025374021381139755,
      "learning_rate": 1e-06,
      "loss": 0.0991,
      "step": 2540
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.076171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2768.0,
      "completions/mean_length": 2856.404296875,
      "completions/mean_terminated_length": 1741.0211181640625,
      "completions/min_length": 999.0,
      "completions/min_terminated_length": 999.0,
      "entropy": 0.3115910291671753,
      "epoch": 4.072115384615385,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 627.9964599609375,
      "learning_rate": 1e-06,
      "loss": 0.1078,
      "num_tokens": 1477984861.0,
      "reward": 0.29800140857696533,
      "reward_std": 0.011926448903977871,
      "rewards/progression_diversity/mean": -0.02163849025964737,
      "rewards/progression_diversity/std": 0.07751591503620148,
      "rewards/symbolic_reward_accuracy/mean": 0.123046875,
      "rewards/symbolic_reward_accuracy/std": 0.32881227135658264,
      "rewards/symbolic_reward_partial_score/mean": 0.7486165761947632,
      "rewards/symbolic_reward_partial_score/std": 0.1641637235879898,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0106605291366577,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 40.304405212402344,
      "step": 2541
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3093924969434738,
      "epoch": 4.073717948717949,
      "grad_norm": 0.008785123936831951,
      "learning_rate": 1e-06,
      "loss": 0.0982,
      "step": 2542
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.33030666410923004,
      "epoch": 4.075320512820513,
      "grad_norm": 0.017673548310995102,
      "learning_rate": 1e-06,
      "loss": 0.0605,
      "step": 2543
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.3170679658651352,
      "epoch": 4.076923076923077,
      "grad_norm": 0.013326681219041348,
      "learning_rate": 1e-06,
      "loss": 0.0929,
      "step": 2544
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.072265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2629.0,
      "completions/mean_length": 2784.251953125,
      "completions/mean_terminated_length": 1724.903076171875,
      "completions/min_length": 955.0,
      "completions/min_terminated_length": 955.0,
      "entropy": 0.3214297592639923,
      "epoch": 4.078525641025641,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 304.0992431640625,
      "learning_rate": 1e-06,
      "loss": 0.076,
      "num_tokens": 1480289838.0,
      "reward": 0.3872213363647461,
      "reward_std": 0.013941626995801926,
      "rewards/progression_diversity/mean": -0.0229843370616436,
      "rewards/progression_diversity/std": 0.08313606679439545,
      "rewards/symbolic_reward_accuracy/mean": 0.24609375,
      "rewards/symbolic_reward_accuracy/std": 0.4311550557613373,
      "rewards/symbolic_reward_partial_score/mean": 0.79931640625,
      "rewards/symbolic_reward_partial_score/std": 0.18186363577842712,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0218315124511719,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 34.574668884277344,
      "step": 2545
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.3198620527982712,
      "epoch": 4.080128205128205,
      "grad_norm": 0.009517377242445946,
      "learning_rate": 1e-06,
      "loss": 0.0768,
      "step": 2546
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.30512526631355286,
      "epoch": 4.081730769230769,
      "grad_norm": 0.018004372715950012,
      "learning_rate": 1e-06,
      "loss": 0.1049,
      "step": 2547
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.3438553065061569,
      "epoch": 4.083333333333333,
      "grad_norm": 0.025911420583724976,
      "learning_rate": 1e-06,
      "loss": 0.059,
      "step": 2548
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.076171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2753.0,
      "completions/mean_length": 2879.19921875,
      "completions/mean_terminated_length": 1765.695556640625,
      "completions/min_length": 1122.0,
      "completions/min_terminated_length": 1122.0,
      "entropy": 0.3178988993167877,
      "epoch": 4.084935897435898,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 239.32733154296875,
      "learning_rate": 1e-06,
      "loss": 0.0345,
      "num_tokens": 1482682692.0,
      "reward": 0.35247373580932617,
      "reward_std": 0.008235493674874306,
      "rewards/progression_diversity/mean": -0.025090141221880913,
      "rewards/progression_diversity/std": 0.08833190053701401,
      "rewards/symbolic_reward_accuracy/mean": 0.21875,
      "rewards/symbolic_reward_accuracy/std": 0.41380295157432556,
      "rewards/symbolic_reward_partial_score/mean": 0.7382487058639526,
      "rewards/symbolic_reward_partial_score/std": 0.19196312129497528,
      "rewards/tag_count_reward/mean": 0.0,
      "rewards/tag_count_reward/std": 0.0,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0059094429016113,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 43.50497055053711,
      "step": 2549
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.32996954023838043,
      "epoch": 4.086538461538462,
      "grad_norm": 0.013742972165346146,
      "learning_rate": 1e-06,
      "loss": 0.0517,
      "step": 2550
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.31920306384563446,
      "epoch": 4.0881410256410255,
      "grad_norm": 0.011623159982264042,
      "learning_rate": 1e-06,
      "loss": 0.1203,
      "step": 2551
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.30411607027053833,
      "epoch": 4.089743589743589,
      "grad_norm": 0.01574169658124447,
      "learning_rate": 1e-06,
      "loss": 0.1498,
      "step": 2552
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.041015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3040.0,
      "completions/mean_length": 2363.705078125,
      "completions/mean_terminated_length": 1764.0592041015625,
      "completions/min_length": 1029.0,
      "completions/min_terminated_length": 1029.0,
      "entropy": 0.3521651029586792,
      "epoch": 4.091346153846154,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 662.1588745117188,
      "learning_rate": 1e-06,
      "loss": 0.0482,
      "num_tokens": 1484783181.0,
      "reward": 0.26874637603759766,
      "reward_std": 0.013555062934756279,
      "rewards/progression_diversity/mean": -0.014036715030670166,
      "rewards/progression_diversity/std": 0.06895329058170319,
      "rewards/symbolic_reward_accuracy/mean": 0.09375,
      "rewards/symbolic_reward_accuracy/std": 0.29176566004753113,
      "rewards/symbolic_reward_partial_score/mean": 0.7100911140441895,
      "rewards/symbolic_reward_partial_score/std": 0.17183727025985718,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0332553386688232,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 29.85314178466797,
      "step": 2553
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.34595395624637604,
      "epoch": 4.092948717948718,
      "grad_norm": 4.502384185791016,
      "learning_rate": 1e-06,
      "loss": 0.0216,
      "step": 2554
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.3410833477973938,
      "epoch": 4.094551282051282,
      "grad_norm": 0.011607046239078045,
      "learning_rate": 1e-06,
      "loss": 0.0667,
      "step": 2555
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.33509866893291473,
      "epoch": 4.096153846153846,
      "grad_norm": 0.014269710518419743,
      "learning_rate": 1e-06,
      "loss": 0.0963,
      "step": 2556
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3428.0,
      "completions/mean_length": 2880.39453125,
      "completions/mean_terminated_length": 1859.113525390625,
      "completions/min_length": 825.0,
      "completions/min_terminated_length": 825.0,
      "entropy": 0.32943619787693024,
      "epoch": 4.097756410256411,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 1257.3856201171875,
      "learning_rate": 1e-06,
      "loss": 0.0543,
      "num_tokens": 1487087367.0,
      "reward": 0.3605473041534424,
      "reward_std": 0.030462127178907394,
      "rewards/progression_diversity/mean": -0.02095445990562439,
      "rewards/progression_diversity/std": 0.07725287228822708,
      "rewards/symbolic_reward_accuracy/mean": 0.23046875,
      "rewards/symbolic_reward_accuracy/std": 0.42154473066329956,
      "rewards/symbolic_reward_partial_score/mean": 0.7422363758087158,
      "rewards/symbolic_reward_partial_score/std": 0.19137205183506012,
      "rewards/tag_count_reward/mean": -0.001953125,
      "rewards/tag_count_reward/std": 0.04419417306780815,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0257246494293213,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 34.513587951660156,
      "step": 2557
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3395875543355942,
      "epoch": 4.0993589743589745,
      "grad_norm": 1.989341139793396,
      "learning_rate": 1e-06,
      "loss": 0.0551,
      "step": 2558
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.33270154893398285,
      "epoch": 4.100961538461538,
      "grad_norm": 0.0756467655301094,
      "learning_rate": 1e-06,
      "loss": 0.1137,
      "step": 2559
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.3363744169473648,
      "epoch": 4.102564102564102,
      "grad_norm": 0.012534077279269695,
      "learning_rate": 1e-06,
      "loss": 0.0671,
      "step": 2560
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.068359375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3092.0,
      "completions/mean_length": 2916.15234375,
      "completions/mean_terminated_length": 1927.9454345703125,
      "completions/min_length": 1158.0,
      "completions/min_terminated_length": 1158.0,
      "entropy": 0.33795928955078125,
      "epoch": 4.104166666666667,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 486.66265869140625,
      "learning_rate": 1e-06,
      "loss": 0.0406,
      "num_tokens": 1489486389.0,
      "reward": 0.3729179799556732,
      "reward_std": 0.019203029572963715,
      "rewards/progression_diversity/mean": -0.019237220287322998,
      "rewards/progression_diversity/std": 0.07287276536226273,
      "rewards/symbolic_reward_accuracy/mean": 0.248046875,
      "rewards/symbolic_reward_accuracy/std": 0.4323015511035919,
      "rewards/symbolic_reward_partial_score/mean": 0.7495605945587158,
      "rewards/symbolic_reward_partial_score/std": 0.2092960774898529,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0225183963775635,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 35.221961975097656,
      "step": 2561
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.33110561966896057,
      "epoch": 4.105769230769231,
      "grad_norm": 13.403287887573242,
      "learning_rate": 1e-06,
      "loss": 0.0302,
      "step": 2562
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.3245306611061096,
      "epoch": 4.107371794871795,
      "grad_norm": 0.013325260020792484,
      "learning_rate": 1e-06,
      "loss": 0.1297,
      "step": 2563
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3338320553302765,
      "epoch": 4.108974358974359,
      "grad_norm": 1.518842101097107,
      "learning_rate": 1e-06,
      "loss": 0.0908,
      "step": 2564
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.107421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3086.0,
      "completions/mean_length": 3535.826171875,
      "completions/mean_terminated_length": 1989.547119140625,
      "completions/min_length": 1126.0,
      "completions/min_terminated_length": 1126.0,
      "entropy": 0.31208691000938416,
      "epoch": 4.110576923076923,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 2256.84619140625,
      "learning_rate": 1e-06,
      "loss": 0.1228,
      "num_tokens": 1492263788.0,
      "reward": 0.3425925374031067,
      "reward_std": 0.025561640039086342,
      "rewards/progression_diversity/mean": -0.030298512428998947,
      "rewards/progression_diversity/std": 0.08967959880828857,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.7237141728401184,
      "rewards/symbolic_reward_partial_score/std": 0.20818644762039185,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9913551807403564,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 50.347373962402344,
      "step": 2565
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.3200324475765228,
      "epoch": 4.112179487179487,
      "grad_norm": 1.371201515197754,
      "learning_rate": 1e-06,
      "loss": 0.1,
      "step": 2566
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.30504705011844635,
      "epoch": 4.113782051282051,
      "grad_norm": 2.302720308303833,
      "learning_rate": 1e-06,
      "loss": 0.1468,
      "step": 2567
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.3112814277410507,
      "epoch": 4.115384615384615,
      "grad_norm": 7566.38720703125,
      "learning_rate": 1e-06,
      "loss": 0.2698,
      "step": 2568
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.080078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3291.0,
      "completions/mean_length": 3172.8828125,
      "completions/mean_terminated_length": 2022.87060546875,
      "completions/min_length": 1223.0,
      "completions/min_terminated_length": 1223.0,
      "entropy": 0.35531145334243774,
      "epoch": 4.11698717948718,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 797.2550048828125,
      "learning_rate": 1e-06,
      "loss": 0.0721,
      "num_tokens": 1494671600.0,
      "reward": 0.32184159755706787,
      "reward_std": 0.0444670133292675,
      "rewards/progression_diversity/mean": -0.02189837582409382,
      "rewards/progression_diversity/std": 0.07626625150442123,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.7167643308639526,
      "rewards/symbolic_reward_partial_score/std": 0.23638619482517242,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.007836103439331,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 40.60625457763672,
      "step": 2569
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.3130183517932892,
      "epoch": 4.118589743589744,
      "grad_norm": 21.2623291015625,
      "learning_rate": 1e-06,
      "loss": 0.0972,
      "step": 2570
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.3290316015481949,
      "epoch": 4.1201923076923075,
      "grad_norm": 1913.60205078125,
      "learning_rate": 1e-06,
      "loss": 0.2226,
      "step": 2571
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.3343227505683899,
      "epoch": 4.121794871794871,
      "grad_norm": 0.022048160433769226,
      "learning_rate": 1e-06,
      "loss": 0.1135,
      "step": 2572
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.09375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3234.0,
      "completions/mean_length": 3387.615234375,
      "completions/mean_terminated_length": 2043.16162109375,
      "completions/min_length": 911.0,
      "completions/min_terminated_length": 911.0,
      "entropy": 0.3383503258228302,
      "epoch": 4.123397435897436,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 572.839111328125,
      "learning_rate": 1e-06,
      "loss": 0.1123,
      "num_tokens": 1497270795.0,
      "reward": 0.41935285925865173,
      "reward_std": 0.022548135370016098,
      "rewards/progression_diversity/mean": -0.02565228007733822,
      "rewards/progression_diversity/std": 0.08250828087329865,
      "rewards/symbolic_reward_accuracy/mean": 0.310546875,
      "rewards/symbolic_reward_accuracy/std": 0.46317005157470703,
      "rewards/symbolic_reward_partial_score/mean": 0.7795573472976685,
      "rewards/symbolic_reward_partial_score/std": 0.20047199726104736,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0107229948043823,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 37.779056549072266,
      "step": 2573
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.3201807737350464,
      "epoch": 4.125,
      "grad_norm": 0.01932665705680847,
      "learning_rate": 1e-06,
      "loss": 0.1213,
      "step": 2574
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.31810304522514343,
      "epoch": 4.126602564102564,
      "grad_norm": 0.022387662902474403,
      "learning_rate": 1e-06,
      "loss": 0.1087,
      "step": 2575
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.3269665241241455,
      "epoch": 4.128205128205128,
      "grad_norm": 0.013768312521278858,
      "learning_rate": 1e-06,
      "loss": 0.1118,
      "step": 2576
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3283.0,
      "completions/mean_length": 2913.330078125,
      "completions/mean_terminated_length": 2015.2855224609375,
      "completions/min_length": 1204.0,
      "completions/min_terminated_length": 1204.0,
      "entropy": 0.3411347270011902,
      "epoch": 4.1298076923076925,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 542.296875,
      "learning_rate": 1e-06,
      "loss": 0.0579,
      "num_tokens": 1499596596.0,
      "reward": 0.2602491080760956,
      "reward_std": 0.025792162865400314,
      "rewards/progression_diversity/mean": -0.015127346850931644,
      "rewards/progression_diversity/std": 0.06053609028458595,
      "rewards/symbolic_reward_accuracy/mean": 0.095703125,
      "rewards/symbolic_reward_accuracy/std": 0.2944713830947876,
      "rewards/symbolic_reward_partial_score/mean": 0.6778970956802368,
      "rewards/symbolic_reward_partial_score/std": 0.19333595037460327,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0344008207321167,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 27.88906478881836,
      "step": 2577
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.3439747095108032,
      "epoch": 4.131410256410256,
      "grad_norm": 0.02023777738213539,
      "learning_rate": 1e-06,
      "loss": 0.0927,
      "step": 2578
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.34476563334465027,
      "epoch": 4.13301282051282,
      "grad_norm": 7.5225725173950195,
      "learning_rate": 1e-06,
      "loss": 0.0408,
      "step": 2579
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.3402794599533081,
      "epoch": 4.134615384615385,
      "grad_norm": 0.021249419078230858,
      "learning_rate": 1e-06,
      "loss": 0.0551,
      "step": 2580
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.048828125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3805.0,
      "completions/mean_length": 2671.884765625,
      "completions/mean_terminated_length": 1967.9775390625,
      "completions/min_length": 597.0,
      "completions/min_terminated_length": 597.0,
      "entropy": 0.33940988779067993,
      "epoch": 4.136217948717949,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 498.6994323730469,
      "learning_rate": 1e-06,
      "loss": 0.0785,
      "num_tokens": 1501793977.0,
      "reward": 0.3605765700340271,
      "reward_std": 0.06087111681699753,
      "rewards/progression_diversity/mean": -0.01119298581033945,
      "rewards/progression_diversity/std": 0.05226728320121765,
      "rewards/symbolic_reward_accuracy/mean": 0.2265625,
      "rewards/symbolic_reward_accuracy/std": 0.4190165400505066,
      "rewards/symbolic_reward_partial_score/mean": 0.7517740726470947,
      "rewards/symbolic_reward_partial_score/std": 0.18635277450084686,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0438597202301025,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 23.59453582763672,
      "step": 2581
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3429712653160095,
      "epoch": 4.137820512820513,
      "grad_norm": 472.5995788574219,
      "learning_rate": 1e-06,
      "loss": 0.0979,
      "step": 2582
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.3512907773256302,
      "epoch": 4.139423076923077,
      "grad_norm": 0.022063203155994415,
      "learning_rate": 1e-06,
      "loss": 0.0575,
      "step": 2583
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.36512602865695953,
      "epoch": 4.141025641025641,
      "grad_norm": 0.027568161487579346,
      "learning_rate": 1e-06,
      "loss": 0.043,
      "step": 2584
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.056640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3768.0,
      "completions/mean_length": 2782.953125,
      "completions/mean_terminated_length": 1966.3271484375,
      "completions/min_length": 1182.0,
      "completions/min_terminated_length": 1182.0,
      "entropy": 0.33962003886699677,
      "epoch": 4.142628205128205,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1543.4376220703125,
      "learning_rate": 1e-06,
      "loss": 0.0607,
      "num_tokens": 1504024065.0,
      "reward": 0.3242529630661011,
      "reward_std": 0.023926857858896255,
      "rewards/progression_diversity/mean": -0.014647168107330799,
      "rewards/progression_diversity/std": 0.0613836906850338,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.716748058795929,
      "rewards/symbolic_reward_partial_score/std": 0.1955670863389969,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.031816005706787,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 30.275238037109375,
      "step": 2585
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.33402132987976074,
      "epoch": 4.144230769230769,
      "grad_norm": 144.89903259277344,
      "learning_rate": 1e-06,
      "loss": 1.3615,
      "step": 2586
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3640512526035309,
      "epoch": 4.145833333333333,
      "grad_norm": 0.025912059471011162,
      "learning_rate": 1e-06,
      "loss": 0.0157,
      "step": 2587
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.346598282456398,
      "epoch": 4.147435897435898,
      "grad_norm": 0.017162565141916275,
      "learning_rate": 1e-06,
      "loss": 0.065,
      "step": 2588
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.07421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3701.0,
      "completions/mean_length": 3073.26953125,
      "completions/mean_terminated_length": 2006.1644287109375,
      "completions/min_length": 1134.0,
      "completions/min_terminated_length": 1134.0,
      "entropy": 0.32446981966495514,
      "epoch": 4.149038461538462,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 570.8603515625,
      "learning_rate": 1e-06,
      "loss": 0.1123,
      "num_tokens": 1506499115.0,
      "reward": 0.3911179006099701,
      "reward_std": 0.021029271185398102,
      "rewards/progression_diversity/mean": -0.018583200871944427,
      "rewards/progression_diversity/std": 0.06753162294626236,
      "rewards/symbolic_reward_accuracy/mean": 0.27734375,
      "rewards/symbolic_reward_accuracy/std": 0.4481254518032074,
      "rewards/symbolic_reward_partial_score/mean": 0.7516113519668579,
      "rewards/symbolic_reward_partial_score/std": 0.21333743631839752,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0214645862579346,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 35.46625900268555,
      "step": 2589
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.35942013561725616,
      "epoch": 4.1506410256410255,
      "grad_norm": 4977.77294921875,
      "learning_rate": 1e-06,
      "loss": 0.2231,
      "step": 2590
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.3404110074043274,
      "epoch": 4.152243589743589,
      "grad_norm": 97.81608581542969,
      "learning_rate": 1e-06,
      "loss": 0.1382,
      "step": 2591
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.3034721314907074,
      "epoch": 4.153846153846154,
      "grad_norm": 10592.0458984375,
      "learning_rate": 1e-06,
      "loss": 0.6398,
      "step": 2592
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3079.0,
      "completions/mean_length": 3393.611328125,
      "completions/mean_terminated_length": 1925.132568359375,
      "completions/min_length": 1143.0,
      "completions/min_terminated_length": 1143.0,
      "entropy": 0.314235121011734,
      "epoch": 4.155448717948718,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 377.22088623046875,
      "learning_rate": 1e-06,
      "loss": 0.0613,
      "num_tokens": 1509122212.0,
      "reward": 0.3483143150806427,
      "reward_std": 0.027692176401615143,
      "rewards/progression_diversity/mean": -0.024526942521333694,
      "rewards/progression_diversity/std": 0.07418946921825409,
      "rewards/symbolic_reward_accuracy/mean": 0.216796875,
      "rewards/symbolic_reward_accuracy/std": 0.4124660789966583,
      "rewards/symbolic_reward_partial_score/mean": 0.7308756113052368,
      "rewards/symbolic_reward_partial_score/std": 0.21280378103256226,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.010713815689087,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 40.72765350341797,
      "step": 2593
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.3067990243434906,
      "epoch": 4.157051282051282,
      "grad_norm": 645088.0625,
      "learning_rate": 1e-06,
      "loss": 45.8723,
      "step": 2594
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.3317181318998337,
      "epoch": 4.158653846153846,
      "grad_norm": 10997.2705078125,
      "learning_rate": 1e-06,
      "loss": 1.253,
      "step": 2595
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.3151821494102478,
      "epoch": 4.160256410256411,
      "grad_norm": 50563.86328125,
      "learning_rate": 1e-06,
      "loss": 9.5746,
      "step": 2596
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.19140625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3123.0,
      "completions/mean_length": 4673.439453125,
      "completions/mean_terminated_length": 1901.3743896484375,
      "completions/min_length": 1096.0,
      "completions/min_terminated_length": 1096.0,
      "entropy": 0.2506090775132179,
      "epoch": 4.1618589743589745,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 1328.9156494140625,
      "learning_rate": 1e-06,
      "loss": 0.1149,
      "num_tokens": 1512400453.0,
      "reward": 0.44010183215141296,
      "reward_std": 0.06620557606220245,
      "rewards/progression_diversity/mean": -0.042061757296323776,
      "rewards/progression_diversity/std": 0.08827082067728043,
      "rewards/symbolic_reward_accuracy/mean": 0.361328125,
      "rewards/symbolic_reward_accuracy/std": 0.48085519671440125,
      "rewards/symbolic_reward_partial_score/mean": 0.7477050423622131,
      "rewards/symbolic_reward_partial_score/std": 0.23697598278522491,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9721221923828125,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 55.41648864746094,
      "step": 2597
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.24983707815408707,
      "epoch": 4.163461538461538,
      "grad_norm": 100176.0,
      "learning_rate": 1e-06,
      "loss": 29.2153,
      "step": 2598
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2740893214941025,
      "epoch": 4.165064102564102,
      "grad_norm": 18222.609375,
      "learning_rate": 1e-06,
      "loss": 1.2944,
      "step": 2599
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.3012789934873581,
      "epoch": 4.166666666666667,
      "grad_norm": 35.68260192871094,
      "learning_rate": 1e-06,
      "loss": 0.0814,
      "step": 2600
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.208984375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2746.0,
      "completions/mean_length": 4911.619140625,
      "completions/mean_terminated_length": 1880.64453125,
      "completions/min_length": 1124.0,
      "completions/min_terminated_length": 1124.0,
      "entropy": 0.24826913326978683,
      "epoch": 4.168269230769231,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 513.0524291992188,
      "learning_rate": 1e-06,
      "loss": 0.1546,
      "num_tokens": 1515790834.0,
      "reward": 0.3871710002422333,
      "reward_std": 0.04217392951250076,
      "rewards/progression_diversity/mean": -0.0426679290831089,
      "rewards/progression_diversity/std": 0.08528563380241394,
      "rewards/symbolic_reward_accuracy/mean": 0.27734375,
      "rewards/symbolic_reward_accuracy/std": 0.4481254518032074,
      "rewards/symbolic_reward_partial_score/mean": 0.7399088740348816,
      "rewards/symbolic_reward_partial_score/std": 0.21892520785331726,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9704339504241943,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 54.93540573120117,
      "step": 2601
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2412198781967163,
      "epoch": 4.169871794871795,
      "grad_norm": 26214.958984375,
      "learning_rate": 1e-06,
      "loss": 2.6327,
      "step": 2602
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.29308077692985535,
      "epoch": 4.171474358974359,
      "grad_norm": 19403.22265625,
      "learning_rate": 1e-06,
      "loss": 0.3375,
      "step": 2603
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.2676732540130615,
      "epoch": 4.173076923076923,
      "grad_norm": 0.013621392659842968,
      "learning_rate": 1e-06,
      "loss": 0.1484,
      "step": 2604
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.20703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3097.0,
      "completions/mean_length": 4899.4921875,
      "completions/mean_terminated_length": 1901.0738525390625,
      "completions/min_length": 1051.0,
      "completions/min_terminated_length": 1051.0,
      "entropy": 0.2361399084329605,
      "epoch": 4.174679487179487,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 1089.3848876953125,
      "learning_rate": 1e-06,
      "loss": 0.1781,
      "num_tokens": 1519098398.0,
      "reward": 0.32966169714927673,
      "reward_std": 0.037758007645606995,
      "rewards/progression_diversity/mean": -0.039204493165016174,
      "rewards/progression_diversity/std": 0.07915779203176498,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.731689453125,
      "rewards/symbolic_reward_partial_score/std": 0.20694516599178314,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9641684889793396,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 58.79726028442383,
      "step": 2605
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.29851216077804565,
      "epoch": 4.176282051282051,
      "grad_norm": 0.012859205715358257,
      "learning_rate": 1e-06,
      "loss": 0.0952,
      "step": 2606
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.26364219933748245,
      "epoch": 4.177884615384615,
      "grad_norm": 0.01124588306993246,
      "learning_rate": 1e-06,
      "loss": 0.1605,
      "step": 2607
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.25536324083805084,
      "epoch": 4.17948717948718,
      "grad_norm": 0.014565329998731613,
      "learning_rate": 1e-06,
      "loss": 0.1756,
      "step": 2608
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.263671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3179.0,
      "completions/mean_length": 5711.21484375,
      "completions/mean_terminated_length": 1889.3951416015625,
      "completions/min_length": 939.0,
      "completions/min_terminated_length": 939.0,
      "entropy": 0.2375631034374237,
      "epoch": 4.181089743589744,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1342.1802978515625,
      "learning_rate": 1e-06,
      "loss": 0.1399,
      "num_tokens": 1522799932.0,
      "reward": 0.375043660402298,
      "reward_std": 0.06779703497886658,
      "rewards/progression_diversity/mean": -0.05374106019735336,
      "rewards/progression_diversity/std": 0.09237094968557358,
      "rewards/symbolic_reward_accuracy/mean": 0.265625,
      "rewards/symbolic_reward_accuracy/std": 0.44209739565849304,
      "rewards/symbolic_reward_partial_score/mean": 0.7239420413970947,
      "rewards/symbolic_reward_partial_score/std": 0.2347642481327057,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.932857871055603,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 76.66647338867188,
      "step": 2609
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2648901790380478,
      "epoch": 4.1826923076923075,
      "grad_norm": 23.981592178344727,
      "learning_rate": 1e-06,
      "loss": 0.1384,
      "step": 2610
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.2275170087814331,
      "epoch": 4.184294871794872,
      "grad_norm": 0.01362849585711956,
      "learning_rate": 1e-06,
      "loss": 0.2444,
      "step": 2611
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.515625,
      "entropy": 0.21743150800466537,
      "epoch": 4.185897435897436,
      "grad_norm": 0.012125054374337196,
      "learning_rate": 1e-06,
      "loss": 0.2923,
      "step": 2612
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.28125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3040.0,
      "completions/mean_length": 5987.09375,
      "completions/mean_terminated_length": 1918.7391357421875,
      "completions/min_length": 973.0,
      "completions/min_terminated_length": 973.0,
      "entropy": 0.24884148687124252,
      "epoch": 4.1875,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 962.3455810546875,
      "learning_rate": 1e-06,
      "loss": 0.1161,
      "num_tokens": 1526650092.0,
      "reward": 0.39309823513031006,
      "reward_std": 0.0952787920832634,
      "rewards/progression_diversity/mean": -0.05785566568374634,
      "rewards/progression_diversity/std": 0.09512782096862793,
      "rewards/symbolic_reward_accuracy/mean": 0.279296875,
      "rewards/symbolic_reward_accuracy/std": 0.44909247756004333,
      "rewards/symbolic_reward_partial_score/mean": 0.759521484375,
      "rewards/symbolic_reward_partial_score/std": 0.2274795025587082,
      "rewards/tag_count_reward/mean": -0.017578125,
      "rewards/tag_count_reward/std": 0.13154059648513794,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.934838056564331,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 76.90158081054688,
      "step": 2613
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.19819628447294235,
      "epoch": 4.189102564102564,
      "grad_norm": 79746.578125,
      "learning_rate": 1e-06,
      "loss": 0.2315,
      "step": 2614
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.23601322621107101,
      "epoch": 4.190705128205128,
      "grad_norm": 0.25907084345817566,
      "learning_rate": 1e-06,
      "loss": 0.1821,
      "step": 2615
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.23588327318429947,
      "epoch": 4.1923076923076925,
      "grad_norm": 0.07316934317350388,
      "learning_rate": 1e-06,
      "loss": 0.1802,
      "step": 2616
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.279296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 2953.0,
      "completions/mean_length": 5958.806640625,
      "completions/mean_terminated_length": 1918.6910400390625,
      "completions/min_length": 1163.0,
      "completions/min_terminated_length": 1163.0,
      "entropy": 0.21116046607494354,
      "epoch": 4.193910256410256,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1958.6832275390625,
      "learning_rate": 1e-06,
      "loss": 0.1101,
      "num_tokens": 1530549113.0,
      "reward": 0.3267575204372406,
      "reward_std": 0.08475440740585327,
      "rewards/progression_diversity/mean": -0.06009019538760185,
      "rewards/progression_diversity/std": 0.09868360310792923,
      "rewards/symbolic_reward_accuracy/mean": 0.193359375,
      "rewards/symbolic_reward_accuracy/std": 0.39531853795051575,
      "rewards/symbolic_reward_partial_score/mean": 0.7077311277389526,
      "rewards/symbolic_reward_partial_score/std": 0.24380625784397125,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9315183758735657,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 78.36473083496094,
      "step": 2617
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.22109153121709824,
      "epoch": 4.19551282051282,
      "grad_norm": 3.5450334548950195,
      "learning_rate": 1e-06,
      "loss": 0.2224,
      "step": 2618
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.24128204584121704,
      "epoch": 4.197115384615385,
      "grad_norm": 0.01579132489860058,
      "learning_rate": 1e-06,
      "loss": 0.1629,
      "step": 2619
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.22772854566574097,
      "epoch": 4.198717948717949,
      "grad_norm": 0.02087291143834591,
      "learning_rate": 1e-06,
      "loss": 0.0817,
      "step": 2620
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.279296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4516.0,
      "completions/mean_length": 5997.45703125,
      "completions/mean_terminated_length": 1972.31982421875,
      "completions/min_length": 1246.0,
      "completions/min_terminated_length": 1246.0,
      "entropy": 0.2100074663758278,
      "epoch": 4.200320512820513,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1239.959228515625,
      "learning_rate": 1e-06,
      "loss": 0.1074,
      "num_tokens": 1534550627.0,
      "reward": 0.26383617520332336,
      "reward_std": 0.05979030951857567,
      "rewards/progression_diversity/mean": -0.06218412518501282,
      "rewards/progression_diversity/std": 0.1025165319442749,
      "rewards/symbolic_reward_accuracy/mean": 0.103515625,
      "rewards/symbolic_reward_accuracy/std": 0.30492907762527466,
      "rewards/symbolic_reward_partial_score/mean": 0.6784017086029053,
      "rewards/symbolic_reward_partial_score/std": 0.21352458000183105,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9165331125259399,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 87.43069458007812,
      "step": 2621
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.24080143868923187,
      "epoch": 4.201923076923077,
      "grad_norm": 2.7573001384735107,
      "learning_rate": 1e-06,
      "loss": 0.1249,
      "step": 2622
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.21514707058668137,
      "epoch": 4.203525641025641,
      "grad_norm": 0.024495212361216545,
      "learning_rate": 1e-06,
      "loss": 0.1772,
      "step": 2623
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.20372025668621063,
      "epoch": 4.205128205128205,
      "grad_norm": 0.013811548240482807,
      "learning_rate": 1e-06,
      "loss": 0.2065,
      "step": 2624
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.28515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3565.0,
      "completions/mean_length": 6081.05078125,
      "completions/mean_terminated_length": 1971.131103515625,
      "completions/min_length": 945.0,
      "completions/min_terminated_length": 945.0,
      "entropy": 0.21161595731973648,
      "epoch": 4.206730769230769,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 856.1641845703125,
      "learning_rate": 1e-06,
      "loss": 0.154,
      "num_tokens": 1538566221.0,
      "reward": 0.3104988634586334,
      "reward_std": 0.025025444105267525,
      "rewards/progression_diversity/mean": -0.06339757144451141,
      "rewards/progression_diversity/std": 0.10305111110210419,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.7278645634651184,
      "rewards/symbolic_reward_partial_score/std": 0.1853560358285904,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9161036014556885,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 87.99066925048828,
      "step": 2625
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.23871337622404099,
      "epoch": 4.208333333333333,
      "grad_norm": 36.576744079589844,
      "learning_rate": 1e-06,
      "loss": 0.1355,
      "step": 2626
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.21872205287218094,
      "epoch": 4.209935897435898,
      "grad_norm": 27.23736000061035,
      "learning_rate": 1e-06,
      "loss": 0.1091,
      "step": 2627
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.1995127573609352,
      "epoch": 4.211538461538462,
      "grad_norm": 0.022180048748850822,
      "learning_rate": 1e-06,
      "loss": 0.2113,
      "step": 2628
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.310546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4055.0,
      "completions/mean_length": 6442.0234375,
      "completions/mean_terminated_length": 1963.9093017578125,
      "completions/min_length": 656.0,
      "completions/min_terminated_length": 656.0,
      "entropy": 0.20760582387447357,
      "epoch": 4.2131410256410255,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 3401.139404296875,
      "learning_rate": 1e-06,
      "loss": 0.1055,
      "num_tokens": 1542709657.0,
      "reward": 0.2375459372997284,
      "reward_std": 0.05149232968688011,
      "rewards/progression_diversity/mean": -0.07206659764051437,
      "rewards/progression_diversity/std": 0.11053474247455597,
      "rewards/symbolic_reward_accuracy/mean": 0.052734375,
      "rewards/symbolic_reward_accuracy/std": 0.22372129559516907,
      "rewards/symbolic_reward_partial_score/mean": 0.6946126222610474,
      "rewards/symbolic_reward_partial_score/std": 0.21639835834503174,
      "rewards/tag_count_reward/mean": -0.017578125,
      "rewards/tag_count_reward/std": 0.13154059648513794,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9067912697792053,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 92.28494262695312,
      "step": 2629
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.2265625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.1819758117198944,
      "epoch": 4.214743589743589,
      "grad_norm": 1448.0584716796875,
      "learning_rate": 1e-06,
      "loss": 0.3297,
      "step": 2630
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.22094269841909409,
      "epoch": 4.216346153846154,
      "grad_norm": 0.1524176448583603,
      "learning_rate": 1e-06,
      "loss": 0.1131,
      "step": 2631
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.21019763499498367,
      "epoch": 4.217948717948718,
      "grad_norm": 0.015752186998724937,
      "learning_rate": 1e-06,
      "loss": 0.1985,
      "step": 2632
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.330078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3971.0,
      "completions/mean_length": 6724.53125,
      "completions/mean_terminated_length": 1965.2012939453125,
      "completions/min_length": 1115.0,
      "completions/min_terminated_length": 1115.0,
      "entropy": 0.19603969156742096,
      "epoch": 4.219551282051282,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1366.21484375,
      "learning_rate": 1e-06,
      "loss": 0.1466,
      "num_tokens": 1547020169.0,
      "reward": 0.35085177421569824,
      "reward_std": 0.07054072618484497,
      "rewards/progression_diversity/mean": -0.07302698493003845,
      "rewards/progression_diversity/std": 0.10713056474924088,
      "rewards/symbolic_reward_accuracy/mean": 0.232421875,
      "rewards/symbolic_reward_accuracy/std": 0.42278963327407837,
      "rewards/symbolic_reward_partial_score/mean": 0.7110025882720947,
      "rewards/symbolic_reward_partial_score/std": 0.22401832044124603,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.910980224609375,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 90.13082122802734,
      "step": 2633
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.19943545758724213,
      "epoch": 4.221153846153846,
      "grad_norm": 25.857837677001953,
      "learning_rate": 1e-06,
      "loss": 0.1302,
      "step": 2634
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.21115753799676895,
      "epoch": 4.222756410256411,
      "grad_norm": 0.014513122849166393,
      "learning_rate": 1e-06,
      "loss": 0.1383,
      "step": 2635
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.1924782246351242,
      "epoch": 4.2243589743589745,
      "grad_norm": 0.024281147867441177,
      "learning_rate": 1e-06,
      "loss": 0.1726,
      "step": 2636
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.255859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3878.0,
      "completions/mean_length": 5678.255859375,
      "completions/mean_terminated_length": 1997.2781982421875,
      "completions/min_length": 1170.0,
      "completions/min_terminated_length": 1170.0,
      "entropy": 0.21415793150663376,
      "epoch": 4.225961538461538,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1601.75390625,
      "learning_rate": 1e-06,
      "loss": 0.0806,
      "num_tokens": 1550778716.0,
      "reward": 0.34982889890670776,
      "reward_std": 0.06202153116464615,
      "rewards/progression_diversity/mean": -0.05031656473875046,
      "rewards/progression_diversity/std": 0.0884041115641594,
      "rewards/symbolic_reward_accuracy/mean": 0.232421875,
      "rewards/symbolic_reward_accuracy/std": 0.42278963327407837,
      "rewards/symbolic_reward_partial_score/mean": 0.7055338621139526,
      "rewards/symbolic_reward_partial_score/std": 0.22613964974880219,
      "rewards/tag_count_reward/mean": -0.0078125,
      "rewards/tag_count_reward/std": 0.08812850713729858,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9257657527923584,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 81.70069885253906,
      "step": 2637
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.22891773283481598,
      "epoch": 4.227564102564102,
      "grad_norm": 2865.813232421875,
      "learning_rate": 1e-06,
      "loss": 0.0946,
      "step": 2638
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.21290214359760284,
      "epoch": 4.229166666666667,
      "grad_norm": 9.31295394897461,
      "learning_rate": 1e-06,
      "loss": 0.194,
      "step": 2639
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.2276032641530037,
      "epoch": 4.230769230769231,
      "grad_norm": 863.3584594726562,
      "learning_rate": 1e-06,
      "loss": 0.2034,
      "step": 2640
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4074.0,
      "completions/mean_length": 6243.08203125,
      "completions/mean_terminated_length": 1961.3612060546875,
      "completions/min_length": 1208.0,
      "completions/min_terminated_length": 1208.0,
      "entropy": 0.18027716875076294,
      "epoch": 4.232371794871795,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1586.920654296875,
      "learning_rate": 1e-06,
      "loss": 0.1933,
      "num_tokens": 1554829782.0,
      "reward": 0.5201883912086487,
      "reward_std": 0.06066072732210159,
      "rewards/progression_diversity/mean": -0.05831076577305794,
      "rewards/progression_diversity/std": 0.09286186844110489,
      "rewards/symbolic_reward_accuracy/mean": 0.453125,
      "rewards/symbolic_reward_accuracy/std": 0.4982847273349762,
      "rewards/symbolic_reward_partial_score/mean": 0.8316080570220947,
      "rewards/symbolic_reward_partial_score/std": 0.2016787976026535,
      "rewards/tag_count_reward/mean": -0.005859375,
      "rewards/tag_count_reward/std": 0.07639661431312561,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9149261116981506,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 86.6026611328125,
      "step": 2641
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.2154352068901062,
      "epoch": 4.233974358974359,
      "grad_norm": 0.13046927750110626,
      "learning_rate": 1e-06,
      "loss": 0.114,
      "step": 2642
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.20094606280326843,
      "epoch": 4.235576923076923,
      "grad_norm": 0.010223207995295525,
      "learning_rate": 1e-06,
      "loss": 0.1366,
      "step": 2643
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.20492058992385864,
      "epoch": 4.237179487179487,
      "grad_norm": 0.01739051379263401,
      "learning_rate": 1e-06,
      "loss": 0.156,
      "step": 2644
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3361.0,
      "completions/mean_length": 5930.7109375,
      "completions/mean_terminated_length": 1996.6773681640625,
      "completions/min_length": 755.0,
      "completions/min_terminated_length": 755.0,
      "entropy": 0.21672135591506958,
      "epoch": 4.238782051282051,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 1050.203125,
      "learning_rate": 1e-06,
      "loss": 0.1747,
      "num_tokens": 1558679986.0,
      "reward": 0.4322790503501892,
      "reward_std": 0.07013079524040222,
      "rewards/progression_diversity/mean": -0.05236773565411568,
      "rewards/progression_diversity/std": 0.08915333449840546,
      "rewards/symbolic_reward_accuracy/mean": 0.35546875,
      "rewards/symbolic_reward_accuracy/std": 0.47912323474884033,
      "rewards/symbolic_reward_partial_score/mean": 0.7395508289337158,
      "rewards/symbolic_reward_partial_score/std": 0.2565183639526367,
      "rewards/tag_count_reward/mean": -0.0234375,
      "rewards/tag_count_reward/std": 0.15143637359142303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9261719584465027,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 78.85722351074219,
      "step": 2645
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.2054838389158249,
      "epoch": 4.240384615384615,
      "grad_norm": 109.77615356445312,
      "learning_rate": 1e-06,
      "loss": 0.1925,
      "step": 2646
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.24229851365089417,
      "epoch": 4.24198717948718,
      "grad_norm": 0.40434131026268005,
      "learning_rate": 1e-06,
      "loss": 0.1406,
      "step": 2647
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.21324530988931656,
      "epoch": 4.243589743589744,
      "grad_norm": 1434.3251953125,
      "learning_rate": 1e-06,
      "loss": 0.2993,
      "step": 2648
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.287109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3740.0,
      "completions/mean_length": 6136.265625,
      "completions/mean_terminated_length": 2009.095947265625,
      "completions/min_length": 982.0,
      "completions/min_terminated_length": 982.0,
      "entropy": 0.20269504189491272,
      "epoch": 4.2451923076923075,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 778.6556396484375,
      "learning_rate": 1e-06,
      "loss": 0.1087,
      "num_tokens": 1562723466.0,
      "reward": 0.3345257341861725,
      "reward_std": 0.0690338984131813,
      "rewards/progression_diversity/mean": -0.054263561964035034,
      "rewards/progression_diversity/std": 0.08993817865848541,
      "rewards/symbolic_reward_accuracy/mean": 0.21875,
      "rewards/symbolic_reward_accuracy/std": 0.41380295157432556,
      "rewards/symbolic_reward_partial_score/mean": 0.6839518547058105,
      "rewards/symbolic_reward_partial_score/std": 0.21551425755023956,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9316227436065674,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 74.91038513183594,
      "step": 2649
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.359375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.197452612221241,
      "epoch": 4.246794871794872,
      "grad_norm": 7070.578125,
      "learning_rate": 1e-06,
      "loss": 0.9423,
      "step": 2650
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.21947723627090454,
      "epoch": 4.248397435897436,
      "grad_norm": 1.046020746231079,
      "learning_rate": 1e-06,
      "loss": 0.2003,
      "step": 2651
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.21606357395648956,
      "epoch": 4.25,
      "grad_norm": 0.011754968203604221,
      "learning_rate": 1e-06,
      "loss": 0.1609,
      "step": 2652
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.23046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4284.0,
      "completions/mean_length": 5343.4765625,
      "completions/mean_terminated_length": 2036.923828125,
      "completions/min_length": 958.0,
      "completions/min_terminated_length": 958.0,
      "entropy": 0.2657729983329773,
      "epoch": 4.251602564102564,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1515.29736328125,
      "learning_rate": 1e-06,
      "loss": 0.0791,
      "num_tokens": 1566282654.0,
      "reward": 0.28956276178359985,
      "reward_std": 0.050334710627794266,
      "rewards/progression_diversity/mean": -0.04128318279981613,
      "rewards/progression_diversity/std": 0.08098018169403076,
      "rewards/symbolic_reward_accuracy/mean": 0.12109375,
      "rewards/symbolic_reward_accuracy/std": 0.3265552520751953,
      "rewards/symbolic_reward_partial_score/mean": 0.7276529669761658,
      "rewards/symbolic_reward_partial_score/std": 0.1922309845685959,
      "rewards/tag_count_reward/mean": -0.009765625,
      "rewards/tag_count_reward/std": 0.09843364357948303,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9493874907493591,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 63.77783203125,
      "step": 2653
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2197081595659256,
      "epoch": 4.253205128205128,
      "grad_norm": 0.02242899127304554,
      "learning_rate": 1e-06,
      "loss": 0.192,
      "step": 2654
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.220329187810421,
      "epoch": 4.2548076923076925,
      "grad_norm": 0.013821829110383987,
      "learning_rate": 1e-06,
      "loss": 0.1641,
      "step": 2655
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4921875,
      "entropy": 0.21346864849328995,
      "epoch": 4.256410256410256,
      "grad_norm": 0.020072348415851593,
      "learning_rate": 1e-06,
      "loss": 0.2551,
      "step": 2656
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.279296875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4037.0,
      "completions/mean_length": 6021.244140625,
      "completions/mean_terminated_length": 2005.3251953125,
      "completions/min_length": 878.0,
      "completions/min_terminated_length": 878.0,
      "entropy": 0.2017608880996704,
      "epoch": 4.25801282051282,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1055.4764404296875,
      "learning_rate": 1e-06,
      "loss": 0.1218,
      "num_tokens": 1570270299.0,
      "reward": 0.3231794834136963,
      "reward_std": 0.05370699614286423,
      "rewards/progression_diversity/mean": -0.05509728938341141,
      "rewards/progression_diversity/std": 0.09356637299060822,
      "rewards/symbolic_reward_accuracy/mean": 0.189453125,
      "rewards/symbolic_reward_accuracy/std": 0.3922513723373413,
      "rewards/symbolic_reward_partial_score/mean": 0.7047525644302368,
      "rewards/symbolic_reward_partial_score/std": 0.21620500087738037,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.935107946395874,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 70.60637664794922,
      "step": 2657
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.23436833918094635,
      "epoch": 4.259615384615385,
      "grad_norm": 139.91903686523438,
      "learning_rate": 1e-06,
      "loss": 0.1392,
      "step": 2658
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.1886296272277832,
      "epoch": 4.261217948717949,
      "grad_norm": 23.106496810913086,
      "learning_rate": 1e-06,
      "loss": 0.2419,
      "step": 2659
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2260531634092331,
      "epoch": 4.262820512820513,
      "grad_norm": 0.19578726589679718,
      "learning_rate": 1e-06,
      "loss": 0.1691,
      "step": 2660
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.287109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3542.0,
      "completions/mean_length": 6165.703125,
      "completions/mean_terminated_length": 2050.38916015625,
      "completions/min_length": 1040.0,
      "completions/min_terminated_length": 1040.0,
      "entropy": 0.19962295144796371,
      "epoch": 4.264423076923077,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 7102.345703125,
      "learning_rate": 1e-06,
      "loss": 0.1947,
      "num_tokens": 1574379123.0,
      "reward": 0.31796449422836304,
      "reward_std": 0.04501751810312271,
      "rewards/progression_diversity/mean": -0.06097230315208435,
      "rewards/progression_diversity/std": 0.10001429170370102,
      "rewards/symbolic_reward_accuracy/mean": 0.173828125,
      "rewards/symbolic_reward_accuracy/std": 0.3793322443962097,
      "rewards/symbolic_reward_partial_score/mean": 0.7155599594116211,
      "rewards/symbolic_reward_partial_score/std": 0.19974127411842346,
      "rewards/tag_count_reward/mean": -0.00390625,
      "rewards/tag_count_reward/std": 0.06243881583213806,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9347620606422424,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 71.3311767578125,
      "step": 2661
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.20831771194934845,
      "epoch": 4.266025641025641,
      "grad_norm": 1.6806011199951172,
      "learning_rate": 1e-06,
      "loss": 0.2179,
      "step": 2662
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.2336948812007904,
      "epoch": 4.267628205128205,
      "grad_norm": 0.07298646867275238,
      "learning_rate": 1e-06,
      "loss": 0.1278,
      "step": 2663
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5234375,
      "entropy": 0.2085075080394745,
      "epoch": 4.269230769230769,
      "grad_norm": 0.7998039722442627,
      "learning_rate": 1e-06,
      "loss": 0.1634,
      "step": 2664
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.259765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3341.0,
      "completions/mean_length": 5808.7734375,
      "completions/mean_terminated_length": 2097.67822265625,
      "completions/min_length": 1006.0,
      "completions/min_terminated_length": 1006.0,
      "entropy": 0.21773767471313477,
      "epoch": 4.270833333333333,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 1245.658203125,
      "learning_rate": 1e-06,
      "loss": 0.1845,
      "num_tokens": 1578336351.0,
      "reward": 0.32293805480003357,
      "reward_std": 0.05607787147164345,
      "rewards/progression_diversity/mean": -0.048967085778713226,
      "rewards/progression_diversity/std": 0.0883127897977829,
      "rewards/symbolic_reward_accuracy/mean": 0.173828125,
      "rewards/symbolic_reward_accuracy/std": 0.3793322443962097,
      "rewards/symbolic_reward_partial_score/mean": 0.7343424558639526,
      "rewards/symbolic_reward_partial_score/std": 0.213489830493927,
      "rewards/tag_count_reward/mean": -0.01171875,
      "rewards/tag_count_reward/std": 0.10772226005792618,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9438670873641968,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 65.19830322265625,
      "step": 2665
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.20395781844854355,
      "epoch": 4.272435897435898,
      "grad_norm": 1100.1546630859375,
      "learning_rate": 1e-06,
      "loss": 0.2518,
      "step": 2666
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.23760968446731567,
      "epoch": 4.274038461538462,
      "grad_norm": 0.022954348474740982,
      "learning_rate": 1e-06,
      "loss": 0.0944,
      "step": 2667
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.22392556816339493,
      "epoch": 4.2756410256410255,
      "grad_norm": 0.028759047389030457,
      "learning_rate": 1e-06,
      "loss": 0.1897,
      "step": 2668
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.271484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4153.0,
      "completions/mean_length": 5969.759765625,
      "completions/mean_terminated_length": 2088.849853515625,
      "completions/min_length": 986.0,
      "completions/min_terminated_length": 986.0,
      "entropy": 0.22678442299365997,
      "epoch": 4.277243589743589,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 729.8278198242188,
      "learning_rate": 1e-06,
      "loss": 0.0574,
      "num_tokens": 1582333732.0,
      "reward": 0.2548202872276306,
      "reward_std": 0.04770761728286743,
      "rewards/progression_diversity/mean": -0.0428718663752079,
      "rewards/progression_diversity/std": 0.07706419378519058,
      "rewards/symbolic_reward_accuracy/mean": 0.107421875,
      "rewards/symbolic_reward_accuracy/std": 0.30995169281959534,
      "rewards/symbolic_reward_partial_score/mean": 0.6405435800552368,
      "rewards/symbolic_reward_partial_score/std": 0.21080487966537476,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9503331780433655,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 61.90028381347656,
      "step": 2669
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2734375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.20155585557222366,
      "epoch": 4.278846153846154,
      "grad_norm": 0.12458622455596924,
      "learning_rate": 1e-06,
      "loss": 0.253,
      "step": 2670
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2578125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.21851865947246552,
      "epoch": 4.280448717948718,
      "grad_norm": 0.17406442761421204,
      "learning_rate": 1e-06,
      "loss": 0.1624,
      "step": 2671
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.2346644550561905,
      "epoch": 4.282051282051282,
      "grad_norm": 0.018945371732115746,
      "learning_rate": 1e-06,
      "loss": 0.1564,
      "step": 2672
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.23828125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3899.0,
      "completions/mean_length": 5541.857421875,
      "completions/mean_terminated_length": 2150.212890625,
      "completions/min_length": 1054.0,
      "completions/min_terminated_length": 1054.0,
      "entropy": 0.2370782345533371,
      "epoch": 4.283653846153846,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 549.9456787109375,
      "learning_rate": 1e-06,
      "loss": 0.0235,
      "num_tokens": 1585986235.0,
      "reward": 0.34320884943008423,
      "reward_std": 0.08159251511096954,
      "rewards/progression_diversity/mean": -0.03165600448846817,
      "rewards/progression_diversity/std": 0.06647796928882599,
      "rewards/symbolic_reward_accuracy/mean": 0.220703125,
      "rewards/symbolic_reward_accuracy/std": 0.4151262938976288,
      "rewards/symbolic_reward_partial_score/mean": 0.7088867425918579,
      "rewards/symbolic_reward_partial_score/std": 0.2290610820055008,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9550345540046692,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 60.594451904296875,
      "step": 2673
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.28125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.5,
      "entropy": 0.21898599714040756,
      "epoch": 4.285256410256411,
      "grad_norm": 0.42992645502090454,
      "learning_rate": 1e-06,
      "loss": 0.2189,
      "step": 2674
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.24925994873046875,
      "epoch": 4.2868589743589745,
      "grad_norm": 0.020100675523281097,
      "learning_rate": 1e-06,
      "loss": 0.0946,
      "step": 2675
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2421875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.2393641099333763,
      "epoch": 4.288461538461538,
      "grad_norm": 0.01565798930823803,
      "learning_rate": 1e-06,
      "loss": 0.2389,
      "step": 2676
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.21484375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4642.0,
      "completions/mean_length": 5307.583984375,
      "completions/mean_terminated_length": 2276.723876953125,
      "completions/min_length": 855.0,
      "completions/min_terminated_length": 855.0,
      "entropy": 0.2343580722808838,
      "epoch": 4.290064102564102,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1280.0311279296875,
      "learning_rate": 1e-06,
      "loss": 0.0584,
      "num_tokens": 1589516614.0,
      "reward": 0.3476376533508301,
      "reward_std": 0.125152587890625,
      "rewards/progression_diversity/mean": -0.03457668796181679,
      "rewards/progression_diversity/std": 0.08347929269075394,
      "rewards/symbolic_reward_accuracy/mean": 0.236328125,
      "rewards/symbolic_reward_accuracy/std": 0.42524150013923645,
      "rewards/symbolic_reward_partial_score/mean": 0.7035644054412842,
      "rewards/symbolic_reward_partial_score/std": 0.2703304886817932,
      "rewards/tag_count_reward/mean": -0.048828125,
      "rewards/tag_count_reward/std": 0.2157193273305893,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9533030986785889,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 62.77830505371094,
      "step": 2677
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.26719367504119873,
      "epoch": 4.291666666666667,
      "grad_norm": 0.26546433568000793,
      "learning_rate": 1e-06,
      "loss": 0.1209,
      "step": 2678
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.2331203892827034,
      "epoch": 4.293269230769231,
      "grad_norm": 0.01867266371846199,
      "learning_rate": 1e-06,
      "loss": 0.2064,
      "step": 2679
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.23465164005756378,
      "epoch": 4.294871794871795,
      "grad_norm": 0.015481102280318737,
      "learning_rate": 1e-06,
      "loss": 0.1809,
      "step": 2680
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.2734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4966.0,
      "completions/mean_length": 6191.328125,
      "completions/mean_terminated_length": 2355.376220703125,
      "completions/min_length": 918.0,
      "completions/min_terminated_length": 918.0,
      "entropy": 0.18589715659618378,
      "epoch": 4.296474358974359,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 1452.932373046875,
      "learning_rate": 1e-06,
      "loss": 0.1798,
      "num_tokens": 1593491662.0,
      "reward": 0.35412442684173584,
      "reward_std": 0.1382516324520111,
      "rewards/progression_diversity/mean": -0.046052753925323486,
      "rewards/progression_diversity/std": 0.09779871255159378,
      "rewards/symbolic_reward_accuracy/mean": 0.2578125,
      "rewards/symbolic_reward_accuracy/std": 0.43785804510116577,
      "rewards/symbolic_reward_partial_score/mean": 0.6865071654319763,
      "rewards/symbolic_reward_partial_score/std": 0.29961350560188293,
      "rewards/tag_count_reward/mean": -0.060546875,
      "rewards/tag_count_reward/std": 0.2387305200099945,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.935286283493042,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 69.72281646728516,
      "step": 2681
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.21021146327257156,
      "epoch": 4.298076923076923,
      "grad_norm": 0.04195324704051018,
      "learning_rate": 1e-06,
      "loss": 0.1036,
      "step": 2682
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.20793236792087555,
      "epoch": 4.299679487179487,
      "grad_norm": 0.15370796620845795,
      "learning_rate": 1e-06,
      "loss": 0.1581,
      "step": 2683
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.19368456304073334,
      "epoch": 4.301282051282051,
      "grad_norm": 0.04016382619738579,
      "learning_rate": 1e-06,
      "loss": 0.2206,
      "step": 2684
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4625.0,
      "completions/mean_length": 5578.798828125,
      "completions/mean_terminated_length": 2271.084228515625,
      "completions/min_length": 590.0,
      "completions/min_terminated_length": 590.0,
      "entropy": 0.21986397355794907,
      "epoch": 4.302884615384615,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 847.1763305664062,
      "learning_rate": 1e-06,
      "loss": 0.1179,
      "num_tokens": 1597246327.0,
      "reward": 0.2654598355293274,
      "reward_std": 0.09309060871601105,
      "rewards/progression_diversity/mean": -0.039955541491508484,
      "rewards/progression_diversity/std": 0.08839154243469238,
      "rewards/symbolic_reward_accuracy/mean": 0.134765625,
      "rewards/symbolic_reward_accuracy/std": 0.3418070077896118,
      "rewards/symbolic_reward_partial_score/mean": 0.6368489265441895,
      "rewards/symbolic_reward_partial_score/std": 0.2666816711425781,
      "rewards/tag_count_reward/mean": -0.060546875,
      "rewards/tag_count_reward/std": 0.2387305200099945,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.944598376750946,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 68.18759155273438,
      "step": 2685
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.20748194307088852,
      "epoch": 4.30448717948718,
      "grad_norm": 1317.310302734375,
      "learning_rate": 1e-06,
      "loss": 0.4452,
      "step": 2686
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23975148051977158,
      "epoch": 4.306089743589744,
      "grad_norm": 0.02449853904545307,
      "learning_rate": 1e-06,
      "loss": 0.0999,
      "step": 2687
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.21875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.2513374388217926,
      "epoch": 4.3076923076923075,
      "grad_norm": 0.0537121407687664,
      "learning_rate": 1e-06,
      "loss": 0.1342,
      "step": 2688
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.173828125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4082.0,
      "completions/mean_length": 4746.345703125,
      "completions/mean_terminated_length": 2297.76123046875,
      "completions/min_length": 1121.0,
      "completions/min_terminated_length": 1121.0,
      "entropy": 0.264044925570488,
      "epoch": 4.309294871794872,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 532.7546997070312,
      "learning_rate": 1e-06,
      "loss": 0.1721,
      "num_tokens": 1600521912.0,
      "reward": 0.3210293650627136,
      "reward_std": 0.10824176669120789,
      "rewards/progression_diversity/mean": -0.02890094369649887,
      "rewards/progression_diversity/std": 0.08233704417943954,
      "rewards/symbolic_reward_accuracy/mean": 0.19140625,
      "rewards/symbolic_reward_accuracy/std": 0.3937928080558777,
      "rewards/symbolic_reward_partial_score/mean": 0.7064778804779053,
      "rewards/symbolic_reward_partial_score/std": 0.26733410358428955,
      "rewards/tag_count_reward/mean": -0.0546875,
      "rewards/tag_count_reward/std": 0.2275916188955307,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.965470552444458,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 58.45082473754883,
      "step": 2689
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.390625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.27652691304683685,
      "epoch": 4.310897435897436,
      "grad_norm": 2549.2978515625,
      "learning_rate": 1e-06,
      "loss": 0.0809,
      "step": 2690
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.23626501858234406,
      "epoch": 4.3125,
      "grad_norm": 25.48175048828125,
      "learning_rate": 1e-06,
      "loss": 0.1749,
      "step": 2691
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.26965607702732086,
      "epoch": 4.314102564102564,
      "grad_norm": 21.068880081176758,
      "learning_rate": 1e-06,
      "loss": 0.158,
      "step": 2692
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.173828125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4066.0,
      "completions/mean_length": 4708.7734375,
      "completions/mean_terminated_length": 2252.28369140625,
      "completions/min_length": 894.0,
      "completions/min_terminated_length": 894.0,
      "entropy": 0.2753334045410156,
      "epoch": 4.315705128205128,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 2231.534423828125,
      "learning_rate": 1e-06,
      "loss": 0.1243,
      "num_tokens": 1603895476.0,
      "reward": 0.3245422840118408,
      "reward_std": 0.10428034514188766,
      "rewards/progression_diversity/mean": -0.028682753443717957,
      "rewards/progression_diversity/std": 0.07225596159696579,
      "rewards/symbolic_reward_accuracy/mean": 0.21875,
      "rewards/symbolic_reward_accuracy/std": 0.41380295157432556,
      "rewards/symbolic_reward_partial_score/mean": 0.662841796875,
      "rewards/symbolic_reward_partial_score/std": 0.27776244282722473,
      "rewards/tag_count_reward/mean": -0.052734375,
      "rewards/tag_count_reward/std": 0.22372129559516907,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9683549404144287,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 57.83184051513672,
      "step": 2693
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.2380198836326599,
      "epoch": 4.3173076923076925,
      "grad_norm": 908.7568969726562,
      "learning_rate": 1e-06,
      "loss": 0.1637,
      "step": 2694
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2474098801612854,
      "epoch": 4.318910256410256,
      "grad_norm": 24.16575050354004,
      "learning_rate": 1e-06,
      "loss": 0.213,
      "step": 2695
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.28730684518814087,
      "epoch": 4.32051282051282,
      "grad_norm": 0.17024962604045868,
      "learning_rate": 1e-06,
      "loss": 0.027,
      "step": 2696
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.17578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5872.0,
      "completions/mean_length": 4746.83203125,
      "completions/mean_terminated_length": 2264.9716796875,
      "completions/min_length": 1135.0,
      "completions/min_terminated_length": 1135.0,
      "entropy": 0.25719259679317474,
      "epoch": 4.322115384615385,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 801.6245727539062,
      "learning_rate": 1e-06,
      "loss": 0.1019,
      "num_tokens": 1607156350.0,
      "reward": 0.30364054441452026,
      "reward_std": 0.11523503810167313,
      "rewards/progression_diversity/mean": -0.023154182359576225,
      "rewards/progression_diversity/std": 0.07085532695055008,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.668505847454071,
      "rewards/symbolic_reward_partial_score/std": 0.2690269947052002,
      "rewards/tag_count_reward/mean": -0.044921875,
      "rewards/tag_count_reward/std": 0.20733514428138733,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9642022848129272,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 60.5496826171875,
      "step": 2697
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.264162078499794,
      "epoch": 4.323717948717949,
      "grad_norm": 7159.71484375,
      "learning_rate": 1e-06,
      "loss": 0.5042,
      "step": 2698
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3515625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.2615188807249069,
      "epoch": 4.325320512820513,
      "grad_norm": 6776.10302734375,
      "learning_rate": 1e-06,
      "loss": 0.2205,
      "step": 2699
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3515625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2697322964668274,
      "epoch": 4.326923076923077,
      "grad_norm": 0.10481588542461395,
      "learning_rate": 1e-06,
      "loss": 0.1195,
      "step": 2700
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.185546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3512.0,
      "completions/mean_length": 4892.396484375,
      "completions/mean_terminated_length": 2274.4052734375,
      "completions/min_length": 1158.0,
      "completions/min_terminated_length": 1158.0,
      "entropy": 0.27855899930000305,
      "epoch": 4.328525641025641,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 877.5675659179688,
      "learning_rate": 1e-06,
      "loss": 0.0915,
      "num_tokens": 1610557369.0,
      "reward": 0.24516724050045013,
      "reward_std": 0.09491077810525894,
      "rewards/progression_diversity/mean": -0.03063952922821045,
      "rewards/progression_diversity/std": 0.08986818790435791,
      "rewards/symbolic_reward_accuracy/mean": 0.109375,
      "rewards/symbolic_reward_accuracy/std": 0.31241437792778015,
      "rewards/symbolic_reward_partial_score/mean": 0.6248860955238342,
      "rewards/symbolic_reward_partial_score/std": 0.26446953415870667,
      "rewards/tag_count_reward/mean": -0.076171875,
      "rewards/tag_count_reward/std": 0.26553234457969666,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9703221321105957,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 56.01524353027344,
      "step": 2701
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2597070336341858,
      "epoch": 4.330128205128205,
      "grad_norm": 2198.372802734375,
      "learning_rate": 1e-06,
      "loss": 0.3091,
      "step": 2702
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.34375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.26145724952220917,
      "epoch": 4.331730769230769,
      "grad_norm": 1281.1771240234375,
      "learning_rate": 1e-06,
      "loss": 0.1863,
      "step": 2703
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2582560032606125,
      "epoch": 4.333333333333333,
      "grad_norm": 0.06832489371299744,
      "learning_rate": 1e-06,
      "loss": 0.2302,
      "step": 2704
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.13671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4234.0,
      "completions/mean_length": 4208.84765625,
      "completions/mean_terminated_length": 2280.65625,
      "completions/min_length": 1218.0,
      "completions/min_terminated_length": 1218.0,
      "entropy": 0.2881985604763031,
      "epoch": 4.334935897435898,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 2649.09423828125,
      "learning_rate": 1e-06,
      "loss": 0.2109,
      "num_tokens": 1613462507.0,
      "reward": 0.3500940799713135,
      "reward_std": 0.11316157132387161,
      "rewards/progression_diversity/mean": -0.019399691373109818,
      "rewards/progression_diversity/std": 0.07821278274059296,
      "rewards/symbolic_reward_accuracy/mean": 0.24609375,
      "rewards/symbolic_reward_accuracy/std": 0.4311550557613373,
      "rewards/symbolic_reward_partial_score/mean": 0.6936686038970947,
      "rewards/symbolic_reward_partial_score/std": 0.2801879942417145,
      "rewards/tag_count_reward/mean": -0.0546875,
      "rewards/tag_count_reward/std": 0.2275916188955307,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9930620193481445,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 43.674652099609375,
      "step": 2705
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.28035198152065277,
      "epoch": 4.336538461538462,
      "grad_norm": 6.077508449554443,
      "learning_rate": 1e-06,
      "loss": 0.1302,
      "step": 2706
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.29166266322135925,
      "epoch": 4.3381410256410255,
      "grad_norm": 0.6038011908531189,
      "learning_rate": 1e-06,
      "loss": 0.1481,
      "step": 2707
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.28930486738681793,
      "epoch": 4.339743589743589,
      "grad_norm": 0.5783957839012146,
      "learning_rate": 1e-06,
      "loss": 0.1213,
      "step": 2708
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.15625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4111.0,
      "completions/mean_length": 4497.861328125,
      "completions/mean_terminated_length": 2296.724609375,
      "completions/min_length": 1329.0,
      "completions/min_terminated_length": 1329.0,
      "entropy": 0.2855447679758072,
      "epoch": 4.341346153846154,
      "frac_reward_zero_std": 0.0,
      "grad_norm": 770.3568725585938,
      "learning_rate": 1e-06,
      "loss": 0.1363,
      "num_tokens": 1616717908.0,
      "reward": 0.2790448069572449,
      "reward_std": 0.11913672089576721,
      "rewards/progression_diversity/mean": -0.018859868869185448,
      "rewards/progression_diversity/std": 0.06757510453462601,
      "rewards/symbolic_reward_accuracy/mean": 0.134765625,
      "rewards/symbolic_reward_accuracy/std": 0.3418070077896118,
      "rewards/symbolic_reward_partial_score/mean": 0.6872884035110474,
      "rewards/symbolic_reward_partial_score/std": 0.2816750705242157,
      "rewards/tag_count_reward/mean": -0.078125,
      "rewards/tag_count_reward/std": 0.26863065361976624,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9902298450469971,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 45.65616989135742,
      "step": 2709
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.26931584626436234,
      "epoch": 4.342948717948718,
      "grad_norm": 456.98199462890625,
      "learning_rate": 1e-06,
      "loss": 0.2891,
      "step": 2710
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2959611713886261,
      "epoch": 4.344551282051282,
      "grad_norm": 6.0887250900268555,
      "learning_rate": 1e-06,
      "loss": 0.7324,
      "step": 2711
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2712084949016571,
      "epoch": 4.346153846153846,
      "grad_norm": 61.22941970825195,
      "learning_rate": 1e-06,
      "loss": 0.2727,
      "step": 2712
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4333.0,
      "completions/mean_length": 4048.97265625,
      "completions/mean_terminated_length": 2286.825927734375,
      "completions/min_length": 857.0,
      "completions/min_terminated_length": 857.0,
      "entropy": 0.2950481027364731,
      "epoch": 4.347756410256411,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 1410.4874267578125,
      "learning_rate": 1e-06,
      "loss": 0.1737,
      "num_tokens": 1619610758.0,
      "reward": 0.28769299387931824,
      "reward_std": 0.09948226064443588,
      "rewards/progression_diversity/mean": -0.012928030453622341,
      "rewards/progression_diversity/std": 0.05409675091505051,
      "rewards/symbolic_reward_accuracy/mean": 0.1640625,
      "rewards/symbolic_reward_accuracy/std": 0.37069445848464966,
      "rewards/symbolic_reward_partial_score/mean": 0.6508138179779053,
      "rewards/symbolic_reward_partial_score/std": 0.267758846282959,
      "rewards/tag_count_reward/mean": -0.05859375,
      "rewards/tag_count_reward/std": 0.23509246110916138,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0013058185577393,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 40.96929931640625,
      "step": 2713
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3671875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.3074454814195633,
      "epoch": 4.3493589743589745,
      "grad_norm": 1392.9722900390625,
      "learning_rate": 1e-06,
      "loss": 0.1584,
      "step": 2714
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.28932228684425354,
      "epoch": 4.350961538461538,
      "grad_norm": 12.400731086730957,
      "learning_rate": 1e-06,
      "loss": 0.5763,
      "step": 2715
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.28287215530872345,
      "epoch": 4.352564102564102,
      "grad_norm": 488.6679382324219,
      "learning_rate": 1e-06,
      "loss": 0.2059,
      "step": 2716
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.142578125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4276.0,
      "completions/mean_length": 4305.140625,
      "completions/mean_terminated_length": 2296.583251953125,
      "completions/min_length": 714.0,
      "completions/min_terminated_length": 714.0,
      "entropy": 0.2647574245929718,
      "epoch": 4.354166666666667,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 1532.6412353515625,
      "learning_rate": 1e-06,
      "loss": 0.247,
      "num_tokens": 1622678990.0,
      "reward": 0.26650041341781616,
      "reward_std": 0.10990884155035019,
      "rewards/progression_diversity/mean": -0.019881153479218483,
      "rewards/progression_diversity/std": 0.07622101157903671,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.6494140625,
      "rewards/symbolic_reward_partial_score/std": 0.2761240601539612,
      "rewards/tag_count_reward/mean": -0.078125,
      "rewards/tag_count_reward/std": 0.26863065361976624,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9932198524475098,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 43.36308288574219,
      "step": 2717
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3515625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.30068182945251465,
      "epoch": 4.355769230769231,
      "grad_norm": 9.29902172088623,
      "learning_rate": 1e-06,
      "loss": 0.137,
      "step": 2718
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2581791281700134,
      "epoch": 4.357371794871795,
      "grad_norm": 1740.7247314453125,
      "learning_rate": 1e-06,
      "loss": 0.235,
      "step": 2719
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.29872237145900726,
      "epoch": 4.358974358974359,
      "grad_norm": 24.941686630249023,
      "learning_rate": 1e-06,
      "loss": 0.1233,
      "step": 2720
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.138671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5111.0,
      "completions/mean_length": 4246.28515625,
      "completions/mean_terminated_length": 2292.140625,
      "completions/min_length": 1088.0,
      "completions/min_terminated_length": 1088.0,
      "entropy": 0.3010956943035126,
      "epoch": 4.360576923076923,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 2546.631103515625,
      "learning_rate": 1e-06,
      "loss": 0.0579,
      "num_tokens": 1625667760.0,
      "reward": 0.2679121792316437,
      "reward_std": 0.097173772752285,
      "rewards/progression_diversity/mean": -0.019818993285298347,
      "rewards/progression_diversity/std": 0.07718969136476517,
      "rewards/symbolic_reward_accuracy/mean": 0.140625,
      "rewards/symbolic_reward_accuracy/std": 0.3479743003845215,
      "rewards/symbolic_reward_partial_score/mean": 0.6358886957168579,
      "rewards/symbolic_reward_partial_score/std": 0.26983723044395447,
      "rewards/tag_count_reward/mean": -0.0703125,
      "rewards/tag_count_reward/std": 0.25592297315597534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0032100677490234,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 38.43324279785156,
      "step": 2721
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.2788269519805908,
      "epoch": 4.362179487179487,
      "grad_norm": 0.07314433157444,
      "learning_rate": 1e-06,
      "loss": 0.1801,
      "step": 2722
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2793916165828705,
      "epoch": 4.363782051282051,
      "grad_norm": 0.012875649146735668,
      "learning_rate": 1e-06,
      "loss": 0.179,
      "step": 2723
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.2723848670721054,
      "epoch": 4.365384615384615,
      "grad_norm": 0.053210750222206116,
      "learning_rate": 1e-06,
      "loss": 0.2149,
      "step": 2724
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.177734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5363.0,
      "completions/mean_length": 4823.73046875,
      "completions/mean_terminated_length": 2324.955078125,
      "completions/min_length": 895.0,
      "completions/min_terminated_length": 895.0,
      "entropy": 0.26607632637023926,
      "epoch": 4.36698717948718,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 536.0072631835938,
      "learning_rate": 1e-06,
      "loss": 0.1797,
      "num_tokens": 1629215734.0,
      "reward": 0.19067470729351044,
      "reward_std": 0.1059282124042511,
      "rewards/progression_diversity/mean": -0.025302495807409286,
      "rewards/progression_diversity/std": 0.08640986680984497,
      "rewards/symbolic_reward_accuracy/mean": 0.060546875,
      "rewards/symbolic_reward_accuracy/std": 0.2387305200099945,
      "rewards/symbolic_reward_partial_score/mean": 0.5459309816360474,
      "rewards/symbolic_reward_partial_score/std": 0.26790010929107666,
      "rewards/tag_count_reward/mean": -0.091796875,
      "rewards/tag_count_reward/std": 0.289021372795105,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9844135046005249,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 45.332759857177734,
      "step": 2725
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2631705701351166,
      "epoch": 4.368589743589744,
      "grad_norm": 1414.080810546875,
      "learning_rate": 1e-06,
      "loss": 0.3588,
      "step": 2726
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.23933739215135574,
      "epoch": 4.3701923076923075,
      "grad_norm": 841.8580932617188,
      "learning_rate": 1e-06,
      "loss": 0.3975,
      "step": 2727
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.2838355749845505,
      "epoch": 4.371794871794872,
      "grad_norm": 46.018558502197266,
      "learning_rate": 1e-06,
      "loss": 0.133,
      "step": 2728
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.13671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5066.0,
      "completions/mean_length": 4252.61328125,
      "completions/mean_terminated_length": 2331.35302734375,
      "completions/min_length": 1241.0,
      "completions/min_terminated_length": 1241.0,
      "entropy": 0.2777034640312195,
      "epoch": 4.373397435897436,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 674.3890991210938,
      "learning_rate": 1e-06,
      "loss": 0.1737,
      "num_tokens": 1632154656.0,
      "reward": 0.3058478832244873,
      "reward_std": 0.1194763332605362,
      "rewards/progression_diversity/mean": -0.017753848806023598,
      "rewards/progression_diversity/std": 0.07137567549943924,
      "rewards/symbolic_reward_accuracy/mean": 0.197265625,
      "rewards/symbolic_reward_accuracy/std": 0.3983237147331238,
      "rewards/symbolic_reward_partial_score/mean": 0.6541991829872131,
      "rewards/symbolic_reward_partial_score/std": 0.2857528626918793,
      "rewards/tag_count_reward/mean": -0.0859375,
      "rewards/tag_count_reward/std": 0.28054583072662354,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.001778483390808,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 38.360496520996094,
      "step": 2729
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.328125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.3052513748407364,
      "epoch": 4.375,
      "grad_norm": 153.75807189941406,
      "learning_rate": 1e-06,
      "loss": 0.0824,
      "step": 2730
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2919474095106125,
      "epoch": 4.376602564102564,
      "grad_norm": 0.023059062659740448,
      "learning_rate": 1e-06,
      "loss": 0.2305,
      "step": 2731
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.29401521384716034,
      "epoch": 4.378205128205128,
      "grad_norm": 0.014374534599483013,
      "learning_rate": 1e-06,
      "loss": 0.2279,
      "step": 2732
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.13671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4636.0,
      "completions/mean_length": 4243.36328125,
      "completions/mean_terminated_length": 2320.63818359375,
      "completions/min_length": 1151.0,
      "completions/min_terminated_length": 1151.0,
      "entropy": 0.30612772703170776,
      "epoch": 4.3798076923076925,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 1426.8148193359375,
      "learning_rate": 1e-06,
      "loss": 0.1158,
      "num_tokens": 1635162922.0,
      "reward": 0.23149944841861725,
      "reward_std": 0.09988817572593689,
      "rewards/progression_diversity/mean": -0.018513914197683334,
      "rewards/progression_diversity/std": 0.07170087099075317,
      "rewards/symbolic_reward_accuracy/mean": 0.1015625,
      "rewards/symbolic_reward_accuracy/std": 0.30236753821372986,
      "rewards/symbolic_reward_partial_score/mean": 0.5978027582168579,
      "rewards/symbolic_reward_partial_score/std": 0.2723853290081024,
      "rewards/tag_count_reward/mean": -0.0859375,
      "rewards/tag_count_reward/std": 0.28054583072662354,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0040245056152344,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 38.19234848022461,
      "step": 2733
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.277388796210289,
      "epoch": 4.381410256410256,
      "grad_norm": 0.028441831469535828,
      "learning_rate": 1e-06,
      "loss": 0.2186,
      "step": 2734
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.29677003622055054,
      "epoch": 4.38301282051282,
      "grad_norm": 0.014668729156255722,
      "learning_rate": 1e-06,
      "loss": 0.117,
      "step": 2735
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.28144995868206024,
      "epoch": 4.384615384615385,
      "grad_norm": 0.014569677412509918,
      "learning_rate": 1e-06,
      "loss": 0.219,
      "step": 2736
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.130859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3800.0,
      "completions/mean_length": 4181.333984375,
      "completions/mean_terminated_length": 2344.07861328125,
      "completions/min_length": 706.0,
      "completions/min_terminated_length": 706.0,
      "entropy": 0.277862548828125,
      "epoch": 4.386217948717949,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 822.4926147460938,
      "learning_rate": 1e-06,
      "loss": 0.2072,
      "num_tokens": 1638169845.0,
      "reward": 0.3534255623817444,
      "reward_std": 0.12979258596897125,
      "rewards/progression_diversity/mean": -0.015356136485934258,
      "rewards/progression_diversity/std": 0.0585295706987381,
      "rewards/symbolic_reward_accuracy/mean": 0.255859375,
      "rewards/symbolic_reward_accuracy/std": 0.43676990270614624,
      "rewards/symbolic_reward_partial_score/mean": 0.6857584714889526,
      "rewards/symbolic_reward_partial_score/std": 0.2829958498477936,
      "rewards/tag_count_reward/mean": -0.056640625,
      "rewards/tag_count_reward/std": 0.23138070106506348,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9974150657653809,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 42.795005798339844,
      "step": 2737
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.3079565465450287,
      "epoch": 4.387820512820513,
      "grad_norm": 3624.53125,
      "learning_rate": 1e-06,
      "loss": 0.4683,
      "step": 2738
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.2956803888082504,
      "epoch": 4.389423076923077,
      "grad_norm": 605.93310546875,
      "learning_rate": 1e-06,
      "loss": 0.216,
      "step": 2739
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.2977532297372818,
      "epoch": 4.391025641025641,
      "grad_norm": 0.01604938693344593,
      "learning_rate": 1e-06,
      "loss": 0.1553,
      "step": 2740
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.115234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3997.0,
      "completions/mean_length": 3936.32421875,
      "completions/mean_terminated_length": 2315.103759765625,
      "completions/min_length": 616.0,
      "completions/min_terminated_length": 616.0,
      "entropy": 0.288834884762764,
      "epoch": 4.392628205128205,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 13779.6708984375,
      "learning_rate": 1e-06,
      "loss": 0.2399,
      "num_tokens": 1641167563.0,
      "reward": 0.19894778728485107,
      "reward_std": 0.08240419626235962,
      "rewards/progression_diversity/mean": -0.016354724764823914,
      "rewards/progression_diversity/std": 0.06330784410238266,
      "rewards/symbolic_reward_accuracy/mean": 0.056640625,
      "rewards/symbolic_reward_accuracy/std": 0.23138070106506348,
      "rewards/symbolic_reward_partial_score/mean": 0.5764648914337158,
      "rewards/symbolic_reward_partial_score/std": 0.26688218116760254,
      "rewards/tag_count_reward/mean": -0.078125,
      "rewards/tag_count_reward/std": 0.26863065361976624,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.010166883468628,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 37.03101348876953,
      "step": 2741
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.3093618303537369,
      "epoch": 4.394230769230769,
      "grad_norm": 40587.7109375,
      "learning_rate": 1e-06,
      "loss": 2.4703,
      "step": 2742
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.31586118042469025,
      "epoch": 4.395833333333333,
      "grad_norm": 1233050.0,
      "learning_rate": 1e-06,
      "loss": 117.5929,
      "step": 2743
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.46875,
      "entropy": 0.29217708110809326,
      "epoch": 4.397435897435898,
      "grad_norm": 0.014232386834919453,
      "learning_rate": 1e-06,
      "loss": 0.1765,
      "step": 2744
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.107421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4179.0,
      "completions/mean_length": 3859.9765625,
      "completions/mean_terminated_length": 2352.708984375,
      "completions/min_length": 1055.0,
      "completions/min_terminated_length": 1055.0,
      "entropy": 0.2802654653787613,
      "epoch": 4.399038461538462,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 1434.6417236328125,
      "learning_rate": 1e-06,
      "loss": 0.2646,
      "num_tokens": 1644092767.0,
      "reward": 0.2843579947948456,
      "reward_std": 0.11489598453044891,
      "rewards/progression_diversity/mean": -0.01488535013049841,
      "rewards/progression_diversity/std": 0.06615662574768066,
      "rewards/symbolic_reward_accuracy/mean": 0.166015625,
      "rewards/symbolic_reward_accuracy/std": 0.3724585771560669,
      "rewards/symbolic_reward_partial_score/mean": 0.6384602785110474,
      "rewards/symbolic_reward_partial_score/std": 0.2716268301010132,
      "rewards/tag_count_reward/mean": -0.06640625,
      "rewards/tag_count_reward/std": 0.2492343932390213,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0159142017364502,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 34.84583282470703,
      "step": 2745
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.3136250674724579,
      "epoch": 4.4006410256410255,
      "grad_norm": 4538.53564453125,
      "learning_rate": 1e-06,
      "loss": 0.6472,
      "step": 2746
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.3199605792760849,
      "epoch": 4.402243589743589,
      "grad_norm": 3800.058349609375,
      "learning_rate": 1e-06,
      "loss": 0.4546,
      "step": 2747
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.31357628107070923,
      "epoch": 4.403846153846154,
      "grad_norm": 11993.4921875,
      "learning_rate": 1e-06,
      "loss": 1.4673,
      "step": 2748
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.126953125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4939.0,
      "completions/mean_length": 4187.60546875,
      "completions/mean_terminated_length": 2414.08056640625,
      "completions/min_length": 1315.0,
      "completions/min_terminated_length": 1315.0,
      "entropy": 0.30130642652511597,
      "epoch": 4.405448717948718,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 1137.6595458984375,
      "learning_rate": 1e-06,
      "loss": 0.075,
      "num_tokens": 1647127509.0,
      "reward": 0.23044368624687195,
      "reward_std": 0.11951728165149689,
      "rewards/progression_diversity/mean": -0.023503467440605164,
      "rewards/progression_diversity/std": 0.08980035781860352,
      "rewards/symbolic_reward_accuracy/mean": 0.09765625,
      "rewards/symbolic_reward_accuracy/std": 0.29713961482048035,
      "rewards/symbolic_reward_partial_score/mean": 0.597705066204071,
      "rewards/symbolic_reward_partial_score/std": 0.25887858867645264,
      "rewards/tag_count_reward/mean": -0.072265625,
      "rewards/tag_count_reward/std": 0.2591804563999176,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0048857927322388,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 40.13702392578125,
      "step": 2749
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.29961445927619934,
      "epoch": 4.407051282051282,
      "grad_norm": 0.05410754680633545,
      "learning_rate": 1e-06,
      "loss": 0.1237,
      "step": 2750
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.453125,
      "entropy": 0.3096155524253845,
      "epoch": 4.408653846153846,
      "grad_norm": 0.017144974321126938,
      "learning_rate": 1e-06,
      "loss": 0.1487,
      "step": 2751
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.203125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2930636405944824,
      "epoch": 4.410256410256411,
      "grad_norm": 0.6245795488357544,
      "learning_rate": 1e-06,
      "loss": 0.184,
      "step": 2752
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.064453125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4303.0,
      "completions/mean_length": 3234.2734375,
      "completions/mean_terminated_length": 2328.342529296875,
      "completions/min_length": 724.0,
      "completions/min_terminated_length": 724.0,
      "entropy": 0.3245136886835098,
      "epoch": 4.4118589743589745,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 1545.5665283203125,
      "learning_rate": 1e-06,
      "loss": 0.1395,
      "num_tokens": 1649663425.0,
      "reward": 0.3059009313583374,
      "reward_std": 0.08847295492887497,
      "rewards/progression_diversity/mean": -0.014401247724890709,
      "rewards/progression_diversity/std": 0.06953856348991394,
      "rewards/symbolic_reward_accuracy/mean": 0.171875,
      "rewards/symbolic_reward_accuracy/std": 0.3776407241821289,
      "rewards/symbolic_reward_partial_score/mean": 0.6913737058639526,
      "rewards/symbolic_reward_partial_score/std": 0.24967017769813538,
      "rewards/tag_count_reward/mean": -0.044921875,
      "rewards/tag_count_reward/std": 0.20733514428138733,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0390673875808716,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 25.484981536865234,
      "step": 2753
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.31169527769088745,
      "epoch": 4.413461538461538,
      "grad_norm": 0.02673480287194252,
      "learning_rate": 1e-06,
      "loss": 0.1752,
      "step": 2754
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.3431106209754944,
      "epoch": 4.415064102564102,
      "grad_norm": 0.0156412310898304,
      "learning_rate": 1e-06,
      "loss": 0.0603,
      "step": 2755
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.046875,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.3379392623901367,
      "epoch": 4.416666666666667,
      "grad_norm": 0.03630439192056656,
      "learning_rate": 1e-06,
      "loss": 0.0386,
      "step": 2756
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.052734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4258.0,
      "completions/mean_length": 2953.6015625,
      "completions/mean_terminated_length": 2205.929931640625,
      "completions/min_length": 1181.0,
      "completions/min_terminated_length": 1181.0,
      "entropy": 0.33529841899871826,
      "epoch": 4.418269230769231,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 1840.3985595703125,
      "learning_rate": 1e-06,
      "loss": 0.0718,
      "num_tokens": 1652110821.0,
      "reward": 0.27117669582366943,
      "reward_std": 0.089139424264431,
      "rewards/progression_diversity/mean": -0.014164266176521778,
      "rewards/progression_diversity/std": 0.07661490887403488,
      "rewards/symbolic_reward_accuracy/mean": 0.12109375,
      "rewards/symbolic_reward_accuracy/std": 0.3265552520751953,
      "rewards/symbolic_reward_partial_score/mean": 0.6706705689430237,
      "rewards/symbolic_reward_partial_score/std": 0.23558108508586884,
      "rewards/tag_count_reward/mean": -0.025390625,
      "rewards/tag_count_reward/std": 0.15746226906776428,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.038825273513794,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 25.822206497192383,
      "step": 2757
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.3324403911828995,
      "epoch": 4.419871794871795,
      "grad_norm": 0.03022000938653946,
      "learning_rate": 1e-06,
      "loss": 0.0187,
      "step": 2758
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.32936082780361176,
      "epoch": 4.421474358974359,
      "grad_norm": 0.026491805911064148,
      "learning_rate": 1e-06,
      "loss": 0.1138,
      "step": 2759
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.109375,
      "clip_ratio/low_mean": 0.25,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.3215688169002533,
      "epoch": 4.423076923076923,
      "grad_norm": 0.019565416499972343,
      "learning_rate": 1e-06,
      "loss": 0.1158,
      "step": 2760
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4371.0,
      "completions/mean_length": 2875.587890625,
      "completions/mean_terminated_length": 2211.23974609375,
      "completions/min_length": 1132.0,
      "completions/min_terminated_length": 1132.0,
      "entropy": 0.34050194919109344,
      "epoch": 4.424679487179487,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 462.5588684082031,
      "learning_rate": 1e-06,
      "loss": 0.0736,
      "num_tokens": 1654529730.0,
      "reward": 0.3495715260505676,
      "reward_std": 0.09474267810583115,
      "rewards/progression_diversity/mean": -0.010134847834706306,
      "rewards/progression_diversity/std": 0.061041031032800674,
      "rewards/symbolic_reward_accuracy/mean": 0.22265625,
      "rewards/symbolic_reward_accuracy/std": 0.41643625497817993,
      "rewards/symbolic_reward_partial_score/mean": 0.7293782830238342,
      "rewards/symbolic_reward_partial_score/std": 0.21722783148288727,
      "rewards/tag_count_reward/mean": -0.02734375,
      "rewards/tag_count_reward/std": 0.16324250400066376,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0413200855255127,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 24.018436431884766,
      "step": 2761
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.3244868367910385,
      "epoch": 4.426282051282051,
      "grad_norm": 0.03473278507590294,
      "learning_rate": 1e-06,
      "loss": 0.0975,
      "step": 2762
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.3279813975095749,
      "epoch": 4.427884615384615,
      "grad_norm": 0.02159891277551651,
      "learning_rate": 1e-06,
      "loss": 0.0595,
      "step": 2763
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.33927109837532043,
      "epoch": 4.42948717948718,
      "grad_norm": 0.017984116449952126,
      "learning_rate": 1e-06,
      "loss": 0.0433,
      "step": 2764
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3670.0,
      "completions/mean_length": 2952.734375,
      "completions/mean_terminated_length": 2116.763671875,
      "completions/min_length": 1072.0,
      "completions/min_terminated_length": 1072.0,
      "entropy": 0.3449130952358246,
      "epoch": 4.431089743589744,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 0.01726151816546917,
      "learning_rate": 1e-06,
      "loss": 0.0012,
      "num_tokens": 1656940938.0,
      "reward": 0.3437863886356354,
      "reward_std": 0.07033580541610718,
      "rewards/progression_diversity/mean": -0.011987379752099514,
      "rewards/progression_diversity/std": 0.059770114719867706,
      "rewards/symbolic_reward_accuracy/mean": 0.21875,
      "rewards/symbolic_reward_accuracy/std": 0.41380295157432556,
      "rewards/symbolic_reward_partial_score/mean": 0.7179687023162842,
      "rewards/symbolic_reward_partial_score/std": 0.22868217527866364,
      "rewards/tag_count_reward/mean": -0.02734375,
      "rewards/tag_count_reward/std": 0.16324250400066376,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.029953956604004,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 30.53244400024414,
      "step": 2765
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.33045631647109985,
      "epoch": 4.4326923076923075,
      "grad_norm": 1069.1544189453125,
      "learning_rate": 1e-06,
      "loss": 0.1185,
      "step": 2766
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1484375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.3347640484571457,
      "epoch": 4.434294871794872,
      "grad_norm": 0.23746109008789062,
      "learning_rate": 1e-06,
      "loss": 0.0323,
      "step": 2767
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.3243960440158844,
      "epoch": 4.435897435897436,
      "grad_norm": 0.02076563611626625,
      "learning_rate": 1e-06,
      "loss": 0.1661,
      "step": 2768
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.072265625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3358.0,
      "completions/mean_length": 3165.107421875,
      "completions/mean_terminated_length": 2135.425048828125,
      "completions/min_length": 1053.0,
      "completions/min_terminated_length": 1053.0,
      "entropy": 0.3194960504770279,
      "epoch": 4.4375,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 1015.4884643554688,
      "learning_rate": 1e-06,
      "loss": 0.1141,
      "num_tokens": 1659435105.0,
      "reward": 0.3932766914367676,
      "reward_std": 0.10632849484682083,
      "rewards/progression_diversity/mean": -0.015594206750392914,
      "rewards/progression_diversity/std": 0.06812360882759094,
      "rewards/symbolic_reward_accuracy/mean": 0.283203125,
      "rewards/symbolic_reward_accuracy/std": 0.4509948492050171,
      "rewards/symbolic_reward_partial_score/mean": 0.7548013925552368,
      "rewards/symbolic_reward_partial_score/std": 0.23602750897407532,
      "rewards/tag_count_reward/mean": -0.029296875,
      "rewards/tag_count_reward/std": 0.16880230605602264,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0133789777755737,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 38.01353073120117,
      "step": 2769
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.3396587520837784,
      "epoch": 4.439102564102564,
      "grad_norm": 615.4129638671875,
      "learning_rate": 1e-06,
      "loss": 0.0409,
      "step": 2770
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.3090101033449173,
      "epoch": 4.440705128205128,
      "grad_norm": 1137.1129150390625,
      "learning_rate": 1e-06,
      "loss": 0.1933,
      "step": 2771
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3358742445707321,
      "epoch": 4.4423076923076925,
      "grad_norm": 4.300602912902832,
      "learning_rate": 1e-06,
      "loss": 0.1023,
      "step": 2772
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3278.0,
      "completions/mean_length": 3125.611328125,
      "completions/mean_terminated_length": 2122.876220703125,
      "completions/min_length": 751.0,
      "completions/min_terminated_length": 751.0,
      "entropy": 0.3427940756082535,
      "epoch": 4.443910256410256,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 407.4725341796875,
      "learning_rate": 1e-06,
      "loss": 0.0307,
      "num_tokens": 1661810682.0,
      "reward": 0.35953429341316223,
      "reward_std": 0.1048397421836853,
      "rewards/progression_diversity/mean": -0.011415429413318634,
      "rewards/progression_diversity/std": 0.053802087903022766,
      "rewards/symbolic_reward_accuracy/mean": 0.25,
      "rewards/symbolic_reward_accuracy/std": 0.43343618512153625,
      "rewards/symbolic_reward_partial_score/mean": 0.7059895396232605,
      "rewards/symbolic_reward_partial_score/std": 0.25418156385421753,
      "rewards/tag_count_reward/mean": -0.021484375,
      "rewards/tag_count_reward/std": 0.14513419568538666,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.019977331161499,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 34.17755889892578,
      "step": 2773
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.3142138421535492,
      "epoch": 4.44551282051282,
      "grad_norm": 6694.47998046875,
      "learning_rate": 1e-06,
      "loss": 0.4636,
      "step": 2774
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.3414735943078995,
      "epoch": 4.447115384615385,
      "grad_norm": 879.877685546875,
      "learning_rate": 1e-06,
      "loss": 0.0436,
      "step": 2775
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.330064132809639,
      "epoch": 4.448717948717949,
      "grad_norm": 0.05605228990316391,
      "learning_rate": 1e-06,
      "loss": 0.0995,
      "step": 2776
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.099609375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4448.0,
      "completions/mean_length": 3544.04296875,
      "completions/mean_terminated_length": 2123.570556640625,
      "completions/min_length": 870.0,
      "completions/min_terminated_length": 870.0,
      "entropy": 0.3042691648006439,
      "epoch": 4.450320512820513,
      "frac_reward_zero_std": 0.03125,
      "grad_norm": 936.56103515625,
      "learning_rate": 1e-06,
      "loss": 0.1016,
      "num_tokens": 1664474448.0,
      "reward": 0.36131882667541504,
      "reward_std": 0.11240831017494202,
      "rewards/progression_diversity/mean": -0.013626575469970703,
      "rewards/progression_diversity/std": 0.05417032539844513,
      "rewards/symbolic_reward_accuracy/mean": 0.248046875,
      "rewards/symbolic_reward_accuracy/std": 0.4323015511035919,
      "rewards/symbolic_reward_partial_score/mean": 0.7146159410476685,
      "rewards/symbolic_reward_partial_score/std": 0.23360253870487213,
      "rewards/tag_count_reward/mean": -0.017578125,
      "rewards/tag_count_reward/std": 0.13154059648513794,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9984034299850464,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 43.34300994873047,
      "step": 2777
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.31309716403484344,
      "epoch": 4.451923076923077,
      "grad_norm": 12874.2177734375,
      "learning_rate": 1e-06,
      "loss": 0.5435,
      "step": 2778
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.3106722980737686,
      "epoch": 4.453525641025641,
      "grad_norm": 0.4278506338596344,
      "learning_rate": 1e-06,
      "loss": 0.0563,
      "step": 2779
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.307477205991745,
      "epoch": 4.455128205128205,
      "grad_norm": 0.03869732841849327,
      "learning_rate": 1e-06,
      "loss": 0.0845,
      "step": 2780
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3998.0,
      "completions/mean_length": 2993.6328125,
      "completions/mean_terminated_length": 2100.94189453125,
      "completions/min_length": 1022.0,
      "completions/min_terminated_length": 1022.0,
      "entropy": 0.33481790125370026,
      "epoch": 4.456730769230769,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 260.7884826660156,
      "learning_rate": 1e-06,
      "loss": 0.0412,
      "num_tokens": 1666869284.0,
      "reward": 0.3834180235862732,
      "reward_std": 0.1236167624592781,
      "rewards/progression_diversity/mean": -0.010737746953964233,
      "rewards/progression_diversity/std": 0.04949859157204628,
      "rewards/symbolic_reward_accuracy/mean": 0.267578125,
      "rewards/symbolic_reward_accuracy/std": 0.4431293308734894,
      "rewards/symbolic_reward_partial_score/mean": 0.7504231929779053,
      "rewards/symbolic_reward_partial_score/std": 0.22556151449680328,
      "rewards/tag_count_reward/mean": -0.021484375,
      "rewards/tag_count_reward/std": 0.14513419568538666,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0300805568695068,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 27.265310287475586,
      "step": 2781
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3359375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.31526750326156616,
      "epoch": 4.458333333333333,
      "grad_norm": 7211.5888671875,
      "learning_rate": 1e-06,
      "loss": 0.8517,
      "step": 2782
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.329397588968277,
      "epoch": 4.459935897435898,
      "grad_norm": 52.46617126464844,
      "learning_rate": 1e-06,
      "loss": 0.0521,
      "step": 2783
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.329983115196228,
      "epoch": 4.461538461538462,
      "grad_norm": 2546.868896484375,
      "learning_rate": 1e-06,
      "loss": 0.2009,
      "step": 2784
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.041015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3777.0,
      "completions/mean_length": 2726.880859375,
      "completions/mean_terminated_length": 2142.767822265625,
      "completions/min_length": 1010.0,
      "completions/min_terminated_length": 1010.0,
      "entropy": 0.32919782400131226,
      "epoch": 4.4631410256410255,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 794.0968017578125,
      "learning_rate": 1e-06,
      "loss": 0.0914,
      "num_tokens": 1669120647.0,
      "reward": 0.3136161267757416,
      "reward_std": 0.07102187722921371,
      "rewards/progression_diversity/mean": -0.006063465960323811,
      "rewards/progression_diversity/std": 0.034248657524585724,
      "rewards/symbolic_reward_accuracy/mean": 0.17578125,
      "rewards/symbolic_reward_accuracy/std": 0.3810062110424042,
      "rewards/symbolic_reward_partial_score/mean": 0.7024902105331421,
      "rewards/symbolic_reward_partial_score/std": 0.2275681048631668,
      "rewards/tag_count_reward/mean": -0.025390625,
      "rewards/tag_count_reward/std": 0.15746226906776428,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.046392560005188,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 19.699295043945312,
      "step": 2785
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.3417544662952423,
      "epoch": 4.464743589743589,
      "grad_norm": 46041.80078125,
      "learning_rate": 1e-06,
      "loss": 0.8061,
      "step": 2786
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.35066042840480804,
      "epoch": 4.466346153846154,
      "grad_norm": 0.02314118668437004,
      "learning_rate": 1e-06,
      "loss": 0.0297,
      "step": 2787
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.34087491035461426,
      "epoch": 4.467948717948718,
      "grad_norm": 5573.73779296875,
      "learning_rate": 1e-06,
      "loss": 0.1796,
      "step": 2788
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3409.0,
      "completions/mean_length": 2916.076171875,
      "completions/mean_terminated_length": 2195.56982421875,
      "completions/min_length": 1118.0,
      "completions/min_terminated_length": 1118.0,
      "entropy": 0.34819819033145905,
      "epoch": 4.469551282051282,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 49.569732666015625,
      "learning_rate": 1e-06,
      "loss": 0.0002,
      "num_tokens": 1671554350.0,
      "reward": 0.4041130542755127,
      "reward_std": 0.06181896850466728,
      "rewards/progression_diversity/mean": -0.005198919214308262,
      "rewards/progression_diversity/std": 0.02956199087202549,
      "rewards/symbolic_reward_accuracy/mean": 0.3046875,
      "rewards/symbolic_reward_accuracy/std": 0.4607250988483429,
      "rewards/symbolic_reward_partial_score/mean": 0.7482584714889526,
      "rewards/symbolic_reward_partial_score/std": 0.23627658188343048,
      "rewards/tag_count_reward/mean": -0.03125,
      "rewards/tag_count_reward/std": 0.17416280508041382,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.037619948387146,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 22.79637908935547,
      "step": 2789
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.3270893096923828,
      "epoch": 4.471153846153846,
      "grad_norm": 3314.6259765625,
      "learning_rate": 1e-06,
      "loss": 4.353,
      "step": 2790
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.31063470244407654,
      "epoch": 4.472756410256411,
      "grad_norm": 5541.47216796875,
      "learning_rate": 1e-06,
      "loss": 0.4331,
      "step": 2791
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.3323570489883423,
      "epoch": 4.4743589743589745,
      "grad_norm": 0.013452508486807346,
      "learning_rate": 1e-06,
      "loss": 0.0806,
      "step": 2792
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.044921875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4269.0,
      "completions/mean_length": 2823.076171875,
      "completions/mean_terminated_length": 2185.2412109375,
      "completions/min_length": 677.0,
      "completions/min_terminated_length": 677.0,
      "entropy": 0.35660889744758606,
      "epoch": 4.475961538461538,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 459.7804260253906,
      "learning_rate": 1e-06,
      "loss": 0.0029,
      "num_tokens": 1673850565.0,
      "reward": 0.3343041241168976,
      "reward_std": 0.0574406236410141,
      "rewards/progression_diversity/mean": -0.001719512976706028,
      "rewards/progression_diversity/std": 0.017452143132686615,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.6970865726470947,
      "rewards/symbolic_reward_partial_score/std": 0.22637033462524414,
      "rewards/tag_count_reward/mean": -0.013671875,
      "rewards/tag_count_reward/std": 0.1162383034825325,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0488998889923096,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 17.94207000732422,
      "step": 2793
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.340995192527771,
      "epoch": 4.477564102564102,
      "grad_norm": 0.02110283449292183,
      "learning_rate": 1e-06,
      "loss": 0.0215,
      "step": 2794
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.331098347902298,
      "epoch": 4.479166666666667,
      "grad_norm": 0.03017670288681984,
      "learning_rate": 1e-06,
      "loss": 0.0531,
      "step": 2795
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.33193665742874146,
      "epoch": 4.480769230769231,
      "grad_norm": 0.07421034574508667,
      "learning_rate": 1e-06,
      "loss": 0.0802,
      "step": 2796
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4440.0,
      "completions/mean_length": 2845.490234375,
      "completions/mean_terminated_length": 2179.661865234375,
      "completions/min_length": 629.0,
      "completions/min_terminated_length": 629.0,
      "entropy": 0.334924578666687,
      "epoch": 4.482371794871795,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 300.2858581542969,
      "learning_rate": 1e-06,
      "loss": 0.0756,
      "num_tokens": 1676160880.0,
      "reward": 0.3624279201030731,
      "reward_std": 0.06554090976715088,
      "rewards/progression_diversity/mean": -0.0008624562760815024,
      "rewards/progression_diversity/std": 0.009423964656889439,
      "rewards/symbolic_reward_accuracy/mean": 0.244140625,
      "rewards/symbolic_reward_accuracy/std": 0.42999663949012756,
      "rewards/symbolic_reward_partial_score/mean": 0.7289550304412842,
      "rewards/symbolic_reward_partial_score/std": 0.23089763522148132,
      "rewards/tag_count_reward/mean": -0.02734375,
      "rewards/tag_count_reward/std": 0.16324250400066376,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0452535152435303,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 16.53045654296875,
      "step": 2797
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.3460453897714615,
      "epoch": 4.483974358974359,
      "grad_norm": 0.013277917169034481,
      "learning_rate": 1e-06,
      "loss": 0.0197,
      "step": 2798
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3399341404438019,
      "epoch": 4.485576923076923,
      "grad_norm": 0.018666263669729233,
      "learning_rate": 1e-06,
      "loss": 0.078,
      "step": 2799
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.32549184560775757,
      "epoch": 4.487179487179487,
      "grad_norm": 0.0136836227029562,
      "learning_rate": 1e-06,
      "loss": 0.1256,
      "step": 2800
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3747.0,
      "completions/mean_length": 2939.9921875,
      "completions/mean_terminated_length": 2162.239501953125,
      "completions/min_length": 605.0,
      "completions/min_terminated_length": 605.0,
      "entropy": 0.3398710787296295,
      "epoch": 4.488782051282051,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 727.4561157226562,
      "learning_rate": 1e-06,
      "loss": 0.0565,
      "num_tokens": 1678543564.0,
      "reward": 0.2560884952545166,
      "reward_std": 0.05889653041958809,
      "rewards/progression_diversity/mean": -3.998234387836419e-05,
      "rewards/progression_diversity/std": 0.0009046972263604403,
      "rewards/symbolic_reward_accuracy/mean": 0.115234375,
      "rewards/symbolic_reward_accuracy/std": 0.3196168541908264,
      "rewards/symbolic_reward_partial_score/mean": 0.6348795294761658,
      "rewards/symbolic_reward_partial_score/std": 0.23259754478931427,
      "rewards/tag_count_reward/mean": -0.03515625,
      "rewards/tag_count_reward/std": 0.1843547374010086,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0552442073822021,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 7.984652519226074,
      "step": 2801
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.33726924657821655,
      "epoch": 4.490384615384615,
      "grad_norm": 0.024799533188343048,
      "learning_rate": 1e-06,
      "loss": 0.0782,
      "step": 2802
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3415806442499161,
      "epoch": 4.49198717948718,
      "grad_norm": 0.01905086636543274,
      "learning_rate": 1e-06,
      "loss": 0.05,
      "step": 2803
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.32189974188804626,
      "epoch": 4.493589743589744,
      "grad_norm": 0.01600506529211998,
      "learning_rate": 1e-06,
      "loss": 0.1482,
      "step": 2804
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.048828125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4592.0,
      "completions/mean_length": 2884.654296875,
      "completions/mean_terminated_length": 2191.66943359375,
      "completions/min_length": 787.0,
      "completions/min_terminated_length": 787.0,
      "entropy": 0.33240219950675964,
      "epoch": 4.4951923076923075,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1222.6678466796875,
      "learning_rate": 1e-06,
      "loss": 0.065,
      "num_tokens": 1680954971.0,
      "reward": 0.2994934320449829,
      "reward_std": 0.05192206799983978,
      "rewards/progression_diversity/mean": -0.00036605086643248796,
      "rewards/progression_diversity/std": 0.004469391889870167,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.7053548097610474,
      "rewards/symbolic_reward_partial_score/std": 0.2314625382423401,
      "rewards/tag_count_reward/mean": -0.03515625,
      "rewards/tag_count_reward/std": 0.1843547374010086,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0502550601959229,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 13.556268692016602,
      "step": 2805
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.32200056314468384,
      "epoch": 4.496794871794872,
      "grad_norm": 0.03439924120903015,
      "learning_rate": 1e-06,
      "loss": 0.1092,
      "step": 2806
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.34106314182281494,
      "epoch": 4.498397435897436,
      "grad_norm": 0.014802143909037113,
      "learning_rate": 1e-06,
      "loss": 0.0375,
      "step": 2807
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.32713207602500916,
      "epoch": 4.5,
      "grad_norm": 0.030605483800172806,
      "learning_rate": 1e-06,
      "loss": 0.1174,
      "step": 2808
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4037.0,
      "completions/mean_length": 2988.58203125,
      "completions/mean_terminated_length": 2154.842529296875,
      "completions/min_length": 983.0,
      "completions/min_terminated_length": 983.0,
      "entropy": 0.32437823712825775,
      "epoch": 4.501602564102564,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 1198.33984375,
      "learning_rate": 1e-06,
      "loss": 0.0856,
      "num_tokens": 1683353045.0,
      "reward": 0.3443542718887329,
      "reward_std": 0.09359234571456909,
      "rewards/progression_diversity/mean": -0.0006082241889089346,
      "rewards/progression_diversity/std": 0.005995499901473522,
      "rewards/symbolic_reward_accuracy/mean": 0.236328125,
      "rewards/symbolic_reward_accuracy/std": 0.42524150013923645,
      "rewards/symbolic_reward_partial_score/mean": 0.692138671875,
      "rewards/symbolic_reward_partial_score/std": 0.2482290267944336,
      "rewards/tag_count_reward/mean": -0.05078125,
      "rewards/tag_count_reward/std": 0.21976542472839355,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.045188069343567,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 16.6503849029541,
      "step": 2809
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.33782175183296204,
      "epoch": 4.503205128205128,
      "grad_norm": 0.020003831014037132,
      "learning_rate": 1e-06,
      "loss": 0.0686,
      "step": 2810
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.3277200609445572,
      "epoch": 4.5048076923076925,
      "grad_norm": 0.016141105443239212,
      "learning_rate": 1e-06,
      "loss": 0.0929,
      "step": 2811
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.32373303174972534,
      "epoch": 4.506410256410256,
      "grad_norm": 0.03574444353580475,
      "learning_rate": 1e-06,
      "loss": 0.1154,
      "step": 2812
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4322.0,
      "completions/mean_length": 2614.474609375,
      "completions/mean_terminated_length": 2112.7509765625,
      "completions/min_length": 655.0,
      "completions/min_terminated_length": 655.0,
      "entropy": 0.3422088921070099,
      "epoch": 4.50801282051282,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 256.3056335449219,
      "learning_rate": 1e-06,
      "loss": 0.057,
      "num_tokens": 1685530120.0,
      "reward": 0.3261813223361969,
      "reward_std": 0.06060642749071121,
      "rewards/progression_diversity/mean": -0.0005208561196923256,
      "rewards/progression_diversity/std": 0.005044504068791866,
      "rewards/symbolic_reward_accuracy/mean": 0.189453125,
      "rewards/symbolic_reward_accuracy/std": 0.3922513723373413,
      "rewards/symbolic_reward_partial_score/mean": 0.7187988758087158,
      "rewards/symbolic_reward_partial_score/std": 0.2249433696269989,
      "rewards/tag_count_reward/mean": -0.03125,
      "rewards/tag_count_reward/std": 0.17416280508041382,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0556640625,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 14.019671440124512,
      "step": 2813
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.32429981231689453,
      "epoch": 4.509615384615385,
      "grad_norm": 0.016235308721661568,
      "learning_rate": 1e-06,
      "loss": 0.1237,
      "step": 2814
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.34095147252082825,
      "epoch": 4.511217948717949,
      "grad_norm": 0.026136431843042374,
      "learning_rate": 1e-06,
      "loss": 0.0271,
      "step": 2815
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.34518367052078247,
      "epoch": 4.512820512820513,
      "grad_norm": 0.012866949662566185,
      "learning_rate": 1e-06,
      "loss": 0.0377,
      "step": 2816
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3575.0,
      "completions/mean_length": 2740.462890625,
      "completions/mean_terminated_length": 2069.46923828125,
      "completions/min_length": 848.0,
      "completions/min_terminated_length": 848.0,
      "entropy": 0.33681435883045197,
      "epoch": 4.514423076923077,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 524.9415283203125,
      "learning_rate": 1e-06,
      "loss": 0.0291,
      "num_tokens": 1687806261.0,
      "reward": 0.4213888347148895,
      "reward_std": 0.08116651326417923,
      "rewards/progression_diversity/mean": -0.0007689363555982709,
      "rewards/progression_diversity/std": 0.008490425534546375,
      "rewards/symbolic_reward_accuracy/mean": 0.337890625,
      "rewards/symbolic_reward_accuracy/std": 0.4734536409378052,
      "rewards/symbolic_reward_partial_score/mean": 0.7386393547058105,
      "rewards/symbolic_reward_partial_score/std": 0.24396677315235138,
      "rewards/tag_count_reward/mean": -0.029296875,
      "rewards/tag_count_reward/std": 0.16880230605602264,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0432395935058594,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 18.25098419189453,
      "step": 2817
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.3194080591201782,
      "epoch": 4.516025641025641,
      "grad_norm": 0.022730786353349686,
      "learning_rate": 1e-06,
      "loss": 0.1341,
      "step": 2818
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.3404367119073868,
      "epoch": 4.517628205128205,
      "grad_norm": 0.017860587686300278,
      "learning_rate": 1e-06,
      "loss": 0.0195,
      "step": 2819
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.324150949716568,
      "epoch": 4.519230769230769,
      "grad_norm": 0.019054269418120384,
      "learning_rate": 1e-06,
      "loss": 0.0698,
      "step": 2820
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3389.0,
      "completions/mean_length": 3260.208984375,
      "completions/mean_terminated_length": 2026.348388671875,
      "completions/min_length": 650.0,
      "completions/min_terminated_length": 650.0,
      "entropy": 0.3134499490261078,
      "epoch": 4.520833333333333,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 1172.327392578125,
      "learning_rate": 1e-06,
      "loss": 0.119,
      "num_tokens": 1690392976.0,
      "reward": 0.27747392654418945,
      "reward_std": 0.07828624546527863,
      "rewards/progression_diversity/mean": -0.0016331763472408056,
      "rewards/progression_diversity/std": 0.009813708253204823,
      "rewards/symbolic_reward_accuracy/mean": 0.140625,
      "rewards/symbolic_reward_accuracy/std": 0.3479743003845215,
      "rewards/symbolic_reward_partial_score/mean": 0.66455078125,
      "rewards/symbolic_reward_partial_score/std": 0.2490902692079544,
      "rewards/tag_count_reward/mean": -0.0625,
      "rewards/tag_count_reward/std": 0.2422981858253479,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0194971561431885,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 25.523792266845703,
      "step": 2821
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.28017091751098633,
      "epoch": 4.522435897435898,
      "grad_norm": 0.024804720655083656,
      "learning_rate": 1e-06,
      "loss": 0.1724,
      "step": 2822
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.32716619968414307,
      "epoch": 4.524038461538462,
      "grad_norm": 0.02028745971620083,
      "learning_rate": 1e-06,
      "loss": 0.0362,
      "step": 2823
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.293626606464386,
      "epoch": 4.5256410256410255,
      "grad_norm": 0.01859315298497677,
      "learning_rate": 1e-06,
      "loss": 0.152,
      "step": 2824
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4031.0,
      "completions/mean_length": 2742.6484375,
      "completions/mean_terminated_length": 2012.8641357421875,
      "completions/min_length": 796.0,
      "completions/min_terminated_length": 796.0,
      "entropy": 0.33004169166088104,
      "epoch": 4.527243589743589,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 179.26219177246094,
      "learning_rate": 1e-06,
      "loss": 0.0294,
      "num_tokens": 1692678172.0,
      "reward": 0.3905848264694214,
      "reward_std": 0.056361980736255646,
      "rewards/progression_diversity/mean": -0.00011124266166007146,
      "rewards/progression_diversity/std": 0.002228989265859127,
      "rewards/symbolic_reward_accuracy/mean": 0.298828125,
      "rewards/symbolic_reward_accuracy/std": 0.45819199085235596,
      "rewards/symbolic_reward_partial_score/mean": 0.716015636920929,
      "rewards/symbolic_reward_partial_score/std": 0.2630784511566162,
      "rewards/tag_count_reward/mean": -0.03515625,
      "rewards/tag_count_reward/std": 0.1843547374010086,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.041427493095398,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 19.09334945678711,
      "step": 2825
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.310644194483757,
      "epoch": 4.528846153846154,
      "grad_norm": 0.01855999417603016,
      "learning_rate": 1e-06,
      "loss": 0.1514,
      "step": 2826
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.203125,
      "entropy": 0.33532649278640747,
      "epoch": 4.530448717948718,
      "grad_norm": 0.017645422369241714,
      "learning_rate": 1e-06,
      "loss": 0.0642,
      "step": 2827
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.3198052793741226,
      "epoch": 4.532051282051282,
      "grad_norm": 0.01399680133908987,
      "learning_rate": 1e-06,
      "loss": 0.0771,
      "step": 2828
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.11328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4372.0,
      "completions/mean_length": 3640.142578125,
      "completions/mean_terminated_length": 2012.0726318359375,
      "completions/min_length": 946.0,
      "completions/min_terminated_length": 946.0,
      "entropy": 0.3003094792366028,
      "epoch": 4.533653846153846,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 824.9033813476562,
      "learning_rate": 1e-06,
      "loss": 0.0933,
      "num_tokens": 1695515621.0,
      "reward": 0.29925772547721863,
      "reward_std": 0.10430952906608582,
      "rewards/progression_diversity/mean": -1.0345164810132701e-05,
      "rewards/progression_diversity/std": 0.00019557196355890483,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.6570312976837158,
      "rewards/symbolic_reward_partial_score/std": 0.28810763359069824,
      "rewards/tag_count_reward/mean": -0.103515625,
      "rewards/tag_count_reward/std": 0.30492907762527466,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0081380605697632,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 32.27723693847656,
      "step": 2829
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2708180397748947,
      "epoch": 4.535256410256411,
      "grad_norm": 0.02845774032175541,
      "learning_rate": 1e-06,
      "loss": 0.1828,
      "step": 2830
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.28906436264514923,
      "epoch": 4.5368589743589745,
      "grad_norm": 0.015066894702613354,
      "learning_rate": 1e-06,
      "loss": 0.1517,
      "step": 2831
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.2919054627418518,
      "epoch": 4.538461538461538,
      "grad_norm": 0.027522152289748192,
      "learning_rate": 1e-06,
      "loss": 0.1443,
      "step": 2832
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4253.0,
      "completions/mean_length": 3101.431640625,
      "completions/mean_terminated_length": 1975.790283203125,
      "completions/min_length": 769.0,
      "completions/min_terminated_length": 769.0,
      "entropy": 0.32231228053569794,
      "epoch": 4.540064102564102,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 864.8489990234375,
      "learning_rate": 1e-06,
      "loss": 0.0609,
      "num_tokens": 1697946786.0,
      "reward": 0.3166502118110657,
      "reward_std": 0.07383254915475845,
      "rewards/progression_diversity/mean": -1.783898551366292e-05,
      "rewards/progression_diversity/std": 0.0004036501923110336,
      "rewards/symbolic_reward_accuracy/mean": 0.19140625,
      "rewards/symbolic_reward_accuracy/std": 0.3937928080558777,
      "rewards/symbolic_reward_partial_score/mean": 0.6961263418197632,
      "rewards/symbolic_reward_partial_score/std": 0.26130229234695435,
      "rewards/tag_count_reward/mean": -0.0703125,
      "rewards/tag_count_reward/std": 0.25592297315597534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0254395008087158,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 25.37073516845703,
      "step": 2833
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.29585976898670197,
      "epoch": 4.541666666666667,
      "grad_norm": 0.015048340894281864,
      "learning_rate": 1e-06,
      "loss": 0.1864,
      "step": 2834
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.31169450283050537,
      "epoch": 4.543269230769231,
      "grad_norm": 0.013053800910711288,
      "learning_rate": 1e-06,
      "loss": 0.1106,
      "step": 2835
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.3154265433549881,
      "epoch": 4.544871794871795,
      "grad_norm": 0.02868885174393654,
      "learning_rate": 1e-06,
      "loss": 0.0882,
      "step": 2836
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.068359375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4116.0,
      "completions/mean_length": 2923.740234375,
      "completions/mean_terminated_length": 1936.090087890625,
      "completions/min_length": 777.0,
      "completions/min_terminated_length": 777.0,
      "entropy": 0.2929193377494812,
      "epoch": 4.546474358974359,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1734.2320556640625,
      "learning_rate": 1e-06,
      "loss": 0.1872,
      "num_tokens": 1700301149.0,
      "reward": 0.2965981960296631,
      "reward_std": 0.07857708632946014,
      "rewards/progression_diversity/mean": -0.0008270339458249509,
      "rewards/progression_diversity/std": 0.01284003909677267,
      "rewards/symbolic_reward_accuracy/mean": 0.169921875,
      "rewards/symbolic_reward_accuracy/std": 0.3759314715862274,
      "rewards/symbolic_reward_partial_score/mean": 0.669677734375,
      "rewards/symbolic_reward_partial_score/std": 0.2715960741043091,
      "rewards/tag_count_reward/mean": -0.0625,
      "rewards/tag_count_reward/std": 0.2422981858253479,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0229657888412476,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 27.50617790222168,
      "step": 2837
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.3156014531850815,
      "epoch": 4.548076923076923,
      "grad_norm": 0.010393503122031689,
      "learning_rate": 1e-06,
      "loss": 0.1186,
      "step": 2838
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.32449236512184143,
      "epoch": 4.549679487179487,
      "grad_norm": 0.015459039248526096,
      "learning_rate": 1e-06,
      "loss": 0.0538,
      "step": 2839
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.31201881170272827,
      "epoch": 4.551282051282051,
      "grad_norm": 0.014963453635573387,
      "learning_rate": 1e-06,
      "loss": 0.0831,
      "step": 2840
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.111328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3925.0,
      "completions/mean_length": 3557.302734375,
      "completions/mean_terminated_length": 1950.44189453125,
      "completions/min_length": 771.0,
      "completions/min_terminated_length": 771.0,
      "entropy": 0.30540113151073456,
      "epoch": 4.552884615384615,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 626.2611083984375,
      "learning_rate": 1e-06,
      "loss": 0.077,
      "num_tokens": 1703068392.0,
      "reward": 0.24340294301509857,
      "reward_std": 0.11287408322095871,
      "rewards/progression_diversity/mean": -4.04381935368292e-05,
      "rewards/progression_diversity/std": 0.0005320300115272403,
      "rewards/symbolic_reward_accuracy/mean": 0.1171875,
      "rewards/symbolic_reward_accuracy/std": 0.32195815443992615,
      "rewards/symbolic_reward_partial_score/mean": 0.6095215082168579,
      "rewards/symbolic_reward_partial_score/std": 0.27623260021209717,
      "rewards/tag_count_reward/mean": -0.09765625,
      "rewards/tag_count_reward/std": 0.29713961482048035,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9910718202590942,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 40.45024108886719,
      "step": 2841
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.2842327356338501,
      "epoch": 4.55448717948718,
      "grad_norm": 0.02323756366968155,
      "learning_rate": 1e-06,
      "loss": 0.1819,
      "step": 2842
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.28665095567703247,
      "epoch": 4.556089743589744,
      "grad_norm": 0.018670253455638885,
      "learning_rate": 1e-06,
      "loss": 0.1907,
      "step": 2843
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.28130021691322327,
      "epoch": 4.5576923076923075,
      "grad_norm": 0.014929386787116528,
      "learning_rate": 1e-06,
      "loss": 0.1567,
      "step": 2844
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.068359375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4265.0,
      "completions/mean_length": 2879.638671875,
      "completions/mean_terminated_length": 1888.7525634765625,
      "completions/min_length": 747.0,
      "completions/min_terminated_length": 747.0,
      "entropy": 0.33259910345077515,
      "epoch": 4.559294871794872,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 539.2830810546875,
      "learning_rate": 1e-06,
      "loss": 0.0357,
      "num_tokens": 1705342479.0,
      "reward": 0.309814453125,
      "reward_std": 0.08594641089439392,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.6902669668197632,
      "rewards/symbolic_reward_partial_score/std": 0.2536517083644867,
      "rewards/tag_count_reward/mean": -0.05078125,
      "rewards/tag_count_reward/std": 0.21976542472839355,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0254346132278442,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 25.892318725585938,
      "step": 2845
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.3099818527698517,
      "epoch": 4.560897435897436,
      "grad_norm": 0.01687266118824482,
      "learning_rate": 1e-06,
      "loss": 0.1278,
      "step": 2846
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.3287712037563324,
      "epoch": 4.5625,
      "grad_norm": 0.017019418999552727,
      "learning_rate": 1e-06,
      "loss": 0.0366,
      "step": 2847
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.3039703518152237,
      "epoch": 4.564102564102564,
      "grad_norm": 0.02699456736445427,
      "learning_rate": 1e-06,
      "loss": 0.1679,
      "step": 2848
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.095703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3132.0,
      "completions/mean_length": 3338.453125,
      "completions/mean_terminated_length": 1957.8228759765625,
      "completions/min_length": 890.0,
      "completions/min_terminated_length": 890.0,
      "entropy": 0.3064081221818924,
      "epoch": 4.565705128205128,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 517.185302734375,
      "learning_rate": 1e-06,
      "loss": 0.109,
      "num_tokens": 1707884951.0,
      "reward": 0.35695183277130127,
      "reward_std": 0.07652982324361801,
      "rewards/progression_diversity/mean": -0.00013049774861428887,
      "rewards/progression_diversity/std": 0.002952827140688896,
      "rewards/symbolic_reward_accuracy/mean": 0.2421875,
      "rewards/symbolic_reward_accuracy/std": 0.42882615327835083,
      "rewards/symbolic_reward_partial_score/mean": 0.7328125238418579,
      "rewards/symbolic_reward_partial_score/std": 0.2740279734134674,
      "rewards/tag_count_reward/mean": -0.08203125,
      "rewards/tag_count_reward/std": 0.2746807038784027,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0161539316177368,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 28.80113983154297,
      "step": 2849
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.3108035773038864,
      "epoch": 4.5673076923076925,
      "grad_norm": 4669.63525390625,
      "learning_rate": 1e-06,
      "loss": 0.3536,
      "step": 2850
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.2923634201288223,
      "epoch": 4.568910256410256,
      "grad_norm": 0.014963356778025627,
      "learning_rate": 1e-06,
      "loss": 0.1355,
      "step": 2851
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.3147512972354889,
      "epoch": 4.57051282051282,
      "grad_norm": 0.009707149118185043,
      "learning_rate": 1e-06,
      "loss": 0.1122,
      "step": 2852
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.134765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4114.0,
      "completions/mean_length": 3972.37890625,
      "completions/mean_terminated_length": 2039.19189453125,
      "completions/min_length": 601.0,
      "completions/min_terminated_length": 601.0,
      "entropy": 0.2673351615667343,
      "epoch": 4.572115384615385,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 499.8143005371094,
      "learning_rate": 1e-06,
      "loss": 0.1695,
      "num_tokens": 1710790361.0,
      "reward": 0.23631691932678223,
      "reward_std": 0.09914463013410568,
      "rewards/progression_diversity/mean": -0.0001452151482226327,
      "rewards/progression_diversity/std": 0.00327769061550498,
      "rewards/symbolic_reward_accuracy/mean": 0.103515625,
      "rewards/symbolic_reward_accuracy/std": 0.30492907762527466,
      "rewards/symbolic_reward_partial_score/mean": 0.6158528327941895,
      "rewards/symbolic_reward_partial_score/std": 0.28579404950141907,
      "rewards/tag_count_reward/mean": -0.10546875,
      "rewards/tag_count_reward/std": 0.3074568510055542,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9861880540847778,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 42.465789794921875,
      "step": 2853
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.301104798913002,
      "epoch": 4.573717948717949,
      "grad_norm": 0.01647542603313923,
      "learning_rate": 1e-06,
      "loss": 0.1265,
      "step": 2854
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.2893615812063217,
      "epoch": 4.575320512820513,
      "grad_norm": 0.02017975226044655,
      "learning_rate": 1e-06,
      "loss": 0.1213,
      "step": 2855
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.26380933821201324,
      "epoch": 4.576923076923077,
      "grad_norm": 0.014531499706208706,
      "learning_rate": 1e-06,
      "loss": 0.2342,
      "step": 2856
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.14453125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3445.0,
      "completions/mean_length": 4089.650390625,
      "completions/mean_terminated_length": 2012.522705078125,
      "completions/min_length": 817.0,
      "completions/min_terminated_length": 817.0,
      "entropy": 0.27838708460330963,
      "epoch": 4.578525641025641,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 3233.53173828125,
      "learning_rate": 1e-06,
      "loss": 0.1156,
      "num_tokens": 1713713190.0,
      "reward": 0.2200063169002533,
      "reward_std": 0.08358844369649887,
      "rewards/progression_diversity/mean": -0.0008350086864084005,
      "rewards/progression_diversity/std": 0.0046890368685126305,
      "rewards/symbolic_reward_accuracy/mean": 0.087890625,
      "rewards/symbolic_reward_accuracy/std": 0.2834126651287079,
      "rewards/symbolic_reward_partial_score/mean": 0.5986166000366211,
      "rewards/symbolic_reward_partial_score/std": 0.27839624881744385,
      "rewards/tag_count_reward/mean": -0.123046875,
      "rewards/tag_count_reward/std": 0.32881227135658264,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9900908470153809,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 42.56203842163086,
      "step": 2857
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.28409095108509064,
      "epoch": 4.580128205128205,
      "grad_norm": 359.2823486328125,
      "learning_rate": 1e-06,
      "loss": 0.1902,
      "step": 2858
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2833191752433777,
      "epoch": 4.581730769230769,
      "grad_norm": 0.015096834860742092,
      "learning_rate": 1e-06,
      "loss": 0.1434,
      "step": 2859
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.2946213483810425,
      "epoch": 4.583333333333333,
      "grad_norm": 0.01923026703298092,
      "learning_rate": 1e-06,
      "loss": 0.16,
      "step": 2860
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3878.0,
      "completions/mean_length": 4379.064453125,
      "completions/mean_terminated_length": 2022.95556640625,
      "completions/min_length": 659.0,
      "completions/min_terminated_length": 659.0,
      "entropy": 0.282021164894104,
      "epoch": 4.584935897435898,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 1836.3408203125,
      "learning_rate": 1e-06,
      "loss": 0.1013,
      "num_tokens": 1716882167.0,
      "reward": 0.23553681373596191,
      "reward_std": 0.10817793756723404,
      "rewards/progression_diversity/mean": -0.0044245729222893715,
      "rewards/progression_diversity/std": 0.01807117834687233,
      "rewards/symbolic_reward_accuracy/mean": 0.103515625,
      "rewards/symbolic_reward_accuracy/std": 0.30492907762527466,
      "rewards/symbolic_reward_partial_score/mean": 0.6192545890808105,
      "rewards/symbolic_reward_partial_score/std": 0.2875533699989319,
      "rewards/tag_count_reward/mean": -0.123046875,
      "rewards/tag_count_reward/std": 0.32881227135658264,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.978479266166687,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 46.914520263671875,
      "step": 2861
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.2763599753379822,
      "epoch": 4.586538461538462,
      "grad_norm": 0.9805912971496582,
      "learning_rate": 1e-06,
      "loss": 0.1614,
      "step": 2862
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.27411825954914093,
      "epoch": 4.5881410256410255,
      "grad_norm": 0.016598911955952644,
      "learning_rate": 1e-06,
      "loss": 0.163,
      "step": 2863
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.2109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.26957446336746216,
      "epoch": 4.589743589743589,
      "grad_norm": 0.02522032894194126,
      "learning_rate": 1e-06,
      "loss": 0.2078,
      "step": 2864
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.13671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3985.0,
      "completions/mean_length": 3941.201171875,
      "completions/mean_terminated_length": 1970.622314453125,
      "completions/min_length": 499.0,
      "completions/min_terminated_length": 499.0,
      "entropy": 0.2950494736433029,
      "epoch": 4.591346153846154,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 2163.189697265625,
      "learning_rate": 1e-06,
      "loss": 0.123,
      "num_tokens": 1719742142.0,
      "reward": 0.42619481682777405,
      "reward_std": 0.12798863649368286,
      "rewards/progression_diversity/mean": -0.001614722772501409,
      "rewards/progression_diversity/std": 0.008756079711019993,
      "rewards/symbolic_reward_accuracy/mean": 0.369140625,
      "rewards/symbolic_reward_accuracy/std": 0.4830440282821655,
      "rewards/symbolic_reward_partial_score/mean": 0.7169270515441895,
      "rewards/symbolic_reward_partial_score/std": 0.3149210810661316,
      "rewards/tag_count_reward/mean": -0.103515625,
      "rewards/tag_count_reward/std": 0.30492907762527466,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9912563562393188,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 41.91343307495117,
      "step": 2865
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2725277543067932,
      "epoch": 4.592948717948718,
      "grad_norm": 5.392195224761963,
      "learning_rate": 1e-06,
      "loss": 0.1291,
      "step": 2866
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.29304346442222595,
      "epoch": 4.594551282051282,
      "grad_norm": 0.035934969782829285,
      "learning_rate": 1e-06,
      "loss": 0.1513,
      "step": 2867
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.3078482747077942,
      "epoch": 4.596153846153846,
      "grad_norm": 0.01671292632818222,
      "learning_rate": 1e-06,
      "loss": 0.1524,
      "step": 2868
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.111328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3282.0,
      "completions/mean_length": 3543.140625,
      "completions/mean_terminated_length": 1934.505615234375,
      "completions/min_length": 658.0,
      "completions/min_terminated_length": 658.0,
      "entropy": 0.3241666555404663,
      "epoch": 4.597756410256411,
      "frac_reward_zero_std": 0.09375,
      "grad_norm": 679.676513671875,
      "learning_rate": 1e-06,
      "loss": 0.1035,
      "num_tokens": 1722401942.0,
      "reward": 0.302420049905777,
      "reward_std": 0.12250546365976334,
      "rewards/progression_diversity/mean": -0.0011605408508330584,
      "rewards/progression_diversity/std": 0.009744822047650814,
      "rewards/symbolic_reward_accuracy/mean": 0.189453125,
      "rewards/symbolic_reward_accuracy/std": 0.3922513723373413,
      "rewards/symbolic_reward_partial_score/mean": 0.656542956829071,
      "rewards/symbolic_reward_partial_score/std": 0.2855222225189209,
      "rewards/tag_count_reward/mean": -0.08203125,
      "rewards/tag_count_reward/std": 0.2746807038784027,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9899616241455078,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 41.70113754272461,
      "step": 2869
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2940093129873276,
      "epoch": 4.5993589743589745,
      "grad_norm": 135809.875,
      "learning_rate": 1e-06,
      "loss": 7.7034,
      "step": 2870
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.3006378263235092,
      "epoch": 4.600961538461538,
      "grad_norm": 0.014255443587899208,
      "learning_rate": 1e-06,
      "loss": 0.122,
      "step": 2871
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.30779320001602173,
      "epoch": 4.602564102564102,
      "grad_norm": 0.01406958419829607,
      "learning_rate": 1e-06,
      "loss": 0.1774,
      "step": 2872
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.138671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5307.0,
      "completions/mean_length": 3947.236328125,
      "completions/mean_terminated_length": 1944.945556640625,
      "completions/min_length": 608.0,
      "completions/min_terminated_length": 608.0,
      "entropy": 0.29377947747707367,
      "epoch": 4.604166666666667,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 797.7410888671875,
      "learning_rate": 1e-06,
      "loss": 0.1212,
      "num_tokens": 1725344047.0,
      "reward": 0.32524576783180237,
      "reward_std": 0.12811732292175293,
      "rewards/progression_diversity/mean": -0.007164230570197105,
      "rewards/progression_diversity/std": 0.03267419710755348,
      "rewards/symbolic_reward_accuracy/mean": 0.2265625,
      "rewards/symbolic_reward_accuracy/std": 0.4190165400505066,
      "rewards/symbolic_reward_partial_score/mean": 0.6670736074447632,
      "rewards/symbolic_reward_partial_score/std": 0.28852909803390503,
      "rewards/tag_count_reward/mean": -0.107421875,
      "rewards/tag_count_reward/std": 0.30995169281959534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9761419892311096,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 46.77880859375,
      "step": 2873
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.28681907057762146,
      "epoch": 4.605769230769231,
      "grad_norm": 0.03097737766802311,
      "learning_rate": 1e-06,
      "loss": 0.168,
      "step": 2874
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.2869716286659241,
      "epoch": 4.607371794871795,
      "grad_norm": 0.017294280230998993,
      "learning_rate": 1e-06,
      "loss": 0.1486,
      "step": 2875
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.26380398869514465,
      "epoch": 4.608974358974359,
      "grad_norm": 0.011689919047057629,
      "learning_rate": 1e-06,
      "loss": 0.204,
      "step": 2876
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.099609375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4096.0,
      "completions/mean_length": 3391.904296875,
      "completions/mean_terminated_length": 1954.6009521484375,
      "completions/min_length": 681.0,
      "completions/min_terminated_length": 681.0,
      "entropy": 0.31273353099823,
      "epoch": 4.610576923076923,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 2286.46435546875,
      "learning_rate": 1e-06,
      "loss": 0.144,
      "num_tokens": 1727923182.0,
      "reward": 0.346618115901947,
      "reward_std": 0.1120685264468193,
      "rewards/progression_diversity/mean": -0.011530693620443344,
      "rewards/progression_diversity/std": 0.04677894711494446,
      "rewards/symbolic_reward_accuracy/mean": 0.23828125,
      "rewards/symbolic_reward_accuracy/std": 0.42644867300987244,
      "rewards/symbolic_reward_partial_score/mean": 0.7052571773529053,
      "rewards/symbolic_reward_partial_score/std": 0.2821883261203766,
      "rewards/tag_count_reward/mean": -0.078125,
      "rewards/tag_count_reward/std": 0.26863065361976624,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.005671739578247,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 34.7049560546875,
      "step": 2877
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.3126651644706726,
      "epoch": 4.612179487179487,
      "grad_norm": 0.12625998258590698,
      "learning_rate": 1e-06,
      "loss": 0.0953,
      "step": 2878
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.30597206950187683,
      "epoch": 4.613782051282051,
      "grad_norm": 0.013404784724116325,
      "learning_rate": 1e-06,
      "loss": 0.1176,
      "step": 2879
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.2983405888080597,
      "epoch": 4.615384615384615,
      "grad_norm": 0.012157008983194828,
      "learning_rate": 1e-06,
      "loss": 0.1615,
      "step": 2880
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.107421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3449.0,
      "completions/mean_length": 3446.08203125,
      "completions/mean_terminated_length": 1889.002197265625,
      "completions/min_length": 558.0,
      "completions/min_terminated_length": 558.0,
      "entropy": 0.2969917505979538,
      "epoch": 4.61698717948718,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 808.1333618164062,
      "learning_rate": 1e-06,
      "loss": 0.1351,
      "num_tokens": 1730537336.0,
      "reward": 0.39082223176956177,
      "reward_std": 0.11836196482181549,
      "rewards/progression_diversity/mean": -0.011041311547160149,
      "rewards/progression_diversity/std": 0.04268914833664894,
      "rewards/symbolic_reward_accuracy/mean": 0.310546875,
      "rewards/symbolic_reward_accuracy/std": 0.46317005157470703,
      "rewards/symbolic_reward_partial_score/mean": 0.7113118767738342,
      "rewards/symbolic_reward_partial_score/std": 0.3017731308937073,
      "rewards/tag_count_reward/mean": -0.087890625,
      "rewards/tag_count_reward/std": 0.2834126651287079,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0123034715652466,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 34.338539123535156,
      "step": 2881
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.3034718632698059,
      "epoch": 4.618589743589744,
      "grad_norm": 0.022329121828079224,
      "learning_rate": 1e-06,
      "loss": 0.1786,
      "step": 2882
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.3132249712944031,
      "epoch": 4.6201923076923075,
      "grad_norm": 0.021133504807949066,
      "learning_rate": 1e-06,
      "loss": 0.1108,
      "step": 2883
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.31891536712646484,
      "epoch": 4.621794871794872,
      "grad_norm": 0.016432074829936028,
      "learning_rate": 1e-06,
      "loss": 0.0594,
      "step": 2884
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.060546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3885.0,
      "completions/mean_length": 2795.095703125,
      "completions/mean_terminated_length": 1919.3035888671875,
      "completions/min_length": 668.0,
      "completions/min_terminated_length": 668.0,
      "entropy": 0.3364175856113434,
      "epoch": 4.623397435897436,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 1410.4578857421875,
      "learning_rate": 1e-06,
      "loss": 0.0957,
      "num_tokens": 1732852201.0,
      "reward": 0.34519851207733154,
      "reward_std": 0.0896041989326477,
      "rewards/progression_diversity/mean": -0.0035866908729076385,
      "rewards/progression_diversity/std": 0.021069582551717758,
      "rewards/symbolic_reward_accuracy/mean": 0.22265625,
      "rewards/symbolic_reward_accuracy/std": 0.41643625497817993,
      "rewards/symbolic_reward_partial_score/mean": 0.7230468988418579,
      "rewards/symbolic_reward_partial_score/std": 0.24687045812606812,
      "rewards/tag_count_reward/mean": -0.052734375,
      "rewards/tag_count_reward/std": 0.22372129559516907,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.028261423110962,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 28.295589447021484,
      "step": 2885
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.32780230045318604,
      "epoch": 4.625,
      "grad_norm": 0.020152615383267403,
      "learning_rate": 1e-06,
      "loss": 0.0954,
      "step": 2886
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.3289777636528015,
      "epoch": 4.626602564102564,
      "grad_norm": 0.012928970158100128,
      "learning_rate": 1e-06,
      "loss": 0.0939,
      "step": 2887
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.33029577136039734,
      "epoch": 4.628205128205128,
      "grad_norm": 0.013741690665483475,
      "learning_rate": 1e-06,
      "loss": 0.114,
      "step": 2888
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4060.0,
      "completions/mean_length": 2811.73828125,
      "completions/mean_terminated_length": 1966.9918212890625,
      "completions/min_length": 586.0,
      "completions/min_terminated_length": 586.0,
      "entropy": 0.3604893386363983,
      "epoch": 4.6298076923076925,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 329.1974792480469,
      "learning_rate": 1e-06,
      "loss": 0.0111,
      "num_tokens": 1735088787.0,
      "reward": 0.31703460216522217,
      "reward_std": 0.06931142508983612,
      "rewards/progression_diversity/mean": -0.0021051131188869476,
      "rewards/progression_diversity/std": 0.014493024908006191,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.7118000984191895,
      "rewards/symbolic_reward_partial_score/std": 0.24182109534740448,
      "rewards/tag_count_reward/mean": -0.04296875,
      "rewards/tag_count_reward/std": 0.2029850035905838,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0364668369293213,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 24.83597183227539,
      "step": 2889
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.3358510881662369,
      "epoch": 4.631410256410256,
      "grad_norm": 0.026600396260619164,
      "learning_rate": 1e-06,
      "loss": 0.0782,
      "step": 2890
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.34051838517189026,
      "epoch": 4.63301282051282,
      "grad_norm": 0.01583019644021988,
      "learning_rate": 1e-06,
      "loss": 0.077,
      "step": 2891
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.3082561045885086,
      "epoch": 4.634615384615385,
      "grad_norm": 0.013296614401042461,
      "learning_rate": 1e-06,
      "loss": 0.1433,
      "step": 2892
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.068359375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3677.0,
      "completions/mean_length": 3038.3359375,
      "completions/mean_terminated_length": 2059.09423828125,
      "completions/min_length": 736.0,
      "completions/min_terminated_length": 736.0,
      "entropy": 0.33847953379154205,
      "epoch": 4.636217948717949,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 867.29736328125,
      "learning_rate": 1e-06,
      "loss": 0.0601,
      "num_tokens": 1737559423.0,
      "reward": 0.36477047204971313,
      "reward_std": 0.08899857103824615,
      "rewards/progression_diversity/mean": -0.001956797670572996,
      "rewards/progression_diversity/std": 0.012748402543365955,
      "rewards/symbolic_reward_accuracy/mean": 0.26171875,
      "rewards/symbolic_reward_accuracy/std": 0.44000017642974854,
      "rewards/symbolic_reward_partial_score/mean": 0.7127115726470947,
      "rewards/symbolic_reward_partial_score/std": 0.27312228083610535,
      "rewards/tag_count_reward/mean": -0.060546875,
      "rewards/tag_count_reward/std": 0.2387305200099945,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.024355411529541,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 30.130250930786133,
      "step": 2893
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.3239692151546478,
      "epoch": 4.637820512820513,
      "grad_norm": 0.012062937021255493,
      "learning_rate": 1e-06,
      "loss": 0.11,
      "step": 2894
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.3260853737592697,
      "epoch": 4.639423076923077,
      "grad_norm": 0.11582206934690475,
      "learning_rate": 1e-06,
      "loss": 0.1047,
      "step": 2895
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.31543244421482086,
      "epoch": 4.641025641025641,
      "grad_norm": 0.020565569400787354,
      "learning_rate": 1e-06,
      "loss": 0.1456,
      "step": 2896
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3929.0,
      "completions/mean_length": 2900.99609375,
      "completions/mean_terminated_length": 2120.987548828125,
      "completions/min_length": 640.0,
      "completions/min_terminated_length": 640.0,
      "entropy": 0.34089401364326477,
      "epoch": 4.642628205128205,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 597.7639770507812,
      "learning_rate": 1e-06,
      "loss": 0.0498,
      "num_tokens": 1739987869.0,
      "reward": 0.27188873291015625,
      "reward_std": 0.067967489361763,
      "rewards/progression_diversity/mean": -9.193217556457967e-05,
      "rewards/progression_diversity/std": 0.0012361678527668118,
      "rewards/symbolic_reward_accuracy/mean": 0.115234375,
      "rewards/symbolic_reward_accuracy/std": 0.3196168541908264,
      "rewards/symbolic_reward_partial_score/mean": 0.689501941204071,
      "rewards/symbolic_reward_partial_score/std": 0.22588656842708588,
      "rewards/tag_count_reward/mean": -0.041015625,
      "rewards/tag_count_reward/std": 0.19852031767368317,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0352983474731445,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 25.13549041748047,
      "step": 2897
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.3242097795009613,
      "epoch": 4.644230769230769,
      "grad_norm": 0.021813800558447838,
      "learning_rate": 1e-06,
      "loss": 0.1245,
      "step": 2898
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.34044331312179565,
      "epoch": 4.645833333333333,
      "grad_norm": 0.016379984095692635,
      "learning_rate": 1e-06,
      "loss": 0.0982,
      "step": 2899
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.33183909952640533,
      "epoch": 4.647435897435898,
      "grad_norm": 0.04149186238646507,
      "learning_rate": 1e-06,
      "loss": 3.0971,
      "step": 2900
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4292.0,
      "completions/mean_length": 2847.33203125,
      "completions/mean_terminated_length": 2064.218994140625,
      "completions/min_length": 771.0,
      "completions/min_terminated_length": 771.0,
      "entropy": 0.32430991530418396,
      "epoch": 4.649038461538462,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 1954.2508544921875,
      "learning_rate": 1e-06,
      "loss": 0.1152,
      "num_tokens": 1742326855.0,
      "reward": 0.3111421465873718,
      "reward_std": 0.07776374369859695,
      "rewards/progression_diversity/mean": -4.3267915316391736e-05,
      "rewards/progression_diversity/std": 0.0009790410986170173,
      "rewards/symbolic_reward_accuracy/mean": 0.1875,
      "rewards/symbolic_reward_accuracy/std": 0.39069411158561707,
      "rewards/symbolic_reward_partial_score/mean": 0.6751627326011658,
      "rewards/symbolic_reward_partial_score/std": 0.2522996664047241,
      "rewards/tag_count_reward/mean": -0.0390625,
      "rewards/tag_count_reward/std": 0.1939331740140915,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0431500673294067,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 21.82623863220215,
      "step": 2901
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.34375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.3435054123401642,
      "epoch": 4.6506410256410255,
      "grad_norm": 0.013913768343627453,
      "learning_rate": 1e-06,
      "loss": 0.0597,
      "step": 2902
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.34366682171821594,
      "epoch": 4.652243589743589,
      "grad_norm": 0.021470922976732254,
      "learning_rate": 1e-06,
      "loss": 0.0525,
      "step": 2903
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.3440418541431427,
      "epoch": 4.653846153846154,
      "grad_norm": 201.25079345703125,
      "learning_rate": 1e-06,
      "loss": 0.0948,
      "step": 2904
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3945.0,
      "completions/mean_length": 2559.75390625,
      "completions/mean_terminated_length": 2026.9736328125,
      "completions/min_length": 664.0,
      "completions/min_terminated_length": 664.0,
      "entropy": 0.34741640090942383,
      "epoch": 4.655448717948718,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 587.72802734375,
      "learning_rate": 1e-06,
      "loss": 0.0481,
      "num_tokens": 1744472953.0,
      "reward": 0.3951566517353058,
      "reward_std": 0.06519210338592529,
      "rewards/progression_diversity/mean": -0.0004501467919908464,
      "rewards/progression_diversity/std": 0.005880733020603657,
      "rewards/symbolic_reward_accuracy/mean": 0.283203125,
      "rewards/symbolic_reward_accuracy/std": 0.4509948492050171,
      "rewards/symbolic_reward_partial_score/mean": 0.7605631351470947,
      "rewards/symbolic_reward_partial_score/std": 0.23610848188400269,
      "rewards/tag_count_reward/mean": -0.029296875,
      "rewards/tag_count_reward/std": 0.16880230605602264,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0574297904968262,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 15.714860916137695,
      "step": 2905
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3476281762123108,
      "epoch": 4.657051282051282,
      "grad_norm": 0.025904379785060883,
      "learning_rate": 1e-06,
      "loss": 0.0288,
      "step": 2906
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.09375,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3408808410167694,
      "epoch": 4.658653846153846,
      "grad_norm": 0.02337014675140381,
      "learning_rate": 1e-06,
      "loss": 0.0792,
      "step": 2907
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1015625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.33761781454086304,
      "epoch": 4.660256410256411,
      "grad_norm": 0.013979962095618248,
      "learning_rate": 1e-06,
      "loss": 0.0681,
      "step": 2908
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3958.0,
      "completions/mean_length": 2804.318359375,
      "completions/mean_terminated_length": 2018.716796875,
      "completions/min_length": 513.0,
      "completions/min_terminated_length": 513.0,
      "entropy": 0.352565199136734,
      "epoch": 4.6618589743589745,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 0.027550842612981796,
      "learning_rate": 1e-06,
      "loss": 0.0123,
      "num_tokens": 1746766092.0,
      "reward": 0.33431804180145264,
      "reward_std": 0.07552418112754822,
      "rewards/progression_diversity/mean": -0.0008154752431437373,
      "rewards/progression_diversity/std": 0.007044443394988775,
      "rewards/symbolic_reward_accuracy/mean": 0.201171875,
      "rewards/symbolic_reward_accuracy/std": 0.4012683033943176,
      "rewards/symbolic_reward_partial_score/mean": 0.7244465947151184,
      "rewards/symbolic_reward_partial_score/std": 0.24222518503665924,
      "rewards/tag_count_reward/mean": -0.037109375,
      "rewards/tag_count_reward/std": 0.18921469151973724,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0389173030853271,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 21.87183380126953,
      "step": 2909
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.3289457857608795,
      "epoch": 4.663461538461538,
      "grad_norm": 0.018005970865488052,
      "learning_rate": 1e-06,
      "loss": 0.1036,
      "step": 2910
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.31602902710437775,
      "epoch": 4.665064102564102,
      "grad_norm": 0.015265545807778835,
      "learning_rate": 1e-06,
      "loss": 0.0927,
      "step": 2911
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.32312436401844025,
      "epoch": 4.666666666666667,
      "grad_norm": 0.010948474518954754,
      "learning_rate": 1e-06,
      "loss": 0.1255,
      "step": 2912
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3403.0,
      "completions/mean_length": 2489.4609375,
      "completions/mean_terminated_length": 1953.9715576171875,
      "completions/min_length": 735.0,
      "completions/min_terminated_length": 735.0,
      "entropy": 0.32907360792160034,
      "epoch": 4.668269230769231,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 1542.9832763671875,
      "learning_rate": 1e-06,
      "loss": 0.0546,
      "num_tokens": 1748830296.0,
      "reward": 0.3275552988052368,
      "reward_std": 0.04779002070426941,
      "rewards/progression_diversity/mean": -0.000332732277456671,
      "rewards/progression_diversity/std": 0.004427206236869097,
      "rewards/symbolic_reward_accuracy/mean": 0.1796875,
      "rewards/symbolic_reward_accuracy/std": 0.38430243730545044,
      "rewards/symbolic_reward_partial_score/mean": 0.7376953363418579,
      "rewards/symbolic_reward_partial_score/std": 0.21765783429145813,
      "rewards/tag_count_reward/mean": -0.015625,
      "rewards/tag_count_reward/std": 0.12414088100194931,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0540077686309814,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 14.962639808654785,
      "step": 2913
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.3347413241863251,
      "epoch": 4.669871794871795,
      "grad_norm": 0.02077837660908699,
      "learning_rate": 1e-06,
      "loss": 0.0577,
      "step": 2914
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.35084839165210724,
      "epoch": 4.671474358974359,
      "grad_norm": 0.1329037845134735,
      "learning_rate": 1e-06,
      "loss": 0.0075,
      "step": 2915
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.3418993204832077,
      "epoch": 4.673076923076923,
      "grad_norm": 0.035508181899785995,
      "learning_rate": 1e-06,
      "loss": 0.0666,
      "step": 2916
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.056640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3737.0,
      "completions/mean_length": 2839.4609375,
      "completions/mean_terminated_length": 2026.227783203125,
      "completions/min_length": 671.0,
      "completions/min_terminated_length": 671.0,
      "entropy": 0.3173777014017105,
      "epoch": 4.674679487179487,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 1100.101806640625,
      "learning_rate": 1e-06,
      "loss": 0.1087,
      "num_tokens": 1751308340.0,
      "reward": 0.3337549567222595,
      "reward_std": 0.0646740198135376,
      "rewards/progression_diversity/mean": -0.000972743786405772,
      "rewards/progression_diversity/std": 0.006808849982917309,
      "rewards/symbolic_reward_accuracy/mean": 0.21875,
      "rewards/symbolic_reward_accuracy/std": 0.41380295157432556,
      "rewards/symbolic_reward_partial_score/mean": 0.690673828125,
      "rewards/symbolic_reward_partial_score/std": 0.2509380280971527,
      "rewards/tag_count_reward/mean": -0.046875,
      "rewards/tag_count_reward/std": 0.21157780289649963,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.034561276435852,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 23.243541717529297,
      "step": 2917
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.3356782793998718,
      "epoch": 4.676282051282051,
      "grad_norm": 0.01395330298691988,
      "learning_rate": 1e-06,
      "loss": 0.043,
      "step": 2918
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.30980144441127777,
      "epoch": 4.677884615384615,
      "grad_norm": 0.015020109713077545,
      "learning_rate": 1e-06,
      "loss": 0.1587,
      "step": 2919
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.3269190788269043,
      "epoch": 4.67948717948718,
      "grad_norm": 0.10749267786741257,
      "learning_rate": 1e-06,
      "loss": 0.0676,
      "step": 2920
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.060546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3986.0,
      "completions/mean_length": 2943.259765625,
      "completions/mean_terminated_length": 2077.0166015625,
      "completions/min_length": 665.0,
      "completions/min_terminated_length": 665.0,
      "entropy": 0.3407384008169174,
      "epoch": 4.681089743589744,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1545.7191162109375,
      "learning_rate": 1e-06,
      "loss": 0.0606,
      "num_tokens": 1753724633.0,
      "reward": 0.3297162652015686,
      "reward_std": 0.0669044628739357,
      "rewards/progression_diversity/mean": -0.0015185039956122637,
      "rewards/progression_diversity/std": 0.012093695811927319,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.6928547620773315,
      "rewards/symbolic_reward_partial_score/std": 0.24812039732933044,
      "rewards/tag_count_reward/mean": -0.046875,
      "rewards/tag_count_reward/std": 0.21157780289649963,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0323323011398315,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 27.282554626464844,
      "step": 2921
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.0390625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.34692129492759705,
      "epoch": 4.6826923076923075,
      "grad_norm": 0.01900269091129303,
      "learning_rate": 1e-06,
      "loss": 0.0044,
      "step": 2922
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.3168996274471283,
      "epoch": 4.684294871794872,
      "grad_norm": 0.011829662136733532,
      "learning_rate": 1e-06,
      "loss": 0.1236,
      "step": 2923
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.3172369599342346,
      "epoch": 4.685897435897436,
      "grad_norm": 0.026483498513698578,
      "learning_rate": 1e-06,
      "loss": 0.1768,
      "step": 2924
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.037109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3942.0,
      "completions/mean_length": 2581.607421875,
      "completions/mean_terminated_length": 2049.66943359375,
      "completions/min_length": 836.0,
      "completions/min_terminated_length": 836.0,
      "entropy": 0.3523339778184891,
      "epoch": 4.6875,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 861.6783447265625,
      "learning_rate": 1e-06,
      "loss": 0.0552,
      "num_tokens": 1755836432.0,
      "reward": 0.2985629141330719,
      "reward_std": 0.05261341854929924,
      "rewards/progression_diversity/mean": -0.0006439232383854687,
      "rewards/progression_diversity/std": 0.005445914342999458,
      "rewards/symbolic_reward_accuracy/mean": 0.1484375,
      "rewards/symbolic_reward_accuracy/std": 0.35588082671165466,
      "rewards/symbolic_reward_partial_score/mean": 0.7100748419761658,
      "rewards/symbolic_reward_partial_score/std": 0.2193126529455185,
      "rewards/tag_count_reward/mean": -0.03515625,
      "rewards/tag_count_reward/std": 0.1843547374010086,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0577545166015625,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 16.59449577331543,
      "step": 2925
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.34911976754665375,
      "epoch": 4.689102564102564,
      "grad_norm": 0.014677703380584717,
      "learning_rate": 1e-06,
      "loss": 0.0362,
      "step": 2926
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.34325067698955536,
      "epoch": 4.690705128205128,
      "grad_norm": 0.016701413318514824,
      "learning_rate": 1e-06,
      "loss": 0.0602,
      "step": 2927
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.34585435688495636,
      "epoch": 4.6923076923076925,
      "grad_norm": 0.026525946334004402,
      "learning_rate": 1e-06,
      "loss": 0.0583,
      "step": 2928
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3789.0,
      "completions/mean_length": 2502.28515625,
      "completions/mean_terminated_length": 2054.48779296875,
      "completions/min_length": 727.0,
      "completions/min_terminated_length": 727.0,
      "entropy": 0.3411019742488861,
      "epoch": 4.693910256410256,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 701.24560546875,
      "learning_rate": 1e-06,
      "loss": 0.1057,
      "num_tokens": 1757960018.0,
      "reward": 0.4438255727291107,
      "reward_std": 0.05670292302966118,
      "rewards/progression_diversity/mean": -0.0007465816452167928,
      "rewards/progression_diversity/std": 0.006228178273886442,
      "rewards/symbolic_reward_accuracy/mean": 0.361328125,
      "rewards/symbolic_reward_accuracy/std": 0.48085519671440125,
      "rewards/symbolic_reward_partial_score/mean": 0.7639485597610474,
      "rewards/symbolic_reward_partial_score/std": 0.24817843735218048,
      "rewards/tag_count_reward/mean": -0.021484375,
      "rewards/tag_count_reward/std": 0.14513419568538666,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0602136850357056,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 15.609932899475098,
      "step": 2929
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.3556542992591858,
      "epoch": 4.69551282051282,
      "grad_norm": 0.02133711613714695,
      "learning_rate": 1e-06,
      "loss": 0.0417,
      "step": 2930
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.3535788506269455,
      "epoch": 4.697115384615385,
      "grad_norm": 0.012668832205235958,
      "learning_rate": 1e-06,
      "loss": 0.0334,
      "step": 2931
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.21875,
      "entropy": 0.34661364555358887,
      "epoch": 4.698717948717949,
      "grad_norm": 0.013216467574238777,
      "learning_rate": 1e-06,
      "loss": 0.0117,
      "step": 2932
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.044921875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3756.0,
      "completions/mean_length": 2681.73046875,
      "completions/mean_terminated_length": 2037.2474365234375,
      "completions/min_length": 798.0,
      "completions/min_terminated_length": 798.0,
      "entropy": 0.3384409695863724,
      "epoch": 4.700320512820513,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 556.1314697265625,
      "learning_rate": 1e-06,
      "loss": 0.0886,
      "num_tokens": 1760146200.0,
      "reward": 0.3850386440753937,
      "reward_std": 0.08345068991184235,
      "rewards/progression_diversity/mean": -0.0010191474575549364,
      "rewards/progression_diversity/std": 0.0071332212537527084,
      "rewards/symbolic_reward_accuracy/mean": 0.279296875,
      "rewards/symbolic_reward_accuracy/std": 0.44909247756004333,
      "rewards/symbolic_reward_partial_score/mean": 0.738574206829071,
      "rewards/symbolic_reward_partial_score/std": 0.2675086259841919,
      "rewards/tag_count_reward/mean": -0.041015625,
      "rewards/tag_count_reward/std": 0.19852031767368317,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0498528480529785,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 20.236854553222656,
      "step": 2933
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.33859123289585114,
      "epoch": 4.701923076923077,
      "grad_norm": 0.012142288498580456,
      "learning_rate": 1e-06,
      "loss": 0.0558,
      "step": 2934
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.3401936739683151,
      "epoch": 4.703525641025641,
      "grad_norm": 0.01944906637072563,
      "learning_rate": 1e-06,
      "loss": 0.0776,
      "step": 2935
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.34829023480415344,
      "epoch": 4.705128205128205,
      "grad_norm": 0.015569426119327545,
      "learning_rate": 1e-06,
      "loss": 0.0491,
      "step": 2936
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3820.0,
      "completions/mean_length": 2710.529296875,
      "completions/mean_terminated_length": 2038.0633544921875,
      "completions/min_length": 245.0,
      "completions/min_terminated_length": 245.0,
      "entropy": 0.3164035826921463,
      "epoch": 4.706730769230769,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1386.5074462890625,
      "learning_rate": 1e-06,
      "loss": 0.1388,
      "num_tokens": 1762370215.0,
      "reward": 0.3807540237903595,
      "reward_std": 0.08757130056619644,
      "rewards/progression_diversity/mean": -0.00028213387122377753,
      "rewards/progression_diversity/std": 0.002852478064596653,
      "rewards/symbolic_reward_accuracy/mean": 0.26953125,
      "rewards/symbolic_reward_accuracy/std": 0.44415023922920227,
      "rewards/symbolic_reward_partial_score/mean": 0.7457519769668579,
      "rewards/symbolic_reward_partial_score/std": 0.2586316764354706,
      "rewards/tag_count_reward/mean": -0.046875,
      "rewards/tag_count_reward/std": 0.21157780289649963,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.044259786605835,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 21.89950942993164,
      "step": 2937
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.35227370262145996,
      "epoch": 4.708333333333333,
      "grad_norm": 0.01955762878060341,
      "learning_rate": 1e-06,
      "loss": 0.0149,
      "step": 2938
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3245883733034134,
      "epoch": 4.709935897435898,
      "grad_norm": 0.016230937093496323,
      "learning_rate": 1e-06,
      "loss": 0.1317,
      "step": 2939
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1171875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.1953125,
      "entropy": 0.349644273519516,
      "epoch": 4.711538461538462,
      "grad_norm": 0.029934851452708244,
      "learning_rate": 1e-06,
      "loss": 0.0294,
      "step": 2940
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.044921875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4126.0,
      "completions/mean_length": 2700.505859375,
      "completions/mean_terminated_length": 2056.90576171875,
      "completions/min_length": 649.0,
      "completions/min_terminated_length": 649.0,
      "entropy": 0.3398941606283188,
      "epoch": 4.7131410256410255,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 1812.2177734375,
      "learning_rate": 1e-06,
      "loss": 0.0752,
      "num_tokens": 1764614458.0,
      "reward": 0.33322569727897644,
      "reward_std": 0.07418441027402878,
      "rewards/progression_diversity/mean": -0.00018588484090287238,
      "rewards/progression_diversity/std": 0.003068706952035427,
      "rewards/symbolic_reward_accuracy/mean": 0.2109375,
      "rewards/symbolic_reward_accuracy/std": 0.4083731174468994,
      "rewards/symbolic_reward_partial_score/mean": 0.7025552988052368,
      "rewards/symbolic_reward_partial_score/std": 0.25797954201698303,
      "rewards/tag_count_reward/mean": -0.041015625,
      "rewards/tag_count_reward/std": 0.19852031767368317,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0462868213653564,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 20.06338882446289,
      "step": 2941
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.34091413021087646,
      "epoch": 4.714743589743589,
      "grad_norm": 0.012703510001301765,
      "learning_rate": 1e-06,
      "loss": 0.0848,
      "step": 2942
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3203125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.3459191471338272,
      "epoch": 4.716346153846154,
      "grad_norm": 0.11944550275802612,
      "learning_rate": 1e-06,
      "loss": 0.0311,
      "step": 2943
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.31565621495246887,
      "epoch": 4.717948717948718,
      "grad_norm": 0.027011489495635033,
      "learning_rate": 1e-06,
      "loss": 0.1477,
      "step": 2944
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.052734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 5223.0,
      "completions/mean_length": 2783.65234375,
      "completions/mean_terminated_length": 2026.5196533203125,
      "completions/min_length": 704.0,
      "completions/min_terminated_length": 704.0,
      "entropy": 0.3283272385597229,
      "epoch": 4.719551282051282,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1712.640625,
      "learning_rate": 1e-06,
      "loss": 0.1465,
      "num_tokens": 1766886824.0,
      "reward": 0.4009649455547333,
      "reward_std": 0.07681870460510254,
      "rewards/progression_diversity/mean": -0.0001883438671939075,
      "rewards/progression_diversity/std": 0.0030330359004437923,
      "rewards/symbolic_reward_accuracy/mean": 0.30078125,
      "rewards/symbolic_reward_accuracy/std": 0.45904624462127686,
      "rewards/symbolic_reward_partial_score/mean": 0.751269519329071,
      "rewards/symbolic_reward_partial_score/std": 0.2660304307937622,
      "rewards/tag_count_reward/mean": -0.048828125,
      "rewards/tag_count_reward/std": 0.2157193273305893,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0471668243408203,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 20.307308197021484,
      "step": 2945
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.34363169968128204,
      "epoch": 4.721153846153846,
      "grad_norm": 0.8427232503890991,
      "learning_rate": 1e-06,
      "loss": 0.0328,
      "step": 2946
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.3473658561706543,
      "epoch": 4.722756410256411,
      "grad_norm": 0.014580237679183483,
      "learning_rate": 1e-06,
      "loss": 0.0303,
      "step": 2947
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.32401567697525024,
      "epoch": 4.7243589743589745,
      "grad_norm": 0.0191756933927536,
      "learning_rate": 1e-06,
      "loss": 0.1417,
      "step": 2948
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.083984375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3855.0,
      "completions/mean_length": 3248.55078125,
      "completions/mean_terminated_length": 2044.234619140625,
      "completions/min_length": 663.0,
      "completions/min_terminated_length": 663.0,
      "entropy": 0.31611892580986023,
      "epoch": 4.725961538461538,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 2376.0654296875,
      "learning_rate": 1e-06,
      "loss": 0.1231,
      "num_tokens": 1769473858.0,
      "reward": 0.37166327238082886,
      "reward_std": 0.13059641420841217,
      "rewards/progression_diversity/mean": -0.00017961920821107924,
      "rewards/progression_diversity/std": 0.00354160089045763,
      "rewards/symbolic_reward_accuracy/mean": 0.26953125,
      "rewards/symbolic_reward_accuracy/std": 0.44415023922920227,
      "rewards/symbolic_reward_partial_score/mean": 0.724560558795929,
      "rewards/symbolic_reward_partial_score/std": 0.29390403628349304,
      "rewards/tag_count_reward/mean": -0.07421875,
      "rewards/tag_count_reward/std": 0.2623828947544098,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.021704912185669,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 30.869539260864258,
      "step": 2949
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4296875,
      "entropy": 0.3012668788433075,
      "epoch": 4.727564102564102,
      "grad_norm": 0.017393101006746292,
      "learning_rate": 1e-06,
      "loss": 0.1819,
      "step": 2950
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.3096124082803726,
      "epoch": 4.729166666666667,
      "grad_norm": 0.22197610139846802,
      "learning_rate": 1e-06,
      "loss": 0.1091,
      "step": 2951
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.3276834934949875,
      "epoch": 4.730769230769231,
      "grad_norm": 0.0192260779440403,
      "learning_rate": 1e-06,
      "loss": 0.0842,
      "step": 2952
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.044921875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4381.0,
      "completions/mean_length": 2684.7734375,
      "completions/mean_terminated_length": 2040.4334716796875,
      "completions/min_length": 447.0,
      "completions/min_terminated_length": 447.0,
      "entropy": 0.3336394876241684,
      "epoch": 4.732371794871795,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 2094.456298828125,
      "learning_rate": 1e-06,
      "loss": 0.0981,
      "num_tokens": 1771661390.0,
      "reward": 0.2928215265274048,
      "reward_std": 0.06420518457889557,
      "rewards/progression_diversity/mean": -7.55006549297832e-05,
      "rewards/progression_diversity/std": 0.001374648418277502,
      "rewards/symbolic_reward_accuracy/mean": 0.1484375,
      "rewards/symbolic_reward_accuracy/std": 0.35588082671165466,
      "rewards/symbolic_reward_partial_score/mean": 0.6941732168197632,
      "rewards/symbolic_reward_partial_score/std": 0.24089013040065765,
      "rewards/tag_count_reward/mean": -0.044921875,
      "rewards/tag_count_reward/std": 0.20733514428138733,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0465376377105713,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 20.815563201904297,
      "step": 2953
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.3418269008398056,
      "epoch": 4.733974358974359,
      "grad_norm": 0.02304861880838871,
      "learning_rate": 1e-06,
      "loss": 0.0803,
      "step": 2954
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3348529040813446,
      "epoch": 4.735576923076923,
      "grad_norm": 0.02030983939766884,
      "learning_rate": 1e-06,
      "loss": 0.0815,
      "step": 2955
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.3480427414178848,
      "epoch": 4.737179487179487,
      "grad_norm": 0.012364787049591541,
      "learning_rate": 1e-06,
      "loss": 0.0586,
      "step": 2956
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.06640625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3864.0,
      "completions/mean_length": 3018.962890625,
      "completions/mean_terminated_length": 2068.311767578125,
      "completions/min_length": 755.0,
      "completions/min_terminated_length": 755.0,
      "entropy": 0.30448389053344727,
      "epoch": 4.738782051282051,
      "frac_reward_zero_std": 0.0625,
      "grad_norm": 8229.33984375,
      "learning_rate": 1e-06,
      "loss": 0.1366,
      "num_tokens": 1774171099.0,
      "reward": 0.29637160897254944,
      "reward_std": 0.10101476311683655,
      "rewards/progression_diversity/mean": -4.6070697862887755e-05,
      "rewards/progression_diversity/std": 0.0010424609063193202,
      "rewards/symbolic_reward_accuracy/mean": 0.154296875,
      "rewards/symbolic_reward_accuracy/std": 0.36158639192581177,
      "rewards/symbolic_reward_partial_score/mean": 0.6994954347610474,
      "rewards/symbolic_reward_partial_score/std": 0.25719350576400757,
      "rewards/tag_count_reward/mean": -0.060546875,
      "rewards/tag_count_reward/std": 0.2387305200099945,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0173419713974,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 31.943614959716797,
      "step": 2957
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4453125,
      "entropy": 0.3160858005285263,
      "epoch": 4.740384615384615,
      "grad_norm": 0.03362959623336792,
      "learning_rate": 1e-06,
      "loss": 0.1305,
      "step": 2958
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.3264904320240021,
      "epoch": 4.74198717948718,
      "grad_norm": 0.018001394346356392,
      "learning_rate": 1e-06,
      "loss": 0.06,
      "step": 2959
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.30660150945186615,
      "epoch": 4.743589743589744,
      "grad_norm": 0.01249605417251587,
      "learning_rate": 1e-06,
      "loss": 0.144,
      "step": 2960
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3706.0,
      "completions/mean_length": 2441.875,
      "completions/mean_terminated_length": 1992.1290283203125,
      "completions/min_length": 291.0,
      "completions/min_terminated_length": 291.0,
      "entropy": 0.35260292887687683,
      "epoch": 4.7451923076923075,
      "frac_reward_zero_std": 0.375,
      "grad_norm": 488.8838806152344,
      "learning_rate": 1e-06,
      "loss": 0.021,
      "num_tokens": 1776175595.0,
      "reward": 0.36955156922340393,
      "reward_std": 0.05210120975971222,
      "rewards/progression_diversity/mean": -0.0004122033715248108,
      "rewards/progression_diversity/std": 0.007343790493905544,
      "rewards/symbolic_reward_accuracy/mean": 0.244140625,
      "rewards/symbolic_reward_accuracy/std": 0.42999663949012756,
      "rewards/symbolic_reward_partial_score/mean": 0.7539876103401184,
      "rewards/symbolic_reward_partial_score/std": 0.2148759663105011,
      "rewards/tag_count_reward/mean": -0.03125,
      "rewards/tag_count_reward/std": 0.17416280508041382,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0614118576049805,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 13.594598770141602,
      "step": 2961
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.347807839512825,
      "epoch": 4.746794871794872,
      "grad_norm": 0.016412392258644104,
      "learning_rate": 1e-06,
      "loss": 0.0841,
      "step": 2962
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1328125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2421875,
      "entropy": 0.33857110142707825,
      "epoch": 4.748397435897436,
      "grad_norm": 0.016963588073849678,
      "learning_rate": 1e-06,
      "loss": 0.0678,
      "step": 2963
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.34765180945396423,
      "epoch": 4.75,
      "grad_norm": 0.03191974386572838,
      "learning_rate": 1e-06,
      "loss": 0.0283,
      "step": 2964
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.083984375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4131.0,
      "completions/mean_length": 3280.611328125,
      "completions/mean_terminated_length": 2079.234619140625,
      "completions/min_length": 263.0,
      "completions/min_terminated_length": 263.0,
      "entropy": 0.29712438583374023,
      "epoch": 4.751602564102564,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 3667.5302734375,
      "learning_rate": 1e-06,
      "loss": 0.1519,
      "num_tokens": 1778749860.0,
      "reward": 0.3474825918674469,
      "reward_std": 0.08697932213544846,
      "rewards/progression_diversity/mean": -0.00027683598455041647,
      "rewards/progression_diversity/std": 0.002760024508461356,
      "rewards/symbolic_reward_accuracy/mean": 0.244140625,
      "rewards/symbolic_reward_accuracy/std": 0.42999663949012756,
      "rewards/symbolic_reward_partial_score/mean": 0.6973470449447632,
      "rewards/symbolic_reward_partial_score/std": 0.2716548442840576,
      "rewards/tag_count_reward/mean": -0.08203125,
      "rewards/tag_count_reward/std": 0.2746807038784027,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0109891891479492,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 36.200584411621094,
      "step": 2965
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.3089023381471634,
      "epoch": 4.753205128205128,
      "grad_norm": 0.01566610299050808,
      "learning_rate": 1e-06,
      "loss": 0.2778,
      "step": 2966
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.2989957928657532,
      "epoch": 4.7548076923076925,
      "grad_norm": 0.03153858333826065,
      "learning_rate": 1e-06,
      "loss": 0.1536,
      "step": 2967
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.3228546977043152,
      "epoch": 4.756410256410256,
      "grad_norm": 0.017119932919740677,
      "learning_rate": 1e-06,
      "loss": 0.0581,
      "step": 2968
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3993.0,
      "completions/mean_length": 3280.58203125,
      "completions/mean_terminated_length": 2048.636962890625,
      "completions/min_length": 214.0,
      "completions/min_terminated_length": 214.0,
      "entropy": 0.303473562002182,
      "epoch": 4.75801282051282,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 2500.171142578125,
      "learning_rate": 1e-06,
      "loss": 0.1769,
      "num_tokens": 1781321806.0,
      "reward": 0.26970726251602173,
      "reward_std": 0.08941252529621124,
      "rewards/progression_diversity/mean": -0.0009572784183546901,
      "rewards/progression_diversity/std": 0.009158154018223286,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.661425769329071,
      "rewards/symbolic_reward_partial_score/std": 0.2673282027244568,
      "rewards/tag_count_reward/mean": -0.083984375,
      "rewards/tag_count_reward/std": 0.2776356339454651,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0137133598327637,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 36.107513427734375,
      "step": 2969
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.3026905506849289,
      "epoch": 4.759615384615385,
      "grad_norm": 0.021056868135929108,
      "learning_rate": 1e-06,
      "loss": 0.14,
      "step": 2970
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.3188410997390747,
      "epoch": 4.761217948717949,
      "grad_norm": 0.01921817846596241,
      "learning_rate": 1e-06,
      "loss": 0.095,
      "step": 2971
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.32021498680114746,
      "epoch": 4.762820512820513,
      "grad_norm": 0.04374043643474579,
      "learning_rate": 1e-06,
      "loss": 0.0584,
      "step": 2972
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.080078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3804.0,
      "completions/mean_length": 3198.404296875,
      "completions/mean_terminated_length": 2050.61376953125,
      "completions/min_length": 758.0,
      "completions/min_terminated_length": 758.0,
      "entropy": 0.3287845551967621,
      "epoch": 4.764423076923077,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 576.1661987304688,
      "learning_rate": 1e-06,
      "loss": 0.0627,
      "num_tokens": 1783830109.0,
      "reward": 0.3199334740638733,
      "reward_std": 0.08054739236831665,
      "rewards/progression_diversity/mean": -0.0003103530907537788,
      "rewards/progression_diversity/std": 0.003829639870673418,
      "rewards/symbolic_reward_accuracy/mean": 0.193359375,
      "rewards/symbolic_reward_accuracy/std": 0.39531853795051575,
      "rewards/symbolic_reward_partial_score/mean": 0.7044758796691895,
      "rewards/symbolic_reward_partial_score/std": 0.27706295251846313,
      "rewards/tag_count_reward/mean": -0.07421875,
      "rewards/tag_count_reward/std": 0.2623828947544098,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.010635495185852,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 784.0,
      "sampling/sampling_logp_difference/mean": 36.60023498535156,
      "step": 2973
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.3081291615962982,
      "epoch": 4.766025641025641,
      "grad_norm": 0.018835294991731644,
      "learning_rate": 1e-06,
      "loss": 0.1503,
      "step": 2974
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.29927410185337067,
      "epoch": 4.767628205128205,
      "grad_norm": 0.013380290009081364,
      "learning_rate": 1e-06,
      "loss": 0.1841,
      "step": 2975
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.322013258934021,
      "epoch": 4.769230769230769,
      "grad_norm": 0.016294099390506744,
      "learning_rate": 1e-06,
      "loss": 0.09,
      "step": 2976
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4037.0,
      "completions/mean_length": 3144.052734375,
      "completions/mean_terminated_length": 2142.71240234375,
      "completions/min_length": 254.0,
      "completions/min_terminated_length": 254.0,
      "entropy": 0.3358401507139206,
      "epoch": 4.770833333333333,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1361.25830078125,
      "learning_rate": 1e-06,
      "loss": 0.0311,
      "num_tokens": 1786338952.0,
      "reward": 0.34467262029647827,
      "reward_std": 0.06627324223518372,
      "rewards/progression_diversity/mean": -2.456928814353887e-05,
      "rewards/progression_diversity/std": 0.000555939506739378,
      "rewards/symbolic_reward_accuracy/mean": 0.24609375,
      "rewards/symbolic_reward_accuracy/std": 0.4311550557613373,
      "rewards/symbolic_reward_partial_score/mean": 0.6788574457168579,
      "rewards/symbolic_reward_partial_score/std": 0.2802160680294037,
      "rewards/tag_count_reward/mean": -0.06640625,
      "rewards/tag_count_reward/std": 0.2492343932390213,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0176117420196533,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 780.0,
      "sampling/sampling_logp_difference/mean": 32.03617858886719,
      "step": 2977
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.31088291108608246,
      "epoch": 4.772435897435898,
      "grad_norm": 0.012580040842294693,
      "learning_rate": 1e-06,
      "loss": 0.1263,
      "step": 2978
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.31900927424430847,
      "epoch": 4.774038461538462,
      "grad_norm": 0.012707662768661976,
      "learning_rate": 1e-06,
      "loss": 0.0779,
      "step": 2979
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.29433473944664,
      "epoch": 4.7756410256410255,
      "grad_norm": 0.01724996045231819,
      "learning_rate": 1e-06,
      "loss": 0.2277,
      "step": 2980
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.115234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3774.0,
      "completions/mean_length": 3808.806640625,
      "completions/mean_terminated_length": 2170.97802734375,
      "completions/min_length": 231.0,
      "completions/min_terminated_length": 231.0,
      "entropy": 0.28534041345119476,
      "epoch": 4.777243589743589,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 1845.77587890625,
      "learning_rate": 1e-06,
      "loss": 0.1926,
      "num_tokens": 1789274645.0,
      "reward": 0.25106510519981384,
      "reward_std": 0.11057647317647934,
      "rewards/progression_diversity/mean": -0.0009115600259974599,
      "rewards/progression_diversity/std": 0.0063006458804011345,
      "rewards/symbolic_reward_accuracy/mean": 0.130859375,
      "rewards/symbolic_reward_accuracy/std": 0.33757632970809937,
      "rewards/symbolic_reward_partial_score/mean": 0.6110025644302368,
      "rewards/symbolic_reward_partial_score/std": 0.29573264718055725,
      "rewards/tag_count_reward/mean": -0.107421875,
      "rewards/tag_count_reward/std": 0.30995169281959534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9930451512336731,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 40.782203674316406,
      "step": 2981
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3515625,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.303585946559906,
      "epoch": 4.778846153846154,
      "grad_norm": 0.012852794490754604,
      "learning_rate": 1e-06,
      "loss": 0.1133,
      "step": 2982
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2768773287534714,
      "epoch": 4.780448717948718,
      "grad_norm": 0.10220827162265778,
      "learning_rate": 1e-06,
      "loss": 0.2116,
      "step": 2983
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.30345627665519714,
      "epoch": 4.782051282051282,
      "grad_norm": 0.06796098500490189,
      "learning_rate": 1e-06,
      "loss": 0.104,
      "step": 2984
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3753.0,
      "completions/mean_length": 3173.6015625,
      "completions/mean_terminated_length": 2174.495849609375,
      "completions/min_length": 809.0,
      "completions/min_terminated_length": 809.0,
      "entropy": 0.30603721737861633,
      "epoch": 4.783653846153846,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 5978.078125,
      "learning_rate": 1e-06,
      "loss": 0.1597,
      "num_tokens": 1791752489.0,
      "reward": 0.3074996769428253,
      "reward_std": 0.0968872681260109,
      "rewards/progression_diversity/mean": -0.0014983013970777392,
      "rewards/progression_diversity/std": 0.009809834882616997,
      "rewards/symbolic_reward_accuracy/mean": 0.19140625,
      "rewards/symbolic_reward_accuracy/std": 0.3937928080558777,
      "rewards/symbolic_reward_partial_score/mean": 0.6656738519668579,
      "rewards/symbolic_reward_partial_score/std": 0.2579022943973541,
      "rewards/tag_count_reward/mean": -0.0703125,
      "rewards/tag_count_reward/std": 0.25592297315597534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0267293453216553,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 26.516414642333984,
      "step": 2985
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.32077351212501526,
      "epoch": 4.785256410256411,
      "grad_norm": 0.03152855858206749,
      "learning_rate": 1e-06,
      "loss": 0.1114,
      "step": 2986
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.3265487253665924,
      "epoch": 4.7868589743589745,
      "grad_norm": 5331029908783104.0,
      "learning_rate": 1e-06,
      "loss": 734216716288.0,
      "step": 2987
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.31940317153930664,
      "epoch": 4.788461538461538,
      "grad_norm": 18845.392578125,
      "learning_rate": 1e-06,
      "loss": 5.7057,
      "step": 2988
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.083984375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4206.0,
      "completions/mean_length": 3355.01953125,
      "completions/mean_terminated_length": 2160.46484375,
      "completions/min_length": 952.0,
      "completions/min_terminated_length": 952.0,
      "entropy": 0.3106093406677246,
      "epoch": 4.790064102564102,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 1507.441650390625,
      "learning_rate": 1e-06,
      "loss": 0.097,
      "num_tokens": 1794362163.0,
      "reward": 0.36152663826942444,
      "reward_std": 0.09895937889814377,
      "rewards/progression_diversity/mean": -0.0021223740186542273,
      "rewards/progression_diversity/std": 0.013677747920155525,
      "rewards/symbolic_reward_accuracy/mean": 0.2578125,
      "rewards/symbolic_reward_accuracy/std": 0.43785804510116577,
      "rewards/symbolic_reward_partial_score/mean": 0.7129719853401184,
      "rewards/symbolic_reward_partial_score/std": 0.2863211929798126,
      "rewards/tag_count_reward/mean": -0.0703125,
      "rewards/tag_count_reward/std": 0.25592297315597534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0154645442962646,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 764.0,
      "sampling/sampling_logp_difference/mean": 28.80075454711914,
      "step": 2989
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.28619883954524994,
      "epoch": 4.791666666666667,
      "grad_norm": 0.02123548462986946,
      "learning_rate": 1e-06,
      "loss": 0.1906,
      "step": 2990
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.29719385504722595,
      "epoch": 4.793269230769231,
      "grad_norm": 0.01841905526816845,
      "learning_rate": 1e-06,
      "loss": 0.1245,
      "step": 2991
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.3073057234287262,
      "epoch": 4.794871794871795,
      "grad_norm": 0.01792910508811474,
      "learning_rate": 1e-06,
      "loss": 0.0712,
      "step": 2992
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.13671875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4189.0,
      "completions/mean_length": 3995.111328125,
      "completions/mean_terminated_length": 2033.0701904296875,
      "completions/min_length": 405.0,
      "completions/min_terminated_length": 405.0,
      "entropy": 0.2574385032057762,
      "epoch": 4.796474358974359,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 2505.857666015625,
      "learning_rate": 1e-06,
      "loss": 0.197,
      "num_tokens": 1797306380.0,
      "reward": 0.26426079869270325,
      "reward_std": 0.08823113143444061,
      "rewards/progression_diversity/mean": -0.000679816585034132,
      "rewards/progression_diversity/std": 0.004317981190979481,
      "rewards/symbolic_reward_accuracy/mean": 0.15234375,
      "rewards/symbolic_reward_accuracy/std": 0.35970520973205566,
      "rewards/symbolic_reward_partial_score/mean": 0.6165690422058105,
      "rewards/symbolic_reward_partial_score/std": 0.3016470968723297,
      "rewards/tag_count_reward/mean": -0.12109375,
      "rewards/tag_count_reward/std": 0.3265552520751953,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9852645397186279,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 33.74715042114258,
      "step": 2993
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.2707083225250244,
      "epoch": 4.798076923076923,
      "grad_norm": 0.09528608620166779,
      "learning_rate": 1e-06,
      "loss": 0.13,
      "step": 2994
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4609375,
      "entropy": 0.2617756277322769,
      "epoch": 4.799679487179487,
      "grad_norm": 0.01799207553267479,
      "learning_rate": 1e-06,
      "loss": 0.1663,
      "step": 2995
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.28101493418216705,
      "epoch": 4.801282051282051,
      "grad_norm": 0.01640203595161438,
      "learning_rate": 1e-06,
      "loss": 0.153,
      "step": 2996
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3428.0,
      "completions/mean_length": 3477.884765625,
      "completions/mean_terminated_length": 2018.9324951171875,
      "completions/min_length": 173.0,
      "completions/min_terminated_length": 173.0,
      "entropy": 0.3022962361574173,
      "epoch": 4.802884615384615,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 2163.138427734375,
      "learning_rate": 1e-06,
      "loss": 0.096,
      "num_tokens": 1799923057.0,
      "reward": 0.2933644652366638,
      "reward_std": 0.09190071374177933,
      "rewards/progression_diversity/mean": -0.00046986271627247334,
      "rewards/progression_diversity/std": 0.004312288947403431,
      "rewards/symbolic_reward_accuracy/mean": 0.162109375,
      "rewards/symbolic_reward_accuracy/std": 0.3689115643501282,
      "rewards/symbolic_reward_partial_score/mean": 0.6836262941360474,
      "rewards/symbolic_reward_partial_score/std": 0.27113112807273865,
      "rewards/tag_count_reward/mean": -0.08984375,
      "rewards/tag_count_reward/std": 0.2862374484539032,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0017528533935547,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 760.0,
      "sampling/sampling_logp_difference/mean": 26.413158416748047,
      "step": 2997
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.2857065796852112,
      "epoch": 4.80448717948718,
      "grad_norm": 0.01477570179849863,
      "learning_rate": 1e-06,
      "loss": 0.2033,
      "step": 2998
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.2832331210374832,
      "epoch": 4.806089743589744,
      "grad_norm": 0.016135964542627335,
      "learning_rate": 1e-06,
      "loss": 0.9264,
      "step": 2999
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.30287647247314453,
      "epoch": 4.8076923076923075,
      "grad_norm": 3408249.25,
      "learning_rate": 1e-06,
      "loss": 1910.0106,
      "step": 3000
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.099609375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4039.0,
      "completions/mean_length": 3437.560546875,
      "completions/mean_terminated_length": 2005.30810546875,
      "completions/min_length": 241.0,
      "completions/min_terminated_length": 241.0,
      "entropy": 0.31259530782699585,
      "epoch": 4.809294871794872,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 722.9392700195312,
      "learning_rate": 1e-06,
      "loss": 0.0731,
      "num_tokens": 1802489712.0,
      "reward": 0.3412483334541321,
      "reward_std": 0.09203283488750458,
      "rewards/progression_diversity/mean": -0.00016795920964796096,
      "rewards/progression_diversity/std": 0.0027168295346200466,
      "rewards/symbolic_reward_accuracy/mean": 0.234375,
      "rewards/symbolic_reward_accuracy/std": 0.42402184009552,
      "rewards/symbolic_reward_partial_score/mean": 0.7013020515441895,
      "rewards/symbolic_reward_partial_score/std": 0.2958465814590454,
      "rewards/tag_count_reward/mean": -0.09765625,
      "rewards/tag_count_reward/std": 0.29713961482048035,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0116255283355713,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 22.6229248046875,
      "step": 3001
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.3023006319999695,
      "epoch": 4.810897435897436,
      "grad_norm": 0.01664135977625847,
      "learning_rate": 1e-06,
      "loss": 0.1274,
      "step": 3002
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.30224159359931946,
      "epoch": 4.8125,
      "grad_norm": 0.016737397760152817,
      "learning_rate": 1e-06,
      "loss": 0.0899,
      "step": 3003
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.0703125,
      "clip_ratio/low_mean": 0.1953125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.2703643888235092,
      "epoch": 4.814102564102564,
      "grad_norm": 0.019325420260429382,
      "learning_rate": 1e-06,
      "loss": 0.2522,
      "step": 3004
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.080078125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3899.0,
      "completions/mean_length": 3158.623046875,
      "completions/mean_terminated_length": 2007.3695068359375,
      "completions/min_length": 29.0,
      "completions/min_terminated_length": 29.0,
      "entropy": 0.29125888645648956,
      "epoch": 4.815705128205128,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 1334.0791015625,
      "learning_rate": 1e-06,
      "loss": 0.1156,
      "num_tokens": 1804920927.0,
      "reward": 0.4101855754852295,
      "reward_std": 0.08342595398426056,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.328125,
      "rewards/symbolic_reward_accuracy/std": 0.4699897766113281,
      "rewards/symbolic_reward_partial_score/mean": 0.7468424439430237,
      "rewards/symbolic_reward_partial_score/std": 0.2777378261089325,
      "rewards/tag_count_reward/mean": -0.107421875,
      "rewards/tag_count_reward/std": 0.30995169281959534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0229686498641968,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 15.124759674072266,
      "step": 3005
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2940514087677002,
      "epoch": 4.8173076923076925,
      "grad_norm": 0.030098246410489082,
      "learning_rate": 1e-06,
      "loss": 0.1046,
      "step": 3006
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.29679183661937714,
      "epoch": 4.818910256410256,
      "grad_norm": 0.13558229804039001,
      "learning_rate": 1e-06,
      "loss": 0.0689,
      "step": 3007
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.28803931176662445,
      "epoch": 4.82051282051282,
      "grad_norm": 0.028529301285743713,
      "learning_rate": 1e-06,
      "loss": 0.0629,
      "step": 3008
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.107421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3583.0,
      "completions/mean_length": 3604.537109375,
      "completions/mean_terminated_length": 2066.52734375,
      "completions/min_length": 344.0,
      "completions/min_terminated_length": 344.0,
      "entropy": 0.2692013829946518,
      "epoch": 4.822115384615385,
      "frac_reward_zero_std": 0.125,
      "grad_norm": 2515.8955078125,
      "learning_rate": 1e-06,
      "loss": 0.1519,
      "num_tokens": 1807674082.0,
      "reward": 0.2645507752895355,
      "reward_std": 0.10678978264331818,
      "rewards/progression_diversity/mean": 0.0,
      "rewards/progression_diversity/std": 0.0,
      "rewards/symbolic_reward_accuracy/mean": 0.138671875,
      "rewards/symbolic_reward_accuracy/std": 0.34594178199768066,
      "rewards/symbolic_reward_partial_score/mean": 0.6370442509651184,
      "rewards/symbolic_reward_partial_score/std": 0.28384286165237427,
      "rewards/tag_count_reward/mean": -0.09765625,
      "rewards/tag_count_reward/std": 0.29713961482048035,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.983115553855896,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 29.70915985107422,
      "step": 3009
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.484375,
      "entropy": 0.26541532576084137,
      "epoch": 4.823717948717949,
      "grad_norm": 0.19816499948501587,
      "learning_rate": 1e-06,
      "loss": 0.1234,
      "step": 3010
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.328125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4765625,
      "entropy": 0.2563439905643463,
      "epoch": 4.825320512820513,
      "grad_norm": 0.014676299877464771,
      "learning_rate": 1e-06,
      "loss": 0.152,
      "step": 3011
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.27064676582813263,
      "epoch": 4.826923076923077,
      "grad_norm": 0.022342152893543243,
      "learning_rate": 1e-06,
      "loss": 0.1299,
      "step": 3012
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.068359375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4520.0,
      "completions/mean_length": 3052.28125,
      "completions/mean_terminated_length": 2074.062744140625,
      "completions/min_length": 358.0,
      "completions/min_terminated_length": 358.0,
      "entropy": 0.2788449823856354,
      "epoch": 4.828525641025641,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 1405.5775146484375,
      "learning_rate": 1e-06,
      "loss": 0.07,
      "num_tokens": 1810095938.0,
      "reward": 0.3012896776199341,
      "reward_std": 0.07198358327150345,
      "rewards/progression_diversity/mean": -0.0009162867208942771,
      "rewards/progression_diversity/std": 0.006650157272815704,
      "rewards/symbolic_reward_accuracy/mean": 0.177734375,
      "rewards/symbolic_reward_accuracy/std": 0.3826628625392914,
      "rewards/symbolic_reward_partial_score/mean": 0.6703450679779053,
      "rewards/symbolic_reward_partial_score/std": 0.2553132474422455,
      "rewards/tag_count_reward/mean": -0.064453125,
      "rewards/tag_count_reward/std": 0.24579854309558868,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0105712413787842,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 764.0,
      "sampling/sampling_logp_difference/mean": 21.2810001373291,
      "step": 3013
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.40625,
      "entropy": 0.258184090256691,
      "epoch": 4.830128205128205,
      "grad_norm": 0.03618193790316582,
      "learning_rate": 1e-06,
      "loss": 0.184,
      "step": 3014
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.0546875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.2832891196012497,
      "epoch": 4.831730769230769,
      "grad_norm": 0.017958376556634903,
      "learning_rate": 1e-06,
      "loss": 0.0317,
      "step": 3015
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2775871604681015,
      "epoch": 4.833333333333333,
      "grad_norm": 0.02259541116654873,
      "learning_rate": 1e-06,
      "loss": 0.0932,
      "step": 3016
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.107421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3844.0,
      "completions/mean_length": 3652.623046875,
      "completions/mean_terminated_length": 2120.400390625,
      "completions/min_length": 610.0,
      "completions/min_terminated_length": 610.0,
      "entropy": 0.25543099641799927,
      "epoch": 4.834935897435898,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 3206.30419921875,
      "learning_rate": 1e-06,
      "loss": 0.1315,
      "num_tokens": 1812886417.0,
      "reward": 0.26881104707717896,
      "reward_std": 0.09635767340660095,
      "rewards/progression_diversity/mean": -0.0021946923807263374,
      "rewards/progression_diversity/std": 0.014525186270475388,
      "rewards/symbolic_reward_accuracy/mean": 0.1328125,
      "rewards/symbolic_reward_accuracy/std": 0.33970388770103455,
      "rewards/symbolic_reward_partial_score/mean": 0.6617349982261658,
      "rewards/symbolic_reward_partial_score/std": 0.27741891145706177,
      "rewards/tag_count_reward/mean": -0.09375,
      "rewards/tag_count_reward/std": 0.29176566004753113,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9844847321510315,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 35.239051818847656,
      "step": 3017
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2537519484758377,
      "epoch": 4.836538461538462,
      "grad_norm": 0.014091641642153263,
      "learning_rate": 1e-06,
      "loss": 0.1088,
      "step": 3018
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3046875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4375,
      "entropy": 0.2543463036417961,
      "epoch": 4.8381410256410255,
      "grad_norm": 0.02137858420610428,
      "learning_rate": 1e-06,
      "loss": 0.148,
      "step": 3019
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.25083983689546585,
      "epoch": 4.839743589743589,
      "grad_norm": 0.021724727004766464,
      "learning_rate": 1e-06,
      "loss": 0.1358,
      "step": 3020
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.111328125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3564.0,
      "completions/mean_length": 3640.005859375,
      "completions/mean_terminated_length": 2043.505615234375,
      "completions/min_length": 812.0,
      "completions/min_terminated_length": 812.0,
      "entropy": 0.25516054779291153,
      "epoch": 4.841346153846154,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 753.2295532226562,
      "learning_rate": 1e-06,
      "loss": 0.0884,
      "num_tokens": 1815684788.0,
      "reward": 0.29512912034988403,
      "reward_std": 0.08512111753225327,
      "rewards/progression_diversity/mean": -0.0027145203202962875,
      "rewards/progression_diversity/std": 0.0170114878565073,
      "rewards/symbolic_reward_accuracy/mean": 0.171875,
      "rewards/symbolic_reward_accuracy/std": 0.3776407241821289,
      "rewards/symbolic_reward_partial_score/mean": 0.6752604246139526,
      "rewards/symbolic_reward_partial_score/std": 0.29889386892318726,
      "rewards/tag_count_reward/mean": -0.10546875,
      "rewards/tag_count_reward/std": 0.3074568510055542,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.986849308013916,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 32.57376480102539,
      "step": 3021
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.23778339475393295,
      "epoch": 4.842948717948718,
      "grad_norm": 0.759289562702179,
      "learning_rate": 1e-06,
      "loss": 0.2053,
      "step": 3022
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.25857195258140564,
      "epoch": 4.844551282051282,
      "grad_norm": 0.02268078364431858,
      "learning_rate": 1e-06,
      "loss": 0.0626,
      "step": 3023
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.24314356595277786,
      "epoch": 4.846153846153846,
      "grad_norm": 0.016103658825159073,
      "learning_rate": 1e-06,
      "loss": 0.1707,
      "step": 3024
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.087890625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 4237.0,
      "completions/mean_length": 3289.69921875,
      "completions/mean_terminated_length": 2027.935791015625,
      "completions/min_length": 690.0,
      "completions/min_terminated_length": 690.0,
      "entropy": 0.27912090718746185,
      "epoch": 4.847756410256411,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 1176.0621337890625,
      "learning_rate": 1e-06,
      "loss": 0.0734,
      "num_tokens": 1818186474.0,
      "reward": 0.2887709140777588,
      "reward_std": 0.047475919127464294,
      "rewards/progression_diversity/mean": -0.0008389006252400577,
      "rewards/progression_diversity/std": 0.008314850740134716,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.6767903566360474,
      "rewards/symbolic_reward_partial_score/std": 0.2791590392589569,
      "rewards/tag_count_reward/mean": -0.080078125,
      "rewards/tag_count_reward/std": 0.271679550409317,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0138845443725586,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 768.0,
      "sampling/sampling_logp_difference/mean": 23.615236282348633,
      "step": 3025
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.28166961669921875,
      "epoch": 4.8493589743589745,
      "grad_norm": 0.01939082331955433,
      "learning_rate": 1e-06,
      "loss": 0.0351,
      "step": 3026
    },
    {
      "clip_ratio/high_max": 0.5,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.25974585115909576,
      "epoch": 4.850961538461538,
      "grad_norm": 0.029129792004823685,
      "learning_rate": 1e-06,
      "loss": 0.1673,
      "step": 3027
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.27130499482154846,
      "epoch": 4.852564102564102,
      "grad_norm": 0.09306855499744415,
      "learning_rate": 1e-06,
      "loss": 0.0714,
      "step": 3028
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.109375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3573.0,
      "completions/mean_length": 3610.380859375,
      "completions/mean_terminated_length": 2041.6907958984375,
      "completions/min_length": 297.0,
      "completions/min_terminated_length": 297.0,
      "entropy": 0.2426222562789917,
      "epoch": 4.854166666666667,
      "frac_reward_zero_std": 0.25,
      "grad_norm": 1595.66259765625,
      "learning_rate": 1e-06,
      "loss": 0.126,
      "num_tokens": 1820930301.0,
      "reward": 0.30804911255836487,
      "reward_std": 0.07695259153842926,
      "rewards/progression_diversity/mean": -0.0027075535617768764,
      "rewards/progression_diversity/std": 0.02146134339272976,
      "rewards/symbolic_reward_accuracy/mean": 0.20703125,
      "rewards/symbolic_reward_accuracy/std": 0.40557438135147095,
      "rewards/symbolic_reward_partial_score/mean": 0.6467121839523315,
      "rewards/symbolic_reward_partial_score/std": 0.3115384578704834,
      "rewards/tag_count_reward/mean": -0.1015625,
      "rewards/tag_count_reward/std": 0.30236753821372986,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9908959269523621,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 32.86491012573242,
      "step": 3029
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.24747144430875778,
      "epoch": 4.855769230769231,
      "grad_norm": 0.017448777332901955,
      "learning_rate": 1e-06,
      "loss": 0.1274,
      "step": 3030
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.26713868975639343,
      "epoch": 4.857371794871795,
      "grad_norm": 0.018266018480062485,
      "learning_rate": 1e-06,
      "loss": 0.0513,
      "step": 3031
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1796875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.23238349705934525,
      "epoch": 4.858974358974359,
      "grad_norm": 0.013434065505862236,
      "learning_rate": 1e-06,
      "loss": 0.2029,
      "step": 3032
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.10546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3829.0,
      "completions/mean_length": 3513.24609375,
      "completions/mean_terminated_length": 1995.733642578125,
      "completions/min_length": 809.0,
      "completions/min_terminated_length": 809.0,
      "entropy": 0.2464209794998169,
      "epoch": 4.860576923076923,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 1448.568115234375,
      "learning_rate": 1e-06,
      "loss": 0.1083,
      "num_tokens": 1823646795.0,
      "reward": 0.3104338049888611,
      "reward_std": 0.06594209372997284,
      "rewards/progression_diversity/mean": -0.003005102276802063,
      "rewards/progression_diversity/std": 0.022076178342103958,
      "rewards/symbolic_reward_accuracy/mean": 0.18359375,
      "rewards/symbolic_reward_accuracy/std": 0.3875311613082886,
      "rewards/symbolic_reward_partial_score/mean": 0.698291003704071,
      "rewards/symbolic_reward_partial_score/std": 0.2843267321586609,
      "rewards/tag_count_reward/mean": -0.091796875,
      "rewards/tag_count_reward/std": 0.289021372795105,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9879831075668335,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 34.731117248535156,
      "step": 3033
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.23648971319198608,
      "epoch": 4.862179487179487,
      "grad_norm": 0.027491990476846695,
      "learning_rate": 1e-06,
      "loss": 0.1345,
      "step": 3034
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23159796744585037,
      "epoch": 4.863782051282051,
      "grad_norm": 0.27813470363616943,
      "learning_rate": 1e-06,
      "loss": 0.1389,
      "step": 3035
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.23856277763843536,
      "epoch": 4.865384615384615,
      "grad_norm": 0.04474368691444397,
      "learning_rate": 1e-06,
      "loss": 0.0929,
      "step": 3036
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3863.0,
      "completions/mean_length": 3395.388671875,
      "completions/mean_terminated_length": 1927.11083984375,
      "completions/min_length": 896.0,
      "completions/min_terminated_length": 896.0,
      "entropy": 0.24479106068611145,
      "epoch": 4.86698717948718,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 1553.0164794921875,
      "learning_rate": 1e-06,
      "loss": 0.1017,
      "num_tokens": 1826229362.0,
      "reward": 0.31477978825569153,
      "reward_std": 0.07595334947109222,
      "rewards/progression_diversity/mean": -0.0034673186019062996,
      "rewards/progression_diversity/std": 0.023447884246706963,
      "rewards/symbolic_reward_accuracy/mean": 0.208984375,
      "rewards/symbolic_reward_accuracy/std": 0.40698084235191345,
      "rewards/symbolic_reward_partial_score/mean": 0.6639648079872131,
      "rewards/symbolic_reward_partial_score/std": 0.2900123596191406,
      "rewards/tag_count_reward/mean": -0.09765625,
      "rewards/tag_count_reward/std": 0.29713961482048035,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9932220578193665,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 33.091861724853516,
      "step": 3037
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.24616575241088867,
      "epoch": 4.868589743589744,
      "grad_norm": 0.10199563205242157,
      "learning_rate": 1e-06,
      "loss": 0.0912,
      "step": 3038
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.421875,
      "entropy": 0.23579637706279755,
      "epoch": 4.8701923076923075,
      "grad_norm": 0.0868377834558487,
      "learning_rate": 1e-06,
      "loss": 0.1658,
      "step": 3039
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.23798397183418274,
      "epoch": 4.871794871794872,
      "grad_norm": 0.019305624067783356,
      "learning_rate": 1e-06,
      "loss": 0.1112,
      "step": 3040
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.09765625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3809.0,
      "completions/mean_length": 3376.041015625,
      "completions/mean_terminated_length": 1968.2532958984375,
      "completions/min_length": 368.0,
      "completions/min_terminated_length": 368.0,
      "entropy": 0.2433418408036232,
      "epoch": 4.873397435897436,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 1277.455810546875,
      "learning_rate": 1e-06,
      "loss": 0.0845,
      "num_tokens": 1828798391.0,
      "reward": 0.32777589559555054,
      "reward_std": 0.09310010075569153,
      "rewards/progression_diversity/mean": -0.002196947578340769,
      "rewards/progression_diversity/std": 0.019314348697662354,
      "rewards/symbolic_reward_accuracy/mean": 0.224609375,
      "rewards/symbolic_reward_accuracy/std": 0.41773295402526855,
      "rewards/symbolic_reward_partial_score/mean": 0.675341784954071,
      "rewards/symbolic_reward_partial_score/std": 0.3014916479587555,
      "rewards/tag_count_reward/mean": -0.095703125,
      "rewards/tag_count_reward/std": 0.2944713830947876,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9986719489097595,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 29.44231605529785,
      "step": 3041
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.25057369470596313,
      "epoch": 4.875,
      "grad_norm": 0.013811684213578701,
      "learning_rate": 1e-06,
      "loss": 0.0995,
      "step": 3042
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.24305298924446106,
      "epoch": 4.876602564102564,
      "grad_norm": 0.014420563355088234,
      "learning_rate": 1e-06,
      "loss": 0.1266,
      "step": 3043
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.24530383944511414,
      "epoch": 4.878205128205128,
      "grad_norm": 0.03273453935980797,
      "learning_rate": 1e-06,
      "loss": 0.0984,
      "step": 3044
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0546875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3740.0,
      "completions/mean_length": 2730.166015625,
      "completions/mean_terminated_length": 1940.274658203125,
      "completions/min_length": 851.0,
      "completions/min_terminated_length": 851.0,
      "entropy": 0.24772102385759354,
      "epoch": 4.8798076923076925,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 1146.00244140625,
      "learning_rate": 1e-06,
      "loss": 0.1173,
      "num_tokens": 1830977612.0,
      "reward": 0.41831594705581665,
      "reward_std": 0.06758980453014374,
      "rewards/progression_diversity/mean": -0.0014180454891175032,
      "rewards/progression_diversity/std": 0.010988151654601097,
      "rewards/symbolic_reward_accuracy/mean": 0.330078125,
      "rewards/symbolic_reward_accuracy/std": 0.47070086002349854,
      "rewards/symbolic_reward_partial_score/mean": 0.7512043714523315,
      "rewards/symbolic_reward_partial_score/std": 0.2662599980831146,
      "rewards/tag_count_reward/mean": -0.05078125,
      "rewards/tag_count_reward/std": 0.21976542472839355,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0178977251052856,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 20.070125579833984,
      "step": 3045
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.2448933646082878,
      "epoch": 4.881410256410256,
      "grad_norm": 0.04779614508152008,
      "learning_rate": 1e-06,
      "loss": 0.1154,
      "step": 3046
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.26386718451976776,
      "epoch": 4.88301282051282,
      "grad_norm": 0.013240370899438858,
      "learning_rate": 1e-06,
      "loss": 0.0541,
      "step": 3047
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.25863589346408844,
      "epoch": 4.884615384615385,
      "grad_norm": 0.013798616826534271,
      "learning_rate": 1e-06,
      "loss": 0.0337,
      "step": 3048
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.091796875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3673.0,
      "completions/mean_length": 3313.994140625,
      "completions/mean_terminated_length": 1992.9398193359375,
      "completions/min_length": 276.0,
      "completions/min_terminated_length": 276.0,
      "entropy": 0.23720625787973404,
      "epoch": 4.886217948717949,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 2632.1220703125,
      "learning_rate": 1e-06,
      "loss": 0.1328,
      "num_tokens": 1833494697.0,
      "reward": 0.3145332932472229,
      "reward_std": 0.07812836021184921,
      "rewards/progression_diversity/mean": -0.0022365176118910313,
      "rewards/progression_diversity/std": 0.015494197607040405,
      "rewards/symbolic_reward_accuracy/mean": 0.205078125,
      "rewards/symbolic_reward_accuracy/std": 0.4041535556316376,
      "rewards/symbolic_reward_partial_score/mean": 0.6689616441726685,
      "rewards/symbolic_reward_partial_score/std": 0.2868025302886963,
      "rewards/tag_count_reward/mean": -0.091796875,
      "rewards/tag_count_reward/std": 0.289021372795105,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.003543734550476,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 26.09931182861328,
      "step": 3049
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.2454667165875435,
      "epoch": 4.887820512820513,
      "grad_norm": 0.01591862179338932,
      "learning_rate": 1e-06,
      "loss": 0.0864,
      "step": 3050
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.24274562299251556,
      "epoch": 4.889423076923077,
      "grad_norm": 0.19580645859241486,
      "learning_rate": 1e-06,
      "loss": 0.0845,
      "step": 3051
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.23570669442415237,
      "epoch": 4.891025641025641,
      "grad_norm": 0.03217773512005806,
      "learning_rate": 1e-06,
      "loss": 0.1089,
      "step": 3052
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.046875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3641.0,
      "completions/mean_length": 2654.248046875,
      "completions/mean_terminated_length": 1979.0142822265625,
      "completions/min_length": 307.0,
      "completions/min_terminated_length": 307.0,
      "entropy": 0.24544284492731094,
      "epoch": 4.892628205128205,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 1223.9429931640625,
      "learning_rate": 1e-06,
      "loss": 0.1096,
      "num_tokens": 1835634792.0,
      "reward": 0.47855299711227417,
      "reward_std": 0.09513897448778152,
      "rewards/progression_diversity/mean": -0.0026136531960219145,
      "rewards/progression_diversity/std": 0.020054563879966736,
      "rewards/symbolic_reward_accuracy/mean": 0.40625,
      "rewards/symbolic_reward_accuracy/std": 0.49161264300346375,
      "rewards/symbolic_reward_partial_score/mean": 0.7983887195587158,
      "rewards/symbolic_reward_partial_score/std": 0.2503470480442047,
      "rewards/tag_count_reward/mean": -0.046875,
      "rewards/tag_count_reward/std": 0.21157780289649963,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.015138864517212,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 20.58639144897461,
      "step": 3053
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.265625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.25392650067806244,
      "epoch": 4.894230769230769,
      "grad_norm": 0.018817421048879623,
      "learning_rate": 1e-06,
      "loss": 0.0569,
      "step": 3054
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3359375,
      "entropy": 0.2459840551018715,
      "epoch": 4.895833333333333,
      "grad_norm": 0.0089957220479846,
      "learning_rate": 1e-06,
      "loss": 0.0989,
      "step": 3055
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.24934467673301697,
      "epoch": 4.897435897435898,
      "grad_norm": 0.02581534907221794,
      "learning_rate": 1e-06,
      "loss": 0.0898,
      "step": 3056
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.115234375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3635.0,
      "completions/mean_length": 3645.541015625,
      "completions/mean_terminated_length": 1986.4481201171875,
      "completions/min_length": 290.0,
      "completions/min_terminated_length": 290.0,
      "entropy": 0.23428595066070557,
      "epoch": 4.899038461538462,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 1076.086669921875,
      "learning_rate": 1e-06,
      "loss": 0.1253,
      "num_tokens": 1838393661.0,
      "reward": 0.22710196673870087,
      "reward_std": 0.09757393598556519,
      "rewards/progression_diversity/mean": -0.0007409834070131183,
      "rewards/progression_diversity/std": 0.006571199279278517,
      "rewards/symbolic_reward_accuracy/mean": 0.0859375,
      "rewards/symbolic_reward_accuracy/std": 0.28054583072662354,
      "rewards/symbolic_reward_partial_score/mean": 0.6222655773162842,
      "rewards/symbolic_reward_partial_score/std": 0.27893805503845215,
      "rewards/tag_count_reward/mean": -0.111328125,
      "rewards/tag_count_reward/std": 0.31484565138816833,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9937641620635986,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 29.560943603515625,
      "step": 3057
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.2234666869044304,
      "epoch": 4.9006410256410255,
      "grad_norm": 0.025841714814305305,
      "learning_rate": 1e-06,
      "loss": 0.1654,
      "step": 3058
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.23846197873353958,
      "epoch": 4.902243589743589,
      "grad_norm": 0.01864476129412651,
      "learning_rate": 1e-06,
      "loss": 0.0935,
      "step": 3059
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.22488000988960266,
      "epoch": 4.903846153846154,
      "grad_norm": 0.20488475263118744,
      "learning_rate": 1e-06,
      "loss": 0.1767,
      "step": 3060
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.044921875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3840.0,
      "completions/mean_length": 2616.1171875,
      "completions/mean_terminated_length": 1968.5479736328125,
      "completions/min_length": 851.0,
      "completions/min_terminated_length": 851.0,
      "entropy": 0.2517363280057907,
      "epoch": 4.905448717948718,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 162.09591674804688,
      "learning_rate": 1e-06,
      "loss": 0.0271,
      "num_tokens": 1840538281.0,
      "reward": 0.3412300944328308,
      "reward_std": 0.05017785727977753,
      "rewards/progression_diversity/mean": -4.024427471449599e-05,
      "rewards/progression_diversity/std": 0.0005465570138767362,
      "rewards/symbolic_reward_accuracy/mean": 0.21484375,
      "rewards/symbolic_reward_accuracy/std": 0.4111155867576599,
      "rewards/symbolic_reward_partial_score/mean": 0.7214192152023315,
      "rewards/symbolic_reward_partial_score/std": 0.2416561096906662,
      "rewards/tag_count_reward/mean": -0.041015625,
      "rewards/tag_count_reward/std": 0.19852031767368317,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0312939882278442,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 772.0,
      "sampling/sampling_logp_difference/mean": 12.591215133666992,
      "step": 3061
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.24662530422210693,
      "epoch": 4.907051282051282,
      "grad_norm": 3917.875732421875,
      "learning_rate": 1e-06,
      "loss": 0.147,
      "step": 3062
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.2500121220946312,
      "epoch": 4.908653846153846,
      "grad_norm": 2733.346923828125,
      "learning_rate": 1e-06,
      "loss": 0.2713,
      "step": 3063
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.23512417078018188,
      "epoch": 4.910256410256411,
      "grad_norm": 0.14125868678092957,
      "learning_rate": 1e-06,
      "loss": 0.1516,
      "step": 3064
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.095703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3573.0,
      "completions/mean_length": 3311.6796875,
      "completions/mean_terminated_length": 1928.2159423828125,
      "completions/min_length": 385.0,
      "completions/min_terminated_length": 385.0,
      "entropy": 0.2293601930141449,
      "epoch": 4.9118589743589745,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 1017.6434326171875,
      "learning_rate": 1e-06,
      "loss": 0.0998,
      "num_tokens": 1843132549.0,
      "reward": 0.29114842414855957,
      "reward_std": 0.08108247071504593,
      "rewards/progression_diversity/mean": -0.00039431118057109416,
      "rewards/progression_diversity/std": 0.004132647532969713,
      "rewards/symbolic_reward_accuracy/mean": 0.1640625,
      "rewards/symbolic_reward_accuracy/std": 0.37069445848464966,
      "rewards/symbolic_reward_partial_score/mean": 0.6723307371139526,
      "rewards/symbolic_reward_partial_score/std": 0.2803674042224884,
      "rewards/tag_count_reward/mean": -0.08984375,
      "rewards/tag_count_reward/std": 0.2862374484539032,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.004433274269104,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 24.679431915283203,
      "step": 3065
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3515625,
      "entropy": 0.22065846621990204,
      "epoch": 4.913461538461538,
      "grad_norm": 0.017783688381314278,
      "learning_rate": 1e-06,
      "loss": 0.1363,
      "step": 3066
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.23686835914850235,
      "epoch": 4.915064102564102,
      "grad_norm": 0.03829358518123627,
      "learning_rate": 1e-06,
      "loss": 0.0933,
      "step": 3067
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.125,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.22281410545110703,
      "epoch": 4.916666666666667,
      "grad_norm": 0.013630473986268044,
      "learning_rate": 1e-06,
      "loss": 0.1489,
      "step": 3068
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0859375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3399.0,
      "completions/mean_length": 3175.4453125,
      "completions/mean_terminated_length": 1933.615478515625,
      "completions/min_length": 307.0,
      "completions/min_terminated_length": 307.0,
      "entropy": 0.23225311189889908,
      "epoch": 4.918269230769231,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 1236.8885498046875,
      "learning_rate": 1e-06,
      "loss": 0.0914,
      "num_tokens": 1845645129.0,
      "reward": 0.29391324520111084,
      "reward_std": 0.04953677952289581,
      "rewards/progression_diversity/mean": -0.0002773654996417463,
      "rewards/progression_diversity/std": 0.003748962190002203,
      "rewards/symbolic_reward_accuracy/mean": 0.15625,
      "rewards/symbolic_reward_accuracy/std": 0.36344730854034424,
      "rewards/symbolic_reward_partial_score/mean": 0.6932617425918579,
      "rewards/symbolic_reward_partial_score/std": 0.2669588029384613,
      "rewards/tag_count_reward/mean": -0.078125,
      "rewards/tag_count_reward/std": 0.26863065361976624,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.011780023574829,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 22.082910537719727,
      "step": 3069
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.234375,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.22503124177455902,
      "epoch": 4.919871794871795,
      "grad_norm": 0.12676434218883514,
      "learning_rate": 1e-06,
      "loss": 0.1061,
      "step": 3070
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.22606287896633148,
      "epoch": 4.921474358974359,
      "grad_norm": 0.02233606018126011,
      "learning_rate": 1e-06,
      "loss": 0.106,
      "step": 3071
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2343372032046318,
      "epoch": 4.923076923076923,
      "grad_norm": 0.014616597443819046,
      "learning_rate": 1e-06,
      "loss": 0.1043,
      "step": 3072
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.087890625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3779.0,
      "completions/mean_length": 3257.8828125,
      "completions/mean_terminated_length": 1993.053466796875,
      "completions/min_length": 304.0,
      "completions/min_terminated_length": 304.0,
      "entropy": 0.21944886445999146,
      "epoch": 4.924679487179487,
      "frac_reward_zero_std": 0.15625,
      "grad_norm": 1409.2808837890625,
      "learning_rate": 1e-06,
      "loss": 0.1516,
      "num_tokens": 1848203293.0,
      "reward": 0.21257862448692322,
      "reward_std": 0.07666586339473724,
      "rewards/progression_diversity/mean": -0.00044064479880034924,
      "rewards/progression_diversity/std": 0.003953089937567711,
      "rewards/symbolic_reward_accuracy/mean": 0.06640625,
      "rewards/symbolic_reward_accuracy/std": 0.2492343932390213,
      "rewards/symbolic_reward_partial_score/mean": 0.6037923097610474,
      "rewards/symbolic_reward_partial_score/std": 0.2407861053943634,
      "rewards/tag_count_reward/mean": -0.083984375,
      "rewards/tag_count_reward/std": 0.2776356339454651,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9995574355125427,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 28.18425941467285,
      "step": 3073
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.28125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2246202975511551,
      "epoch": 4.926282051282051,
      "grad_norm": 0.31535470485687256,
      "learning_rate": 1e-06,
      "loss": 0.1254,
      "step": 3074
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.3125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.390625,
      "entropy": 0.23829004913568497,
      "epoch": 4.927884615384615,
      "grad_norm": 0.02114877477288246,
      "learning_rate": 1e-06,
      "loss": 0.0652,
      "step": 3075
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.25,
      "clip_ratio/low_mean": 0.1484375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3984375,
      "entropy": 0.22181038558483124,
      "epoch": 4.92948717948718,
      "grad_norm": 0.021001050248742104,
      "learning_rate": 1e-06,
      "loss": 0.1621,
      "step": 3076
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.052734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3972.0,
      "completions/mean_length": 2757.482421875,
      "completions/mean_terminated_length": 1998.892822265625,
      "completions/min_length": 963.0,
      "completions/min_terminated_length": 963.0,
      "entropy": 0.2369692623615265,
      "epoch": 4.931089743589744,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 1576.586181640625,
      "learning_rate": 1e-06,
      "loss": 0.1226,
      "num_tokens": 1850362948.0,
      "reward": 0.3155801594257355,
      "reward_std": 0.0678568109869957,
      "rewards/progression_diversity/mean": -9.189260163111612e-05,
      "rewards/progression_diversity/std": 0.0013428285019472241,
      "rewards/symbolic_reward_accuracy/mean": 0.177734375,
      "rewards/symbolic_reward_accuracy/std": 0.3826628625392914,
      "rewards/symbolic_reward_partial_score/mean": 0.7120931148529053,
      "rewards/symbolic_reward_partial_score/std": 0.23257166147232056,
      "rewards/tag_count_reward/mean": -0.046875,
      "rewards/tag_count_reward/std": 0.21157780289649963,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.022636890411377,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 18.015220642089844,
      "step": 3077
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.234375,
      "entropy": 0.2449949011206627,
      "epoch": 4.9326923076923075,
      "grad_norm": 986.05517578125,
      "learning_rate": 1e-06,
      "loss": 0.1112,
      "step": 3078
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.2491198480129242,
      "epoch": 4.934294871794872,
      "grad_norm": 0.01813652738928795,
      "learning_rate": 1e-06,
      "loss": 0.0695,
      "step": 3079
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.25,
      "entropy": 0.24884822964668274,
      "epoch": 4.935897435897436,
      "grad_norm": 0.030230067670345306,
      "learning_rate": 1e-06,
      "loss": 0.0596,
      "step": 3080
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.07421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3653.0,
      "completions/mean_length": 3115.162109375,
      "completions/mean_terminated_length": 2051.41552734375,
      "completions/min_length": 859.0,
      "completions/min_terminated_length": 859.0,
      "entropy": 0.2273581624031067,
      "epoch": 4.9375,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 1620.06494140625,
      "learning_rate": 1e-06,
      "loss": 0.1231,
      "num_tokens": 1852808343.0,
      "reward": 0.3891579508781433,
      "reward_std": 0.09541543573141098,
      "rewards/progression_diversity/mean": -0.0007119444198906422,
      "rewards/progression_diversity/std": 0.004969864152371883,
      "rewards/symbolic_reward_accuracy/mean": 0.298828125,
      "rewards/symbolic_reward_accuracy/std": 0.45819199085235596,
      "rewards/symbolic_reward_partial_score/mean": 0.7229980230331421,
      "rewards/symbolic_reward_partial_score/std": 0.27403724193573,
      "rewards/tag_count_reward/mean": -0.0703125,
      "rewards/tag_count_reward/std": 0.25592297315597534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0014233589172363,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 27.424591064453125,
      "step": 3081
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.22573982179164886,
      "epoch": 4.939102564102564,
      "grad_norm": 0.013195838779211044,
      "learning_rate": 1e-06,
      "loss": 0.1688,
      "step": 3082
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.22488176822662354,
      "epoch": 4.940705128205128,
      "grad_norm": 0.01642797514796257,
      "learning_rate": 1e-06,
      "loss": 0.1161,
      "step": 3083
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.22580020129680634,
      "epoch": 4.9423076923076925,
      "grad_norm": 0.030097253620624542,
      "learning_rate": 1e-06,
      "loss": 0.1038,
      "step": 3084
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.064453125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3909.0,
      "completions/mean_length": 2938.462890625,
      "completions/mean_terminated_length": 2012.1524658203125,
      "completions/min_length": 1015.0,
      "completions/min_terminated_length": 1015.0,
      "entropy": 0.24025756865739822,
      "epoch": 4.943910256410256,
      "frac_reward_zero_std": 0.3125,
      "grad_norm": 427.4413757324219,
      "learning_rate": 1e-06,
      "loss": 0.0368,
      "num_tokens": 1855159284.0,
      "reward": 0.3869500756263733,
      "reward_std": 0.09471500664949417,
      "rewards/progression_diversity/mean": -0.0007927106926217675,
      "rewards/progression_diversity/std": 0.0072387754917144775,
      "rewards/symbolic_reward_accuracy/mean": 0.28515625,
      "rewards/symbolic_reward_accuracy/std": 0.45193037390708923,
      "rewards/symbolic_reward_partial_score/mean": 0.7403808832168579,
      "rewards/symbolic_reward_partial_score/std": 0.2626368999481201,
      "rewards/tag_count_reward/mean": -0.0625,
      "rewards/tag_count_reward/std": 0.2422981858253479,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.011838674545288,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 22.941444396972656,
      "step": 3085
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.1015625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.2293272241950035,
      "epoch": 4.94551282051282,
      "grad_norm": 0.014672359451651573,
      "learning_rate": 1e-06,
      "loss": 0.0897,
      "step": 3086
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.23675543069839478,
      "epoch": 4.947115384615385,
      "grad_norm": 0.03274073079228401,
      "learning_rate": 1e-06,
      "loss": 2.9293,
      "step": 3087
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.109375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.28125,
      "entropy": 0.21905168890953064,
      "epoch": 4.948717948717949,
      "grad_norm": 0.013550637289881706,
      "learning_rate": 1e-06,
      "loss": 0.181,
      "step": 3088
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.044921875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3887.0,
      "completions/mean_length": 2637.939453125,
      "completions/mean_terminated_length": 1991.396728515625,
      "completions/min_length": 296.0,
      "completions/min_terminated_length": 296.0,
      "entropy": 0.23665142059326172,
      "epoch": 4.950320512820513,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 1990.0498046875,
      "learning_rate": 1e-06,
      "loss": 0.1275,
      "num_tokens": 1857301445.0,
      "reward": 0.35283172130584717,
      "reward_std": 0.07654442638158798,
      "rewards/progression_diversity/mean": -0.0010098177008330822,
      "rewards/progression_diversity/std": 0.007632537744939327,
      "rewards/symbolic_reward_accuracy/mean": 0.23828125,
      "rewards/symbolic_reward_accuracy/std": 0.42644867300987244,
      "rewards/symbolic_reward_partial_score/mean": 0.7145507335662842,
      "rewards/symbolic_reward_partial_score/std": 0.2468079924583435,
      "rewards/tag_count_reward/mean": -0.044921875,
      "rewards/tag_count_reward/std": 0.20733514428138733,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0146788358688354,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 22.055877685546875,
      "step": 3089
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2578125,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.2322361320257187,
      "epoch": 4.951923076923077,
      "grad_norm": 0.017291007563471794,
      "learning_rate": 1e-06,
      "loss": 0.1592,
      "step": 3090
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.24952280521392822,
      "epoch": 4.953525641025641,
      "grad_norm": 0.010401350446045399,
      "learning_rate": 1e-06,
      "loss": 0.0266,
      "step": 3091
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.2399914711713791,
      "epoch": 4.955128205128205,
      "grad_norm": 0.35759103298187256,
      "learning_rate": 1e-06,
      "loss": 0.0707,
      "step": 3092
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.076171875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3963.0,
      "completions/mean_length": 3135.34375,
      "completions/mean_terminated_length": 2042.9598388671875,
      "completions/min_length": 554.0,
      "completions/min_terminated_length": 554.0,
      "entropy": 0.2194850966334343,
      "epoch": 4.956730769230769,
      "frac_reward_zero_std": 0.1875,
      "grad_norm": 1517.0322265625,
      "learning_rate": 1e-06,
      "loss": 0.2334,
      "num_tokens": 1859834453.0,
      "reward": 0.2989712357521057,
      "reward_std": 0.10124389827251434,
      "rewards/progression_diversity/mean": -0.0013144873082637787,
      "rewards/progression_diversity/std": 0.011923530139029026,
      "rewards/symbolic_reward_accuracy/mean": 0.185546875,
      "rewards/symbolic_reward_accuracy/std": 0.38912075757980347,
      "rewards/symbolic_reward_partial_score/mean": 0.6483073234558105,
      "rewards/symbolic_reward_partial_score/std": 0.26260486245155334,
      "rewards/tag_count_reward/mean": -0.068359375,
      "rewards/tag_count_reward/std": 0.25260838866233826,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9993857145309448,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 29.454757690429688,
      "step": 3093
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2578125,
      "entropy": 0.23910460621118546,
      "epoch": 4.958333333333333,
      "grad_norm": 0.11438003182411194,
      "learning_rate": 1e-06,
      "loss": 0.0364,
      "step": 3094
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.296875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.4140625,
      "entropy": 0.2314603254199028,
      "epoch": 4.959935897435898,
      "grad_norm": 0.013372881338000298,
      "learning_rate": 1e-06,
      "loss": 0.1093,
      "step": 3095
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.22888445109128952,
      "epoch": 4.961538461538462,
      "grad_norm": 0.04182210564613342,
      "learning_rate": 1e-06,
      "loss": 0.1042,
      "step": 3096
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1015625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3800.0,
      "completions/mean_length": 3504.990234375,
      "completions/mean_terminated_length": 2049.10205078125,
      "completions/min_length": 499.0,
      "completions/min_terminated_length": 499.0,
      "entropy": 0.21291837096214294,
      "epoch": 4.9631410256410255,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 1515.2686767578125,
      "learning_rate": 1e-06,
      "loss": 0.1687,
      "num_tokens": 1862539136.0,
      "reward": 0.38587409257888794,
      "reward_std": 0.12614938616752625,
      "rewards/progression_diversity/mean": -0.0014604174066334963,
      "rewards/progression_diversity/std": 0.010783915407955647,
      "rewards/symbolic_reward_accuracy/mean": 0.310546875,
      "rewards/symbolic_reward_accuracy/std": 0.46317005157470703,
      "rewards/symbolic_reward_partial_score/mean": 0.6984049677848816,
      "rewards/symbolic_reward_partial_score/std": 0.3126681447029114,
      "rewards/tag_count_reward/mean": -0.099609375,
      "rewards/tag_count_reward/std": 0.29977133870124817,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 0.9896366000175476,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 34.82575607299805,
      "step": 3097
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2890625,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.23361308127641678,
      "epoch": 4.964743589743589,
      "grad_norm": 99630.9453125,
      "learning_rate": 1e-06,
      "loss": 1.774,
      "step": 3098
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.15625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.375,
      "entropy": 0.21884525567293167,
      "epoch": 4.966346153846154,
      "grad_norm": 0.037266023457050323,
      "learning_rate": 1e-06,
      "loss": 0.1422,
      "step": 3099
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.203125,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.21776709705591202,
      "epoch": 4.967948717948718,
      "grad_norm": 0.022990232333540916,
      "learning_rate": 1e-06,
      "loss": 0.1704,
      "step": 3100
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.07421875,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3794.0,
      "completions/mean_length": 3065.734375,
      "completions/mean_terminated_length": 1998.0252685546875,
      "completions/min_length": 313.0,
      "completions/min_terminated_length": 313.0,
      "entropy": 0.2290486916899681,
      "epoch": 4.969551282051282,
      "frac_reward_zero_std": 0.21875,
      "grad_norm": 1406.708984375,
      "learning_rate": 1e-06,
      "loss": 0.0795,
      "num_tokens": 1864989176.0,
      "reward": 0.27424493432044983,
      "reward_std": 0.08562411367893219,
      "rewards/progression_diversity/mean": -0.0012889985227957368,
      "rewards/progression_diversity/std": 0.011297719553112984,
      "rewards/symbolic_reward_accuracy/mean": 0.134765625,
      "rewards/symbolic_reward_accuracy/std": 0.3418070077896118,
      "rewards/symbolic_reward_partial_score/mean": 0.6680989265441895,
      "rewards/symbolic_reward_partial_score/std": 0.251154363155365,
      "rewards/tag_count_reward/mean": -0.0703125,
      "rewards/tag_count_reward/std": 0.25592297315597534,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.004550576210022,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 27.251415252685547,
      "step": 3101
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.171875,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3125,
      "entropy": 0.21562334895133972,
      "epoch": 4.971153846153846,
      "grad_norm": 0.22579920291900635,
      "learning_rate": 1e-06,
      "loss": 0.1258,
      "step": 3102
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.22457308322191238,
      "epoch": 4.972756410256411,
      "grad_norm": 0.023017656058073044,
      "learning_rate": 1e-06,
      "loss": 423466112.0,
      "step": 3103
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.15625,
      "clip_ratio/low_mean": 0.140625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.296875,
      "entropy": 0.2277710661292076,
      "epoch": 4.9743589743589745,
      "grad_norm": 0.0158623605966568,
      "learning_rate": 1e-06,
      "loss": 0.088,
      "step": 3104
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0703125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3392.0,
      "completions/mean_length": 2962.814453125,
      "completions/mean_terminated_length": 1947.766845703125,
      "completions/min_length": 239.0,
      "completions/min_terminated_length": 239.0,
      "entropy": 0.22295588999986649,
      "epoch": 4.975961538461538,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 615.591796875,
      "learning_rate": 1e-06,
      "loss": 0.1338,
      "num_tokens": 1867381977.0,
      "reward": 0.31032562255859375,
      "reward_std": 0.08792345225811005,
      "rewards/progression_diversity/mean": -0.001620155293494463,
      "rewards/progression_diversity/std": 0.014947640709578991,
      "rewards/symbolic_reward_accuracy/mean": 0.16796875,
      "rewards/symbolic_reward_accuracy/std": 0.374204158782959,
      "rewards/symbolic_reward_partial_score/mean": 0.7187174558639526,
      "rewards/symbolic_reward_partial_score/std": 0.2543519139289856,
      "rewards/tag_count_reward/mean": -0.060546875,
      "rewards/tag_count_reward/std": 0.2387305200099945,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0067986249923706,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 26.045305252075195,
      "step": 3105
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2734375,
      "entropy": 0.2358352467417717,
      "epoch": 4.977564102564102,
      "grad_norm": 0.02432454563677311,
      "learning_rate": 1e-06,
      "loss": 0.0836,
      "step": 3106
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1796875,
      "clip_ratio/low_mean": 0.125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3046875,
      "entropy": 0.22167550772428513,
      "epoch": 4.979166666666667,
      "grad_norm": 0.01078137755393982,
      "learning_rate": 1e-06,
      "loss": 0.1256,
      "step": 3107
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.34375,
      "entropy": 0.22863591462373734,
      "epoch": 4.980769230769231,
      "grad_norm": 0.019278274849057198,
      "learning_rate": 1e-06,
      "loss": 0.0846,
      "step": 3108
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.048828125,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3294.0,
      "completions/mean_length": 2566.49609375,
      "completions/mean_terminated_length": 1857.1787109375,
      "completions/min_length": 900.0,
      "completions/min_terminated_length": 900.0,
      "entropy": 0.25486068427562714,
      "epoch": 4.982371794871795,
      "frac_reward_zero_std": 0.34375,
      "grad_norm": 590.8904418945312,
      "learning_rate": 1e-06,
      "loss": 0.0226,
      "num_tokens": 1869483767.0,
      "reward": 0.3811664581298828,
      "reward_std": 0.06205814331769943,
      "rewards/progression_diversity/mean": -0.0015223543159663677,
      "rewards/progression_diversity/std": 0.01551423966884613,
      "rewards/symbolic_reward_accuracy/mean": 0.267578125,
      "rewards/symbolic_reward_accuracy/std": 0.4431293308734894,
      "rewards/symbolic_reward_partial_score/mean": 0.7497721314430237,
      "rewards/symbolic_reward_partial_score/std": 0.24262651801109314,
      "rewards/tag_count_reward/mean": -0.04296875,
      "rewards/tag_count_reward/std": 0.2029850035905838,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0239535570144653,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 18.67233657836914,
      "step": 3109
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.2109375,
      "clip_ratio/low_mean": 0.078125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2890625,
      "entropy": 0.23615090548992157,
      "epoch": 4.983974358974359,
      "grad_norm": 37597.58984375,
      "learning_rate": 1e-06,
      "loss": 0.8181,
      "step": 3110
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.1328125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.328125,
      "entropy": 0.24501872807741165,
      "epoch": 4.985576923076923,
      "grad_norm": 0.01006036065518856,
      "learning_rate": 1e-06,
      "loss": 0.0984,
      "step": 3111
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2734375,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3671875,
      "entropy": 0.24807747453451157,
      "epoch": 4.987179487179487,
      "grad_norm": 0.02072441577911377,
      "learning_rate": 1e-06,
      "loss": 0.0622,
      "step": 3112
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.052734375,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3321.0,
      "completions/mean_length": 2686.251953125,
      "completions/mean_terminated_length": 1923.697021484375,
      "completions/min_length": 427.0,
      "completions/min_terminated_length": 427.0,
      "entropy": 0.234246626496315,
      "epoch": 4.988782051282051,
      "frac_reward_zero_std": 0.28125,
      "grad_norm": 554.318603515625,
      "learning_rate": 1e-06,
      "loss": 0.0841,
      "num_tokens": 1871779112.0,
      "reward": 0.3497743010520935,
      "reward_std": 0.06554560363292694,
      "rewards/progression_diversity/mean": -0.0010880029294639826,
      "rewards/progression_diversity/std": 0.008479480631649494,
      "rewards/symbolic_reward_accuracy/mean": 0.236328125,
      "rewards/symbolic_reward_accuracy/std": 0.42524150013923645,
      "rewards/symbolic_reward_partial_score/mean": 0.7102213501930237,
      "rewards/symbolic_reward_partial_score/std": 0.2495347112417221,
      "rewards/tag_count_reward/mean": -0.05078125,
      "rewards/tag_count_reward/std": 0.21976542472839355,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0156009197235107,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 21.600927352905273,
      "step": 3113
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2265625,
      "clip_ratio/low_mean": 0.09375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3203125,
      "entropy": 0.2407403290271759,
      "epoch": 4.990384615384615,
      "grad_norm": 0.019321072846651077,
      "learning_rate": 1e-06,
      "loss": 0.0487,
      "step": 3114
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.2421875,
      "clip_ratio/low_mean": 0.1171875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.359375,
      "entropy": 0.23200398683547974,
      "epoch": 4.99198717948718,
      "grad_norm": 0.015974221751093864,
      "learning_rate": 1e-06,
      "loss": 0.1028,
      "step": 3115
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.21875,
      "clip_ratio/low_mean": 0.1640625,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.3828125,
      "entropy": 0.23310399055480957,
      "epoch": 4.993589743589744,
      "grad_norm": 0.016877546906471252,
      "learning_rate": 1e-06,
      "loss": 0.1081,
      "step": 3116
    },
    {
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03515625,
      "completions/max_length": 16384.0,
      "completions/max_terminated_length": 3118.0,
      "completions/mean_length": 2393.515625,
      "completions/mean_terminated_length": 1883.740966796875,
      "completions/min_length": 592.0,
      "completions/min_terminated_length": 592.0,
      "entropy": 0.24390202015638351,
      "epoch": 4.9951923076923075,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 1346.7171630859375,
      "learning_rate": 1e-06,
      "loss": 0.0759,
      "num_tokens": 1873829408.0,
      "reward": 0.3640533685684204,
      "reward_std": 0.027338774874806404,
      "rewards/progression_diversity/mean": -0.0004282901354599744,
      "rewards/progression_diversity/std": 0.0046591991558671,
      "rewards/symbolic_reward_accuracy/mean": 0.25,
      "rewards/symbolic_reward_accuracy/std": 0.43343618512153625,
      "rewards/symbolic_reward_partial_score/mean": 0.7239420413970947,
      "rewards/symbolic_reward_partial_score/std": 0.23576416075229645,
      "rewards/tag_count_reward/mean": -0.03125,
      "rewards/tag_count_reward/std": 0.17416280508041382,
      "sampling/importance_sampling_ratio/max": 2.0,
      "sampling/importance_sampling_ratio/mean": 1.0342185497283936,
      "sampling/importance_sampling_ratio/min": 0.0,
      "sampling/sampling_logp_difference/max": 776.0,
      "sampling/sampling_logp_difference/mean": 13.47022533416748,
      "step": 3117
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1953125,
      "clip_ratio/low_mean": 0.0703125,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.265625,
      "entropy": 0.24693401902914047,
      "epoch": 4.996794871794872,
      "grad_norm": 2261.0185546875,
      "learning_rate": 1e-06,
      "loss": 0.0913,
      "step": 3118
    },
    {
      "clip_ratio/high_max": 0.75,
      "clip_ratio/high_mean": 0.140625,
      "clip_ratio/low_mean": 0.0859375,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2265625,
      "entropy": 0.24139049649238586,
      "epoch": 4.998397435897436,
      "grad_norm": 0.024692002683877945,
      "learning_rate": 1e-06,
      "loss": 0.0602,
      "step": 3119
    },
    {
      "clip_ratio/high_max": 1.0,
      "clip_ratio/high_mean": 0.1640625,
      "clip_ratio/low_mean": 0.046875,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.2109375,
      "entropy": 0.2525310665369034,
      "epoch": 5.0,
      "grad_norm": 0.013732725754380226,
      "learning_rate": 1e-06,
      "loss": 0.0156,
      "step": 3120
    },
    {
      "epoch": 5.0,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.063720703125,
      "eval_completions/max_length": 16002.59375,
      "eval_completions/max_terminated_length": 3275.65625,
      "eval_completions/mean_length": 2817.200927734375,
      "eval_completions/mean_terminated_length": 1894.647304534912,
      "eval_completions/min_length": 834.84375,
      "eval_completions/min_terminated_length": 834.84375,
      "eval_entropy": 0.22543939435854554,
      "eval_frac_reward_zero_std": 0.28125,
      "eval_loss": 0.04206552356481552,
      "eval_num_tokens": 1873829408.0,
      "eval_reward": 0.24393098056316376,
      "eval_reward_std": 0.05101948481751606,
      "eval_rewards/progression_diversity/mean": -0.001373334529262138,
      "eval_rewards/progression_diversity/std": 0.01101221157978216,
      "eval_rewards/symbolic_reward_accuracy/mean": 0.0966796875,
      "eval_rewards/symbolic_reward_accuracy/std": 0.21552498079836369,
      "eval_rewards/symbolic_reward_partial_score/mean": 0.6393208876252174,
      "eval_rewards/symbolic_reward_partial_score/std": 0.22751979576423764,
      "eval_rewards/tag_count_reward/mean": -0.05859375,
      "eval_rewards/tag_count_reward/std": 0.21680113021284342,
      "eval_runtime": 4356.8719,
      "eval_samples_per_second": 0.057,
      "eval_sampling/importance_sampling_ratio/max": 2.0,
      "eval_sampling/importance_sampling_ratio/mean": 1.0280762165784836,
      "eval_sampling/importance_sampling_ratio/min": 1.5225041253212082e-39,
      "eval_sampling/sampling_logp_difference/max": 754.4348351955414,
      "eval_sampling/sampling_logp_difference/mean": 16.028895314550027,
      "eval_steps_per_second": 0.0,
      "step": 3120
    },
    {
      "epoch": 5.0,
      "step": 3120,
      "total_flos": 0.0,
      "train_loss": 235461599.143655,
      "train_runtime": 175420.2215,
      "train_samples_per_second": 0.143,
      "train_steps_per_second": 0.018
    }
  ],
  "logging_steps": 1,
  "max_steps": 3120,
  "num_input_tokens_seen": 1873829408,
  "num_train_epochs": 5,
  "save_steps": 500,
  "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": 2,
  "trial_name": null,
  "trial_params": null
}