{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.21333333333333335,
  "eval_steps": 500,
  "global_step": 200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "aux_brier/lambda": 0.1,
      "aux_brier/loss": 5.791089203391117e-07,
      "aux_brier/mean_group_std": 0.06289231620091193,
      "aux_brier/mean_r": 0.4665906001184907,
      "aux_brier/n_active_tok": 24.615384615384617,
      "aux_brier/n_groups": 5.3076923076923075,
      "aux_brier/n_step_records": 6.153846153846154,
      "calib/answer_extract_rate": 0.08203125,
      "calib/auroc": 0.6944444444444445,
      "calib/avg_num_step_conf": 0.3359375,
      "calib/ece": 0.6230769230769231,
      "calib/final_conf_rate": 0.05078125,
      "calib/format_rate": 0.04296875,
      "calib/frac_conf_gt_0.9": 0.7692307692307693,
      "calib/gap": 0.03861111111111115,
      "calib/mean_conf": 0.9307692307692309,
      "calib/mu_c": 0.9575,
      "calib/mu_w": 0.9188888888888889,
      "calib/nonempty_final_conf_rate": 0.05078125,
      "calib/nonempty_reasoning_rate": 0.09765625,
      "calib/nonempty_step_conf_rate": 0.0703125,
      "calib/pce": 0.6230769230769231,
      "calib/std_conf": 0.07965903671384378,
      "calib/step_conf_rate": 0.0703125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08984375,
      "completions/max_length": 2955.0,
      "completions/max_terminated_length": 2955.0,
      "completions/mean_length": 613.67578125,
      "completions/mean_terminated_length": 674.2532348632812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0010666666666666667,
      "grad_norm": 0.3237912356853485,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0332,
      "num_tokens": 264685.0,
      "reward": 0.03929531201720238,
      "reward_std": 0.08434611558914185,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.01655624993145466,
      "rewards/format_reward_step_strict": 0.0390625,
      "step": 1
    },
    {
      "aux_brier/lambda": 0.10000000000000003,
      "aux_brier/loss": 2.461345396504181e-08,
      "aux_brier/mean_group_std": 0.046398653263787254,
      "aux_brier/mean_r": 0.430243897442093,
      "aux_brier/n_active_tok": 28.42105263157895,
      "aux_brier/n_groups": 5.894736842105263,
      "aux_brier/n_step_records": 7.105263157894737,
      "calib/answer_extract_rate": 0.13671875,
      "calib/auroc": 0.5338345864661654,
      "calib/avg_num_step_conf": 0.55078125,
      "calib/ece": 0.6261538461538463,
      "calib/final_conf_rate": 0.1015625,
      "calib/format_rate": 0.08984375,
      "calib/frac_conf_gt_0.9": 0.7692307692307693,
      "calib/gap": 0.002406015037593856,
      "calib/mean_conf": 0.8953846153846153,
      "calib/mu_c": 0.897142857142857,
      "calib/mu_w": 0.8947368421052632,
      "calib/nonempty_final_conf_rate": 0.1015625,
      "calib/nonempty_reasoning_rate": 0.14453125,
      "calib/nonempty_step_conf_rate": 0.109375,
      "calib/pce": 0.6261538461538463,
      "calib/std_conf": 0.18653172073466937,
      "calib/step_conf_rate": 0.109375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3001.0,
      "completions/max_terminated_length": 3001.0,
      "completions/mean_length": 646.4609375,
      "completions/mean_terminated_length": 683.8594970703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0021333333333333334,
      "grad_norm": 0.006301195360720158,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0643,
      "num_tokens": 533467.0,
      "reward": 0.08358447253704071,
      "reward_std": 0.15892045199871063,
      "rewards/accuracy_reward_step": 0.03125,
      "rewards/final_brier_reward_step": 0.02965039201080799,
      "rewards/format_reward_step_strict": 0.08984375,
      "step": 2
    },
    {
      "aux_brier/lambda": 0.09999999999999999,
      "aux_brier/loss": -5.62932740422184e-09,
      "aux_brier/mean_group_std": 0.04158255792864185,
      "aux_brier/mean_r": 0.3927569710388179,
      "aux_brier/n_active_tok": 24.0,
      "aux_brier/n_groups": 5.0,
      "aux_brier/n_step_records": 6.0,
      "calib/answer_extract_rate": 0.05859375,
      "calib/auroc": 0.5208333333333333,
      "calib/avg_num_step_conf": 0.2421875,
      "calib/ece": 0.6027272727272728,
      "calib/final_conf_rate": 0.04296875,
      "calib/format_rate": 0.0390625,
      "calib/frac_conf_gt_0.9": 0.9090909090909091,
      "calib/gap": 0.12083333333333324,
      "calib/mean_conf": 0.8754545454545454,
      "calib/mu_c": 0.9633333333333333,
      "calib/mu_w": 0.8425,
      "calib/nonempty_final_conf_rate": 0.04296875,
      "calib/nonempty_reasoning_rate": 0.0625,
      "calib/nonempty_step_conf_rate": 0.046875,
      "calib/pce": 0.6027272727272728,
      "calib/std_conf": 0.2744039370365466,
      "calib/step_conf_rate": 0.046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3047.0,
      "completions/max_terminated_length": 3047.0,
      "completions/mean_length": 696.83203125,
      "completions/mean_terminated_length": 752.6961669921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0032,
      "grad_norm": 0.003430852433666587,
      "learning_rate": 7.5e-07,
      "loss": 0.0347,
      "num_tokens": 817112.0,
      "reward": 0.033635444939136505,
      "reward_std": 0.07786141335964203,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.017354296520352364,
      "rewards/format_reward_step_strict": 0.03515625,
      "step": 3
    },
    {
      "aux_brier/lambda": 0.1,
      "aux_brier/loss": -1.676789982077218e-08,
      "aux_brier/mean_group_std": 0.0222393300341596,
      "aux_brier/mean_r": 0.5159534199701812,
      "aux_brier/n_active_tok": 24.615384615384617,
      "aux_brier/n_groups": 5.538461538461538,
      "aux_brier/n_step_records": 6.153846153846154,
      "calib/answer_extract_rate": 0.0625,
      "calib/auroc": 0.475,
      "calib/avg_num_step_conf": 0.31640625,
      "calib/ece": 0.7883333333333334,
      "calib/final_conf_rate": 0.046875,
      "calib/format_rate": 0.04296875,
      "calib/frac_conf_gt_0.9": 0.8333333333333334,
      "calib/gap": -0.03599999999999992,
      "calib/mean_conf": 0.9199999999999999,
      "calib/mu_c": 0.89,
      "calib/mu_w": 0.9259999999999999,
      "calib/nonempty_final_conf_rate": 0.046875,
      "calib/nonempty_reasoning_rate": 0.07421875,
      "calib/nonempty_step_conf_rate": 0.05859375,
      "calib/pce": 0.7708333333333335,
      "calib/std_conf": 0.10984838035522722,
      "calib/step_conf_rate": 0.05859375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3020.0,
      "completions/max_terminated_length": 3020.0,
      "completions/mean_length": 689.0546875,
      "completions/mean_terminated_length": 747.4491577148438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.004266666666666667,
      "grad_norm": 0.004446873906999826,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.009,
      "num_tokens": 1099678.0,
      "reward": 0.036271486431360245,
      "reward_std": 0.08171947300434113,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.012273438274860382,
      "rewards/format_reward_step_strict": 0.04296875,
      "step": 4
    },
    {
      "aux_brier/lambda": 0.1,
      "aux_brier/loss": 1.6811566839491066e-09,
      "aux_brier/mean_group_std": 0.01697577418694997,
      "aux_brier/mean_r": 0.6002687683778114,
      "aux_brier/n_active_tok": 28.0,
      "aux_brier/n_groups": 6.6923076923076925,
      "aux_brier/n_step_records": 7.0,
      "calib/answer_extract_rate": 0.07421875,
      "calib/auroc": 0.35714285714285715,
      "calib/avg_num_step_conf": 0.359375,
      "calib/ece": 0.7699999999999999,
      "calib/final_conf_rate": 0.03125,
      "calib/format_rate": 0.02734375,
      "calib/frac_conf_gt_0.9": 0.75,
      "calib/gap": 0.07428571428571418,
      "calib/mean_conf": 0.895,
      "calib/mu_c": 0.96,
      "calib/mu_w": 0.8857142857142858,
      "calib/nonempty_final_conf_rate": 0.03125,
      "calib/nonempty_reasoning_rate": 0.09765625,
      "calib/nonempty_step_conf_rate": 0.05859375,
      "calib/pce": 0.7699999999999999,
      "calib/std_conf": 0.1606237840420901,
      "calib/step_conf_rate": 0.05859375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2968.0,
      "completions/max_terminated_length": 2968.0,
      "completions/mean_length": 719.578125,
      "completions/mean_terminated_length": 797.4545288085938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.005333333333333333,
      "grad_norm": 0.0036794233601540327,
      "learning_rate": 1.25e-06,
      "loss": 0.0054,
      "num_tokens": 1390578.0,
      "reward": 0.027565428987145424,
      "reward_std": 0.07796680927276611,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.00869921874254942,
      "rewards/format_reward_step_strict": 0.02734375,
      "step": 5
    },
    {
      "aux_brier/lambda": 0.10000000000000002,
      "aux_brier/loss": 6.092032317495111e-08,
      "aux_brier/mean_group_std": 0.08271630606675115,
      "aux_brier/mean_r": 0.49483007053304573,
      "aux_brier/n_active_tok": 26.25,
      "aux_brier/n_groups": 5.5,
      "aux_brier/n_step_records": 6.5625,
      "calib/answer_extract_rate": 0.0859375,
      "calib/auroc": 0.5641025641025641,
      "calib/avg_num_step_conf": 0.6953125,
      "calib/ece": 0.745625,
      "calib/final_conf_rate": 0.0625,
      "calib/format_rate": 0.0546875,
      "calib/frac_conf_gt_0.9": 0.75,
      "calib/gap": 0.02897435897435896,
      "calib/mean_conf": 0.933125,
      "calib/mu_c": 0.9566666666666667,
      "calib/mu_w": 0.9276923076923077,
      "calib/nonempty_final_conf_rate": 0.0625,
      "calib/nonempty_reasoning_rate": 0.12109375,
      "calib/nonempty_step_conf_rate": 0.09375,
      "calib/pce": 0.745625,
      "calib/std_conf": 0.04958814752539159,
      "calib/step_conf_rate": 0.09375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3053.0,
      "completions/max_terminated_length": 3053.0,
      "completions/mean_length": 630.1640625,
      "completions/mean_terminated_length": 698.3636474609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0064,
      "grad_norm": 0.33460941910743713,
      "learning_rate": 1.5e-06,
      "loss": 0.0258,
      "num_tokens": 1657852.0,
      "reward": 0.04356689751148224,
      "reward_std": 0.09231738746166229,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.01801757887005806,
      "rewards/format_reward_step_strict": 0.0546875,
      "step": 6
    },
    {
      "aux_brier/lambda": 0.09999999999999999,
      "aux_brier/loss": -1.59655296982919e-08,
      "aux_brier/mean_group_std": 0.03944458446963456,
      "aux_brier/mean_r": 0.4118567446097141,
      "aux_brier/n_active_tok": 24.0,
      "aux_brier/n_groups": 5.3,
      "aux_brier/n_step_records": 6.0,
      "calib/answer_extract_rate": 0.0859375,
      "calib/auroc": 0.20370370370370372,
      "calib/avg_num_step_conf": 0.234375,
      "calib/ece": 0.7083333333333334,
      "calib/final_conf_rate": 0.046875,
      "calib/format_rate": 0.03515625,
      "calib/frac_conf_gt_0.9": 0.9166666666666666,
      "calib/gap": -0.028888888888888853,
      "calib/mean_conf": 0.9583333333333334,
      "calib/mu_c": 0.9366666666666666,
      "calib/mu_w": 0.9655555555555555,
      "calib/nonempty_final_conf_rate": 0.046875,
      "calib/nonempty_reasoning_rate": 0.09375,
      "calib/nonempty_step_conf_rate": 0.046875,
      "calib/pce": 0.7083333333333334,
      "calib/std_conf": 0.02939198681424732,
      "calib/step_conf_rate": 0.046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3059.0,
      "completions/max_terminated_length": 3059.0,
      "completions/mean_length": 742.8125,
      "completions/mean_terminated_length": 826.7825927734375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 5.0,
      "epoch": 0.007466666666666667,
      "grad_norm": 0.002587189432233572,
      "learning_rate": 1.75e-06,
      "loss": 0.0202,
      "num_tokens": 1955436.0,
      "reward": 0.03569609299302101,
      "reward_std": 0.07640209794044495,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.00997187476605177,
      "rewards/format_reward_step_strict": 0.03515625,
      "step": 7
    },
    {
      "aux_brier/lambda": 0.10000000000000002,
      "aux_brier/loss": -6.6088248599151165e-09,
      "aux_brier/mean_group_std": 0.016897729228566256,
      "aux_brier/mean_r": 0.4548081345696824,
      "aux_brier/n_active_tok": 23.764705882352942,
      "aux_brier/n_groups": 5.411764705882353,
      "aux_brier/n_step_records": 5.9411764705882355,
      "calib/answer_extract_rate": 0.11328125,
      "calib/auroc": 0.5,
      "calib/avg_num_step_conf": 0.39453125,
      "calib/ece": 0.5121041666666667,
      "calib/final_conf_rate": 0.09375,
      "calib/format_rate": 0.0703125,
      "calib/frac_conf_gt_0.9": 0.75,
      "calib/gap": 0.027035714285714385,
      "calib/mean_conf": 0.8162291666666667,
      "calib/mu_c": 0.8320000000000001,
      "calib/mu_w": 0.8049642857142857,
      "calib/nonempty_final_conf_rate": 0.09375,
      "calib/nonempty_reasoning_rate": 0.12890625,
      "calib/nonempty_step_conf_rate": 0.0859375,
      "calib/pce": 0.4558333333333334,
      "calib/std_conf": 0.3177236082393609,
      "calib/step_conf_rate": 0.0859375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08984375,
      "completions/max_length": 3048.0,
      "completions/max_terminated_length": 3048.0,
      "completions/mean_length": 602.73828125,
      "completions/mean_terminated_length": 662.2360229492188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.008533333333333334,
      "grad_norm": 0.004288786556571722,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0139,
      "num_tokens": 2216249.0,
      "reward": 0.08339668065309525,
      "reward_std": 0.13205386698246002,
      "rewards/accuracy_reward_step": 0.0390625,
      "rewards/final_brier_reward_step": 0.036711715161800385,
      "rewards/format_reward_step_strict": 0.0703125,
      "step": 8
    },
    {
      "aux_brier/lambda": 0.1,
      "aux_brier/loss": -6.720305435946119e-08,
      "aux_brier/mean_group_std": 0.011471152313687427,
      "aux_brier/mean_r": 0.41468368217524537,
      "aux_brier/n_active_tok": 20.857142857142858,
      "aux_brier/n_groups": 4.571428571428571,
      "aux_brier/n_step_records": 5.214285714285714,
      "calib/answer_extract_rate": 0.08984375,
      "calib/auroc": 0.6018518518518519,
      "calib/avg_num_step_conf": 0.2890625,
      "calib/ece": 0.5273333333333334,
      "calib/final_conf_rate": 0.05859375,
      "calib/format_rate": 0.046875,
      "calib/frac_conf_gt_0.9": 0.8666666666666667,
      "calib/gap": 0.060000000000000164,
      "calib/mean_conf": 0.9273333333333335,
      "calib/mu_c": 0.9633333333333334,
      "calib/mu_w": 0.9033333333333332,
      "calib/nonempty_final_conf_rate": 0.05859375,
      "calib/nonempty_reasoning_rate": 0.09765625,
      "calib/nonempty_step_conf_rate": 0.06640625,
      "calib/pce": 0.5273333333333334,
      "calib/std_conf": 0.1376694914964419,
      "calib/step_conf_rate": 0.06640625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3021.0,
      "completions/max_terminated_length": 3021.0,
      "completions/mean_length": 663.39453125,
      "completions/mean_terminated_length": 748.1453247070312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.0096,
      "grad_norm": 0.004142031539231539,
      "learning_rate": 2.25e-06,
      "loss": 0.0572,
      "num_tokens": 2493614.0,
      "reward": 0.05300693213939667,
      "reward_std": 0.11931192874908447,
      "rewards/accuracy_reward_step": 0.0234375,
      "rewards/final_brier_reward_step": 0.02452773228287697,
      "rewards/format_reward_step_strict": 0.046875,
      "step": 9
    },
    {
      "aux_brier/lambda": 0.09999999999999999,
      "aux_brier/loss": 1.40633008038904e-06,
      "aux_brier/mean_group_std": 0.08625601403947647,
      "aux_brier/mean_r": 0.41661931629281,
      "aux_brier/n_active_tok": 32.0,
      "aux_brier/n_groups": 5.2,
      "aux_brier/n_step_records": 8.0,
      "calib/answer_extract_rate": 0.0625,
      "calib/avg_num_step_conf": 0.3125,
      "calib/ece": 0.9378571428571427,
      "calib/final_conf_rate": 0.0546875,
      "calib/format_rate": 0.03515625,
      "calib/frac_conf_gt_0.9": 0.8571428571428571,
      "calib/mean_conf": 0.9378571428571431,
      "calib/mu_c": NaN,
      "calib/mu_w": 0.9378571428571431,
      "calib/nonempty_final_conf_rate": 0.0546875,
      "calib/nonempty_reasoning_rate": 0.08984375,
      "calib/nonempty_step_conf_rate": 0.0703125,
      "calib/pce": 0.9378571428571427,
      "calib/std_conf": 0.0655938337072027,
      "calib/step_conf_rate": 0.0703125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3063.0,
      "completions/max_terminated_length": 3063.0,
      "completions/mean_length": 656.359375,
      "completions/mean_terminated_length": 706.0000610351562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.010666666666666666,
      "grad_norm": 0.016049616038799286,
      "learning_rate": 2.5e-06,
      "loss": 0.0226,
      "num_tokens": 2768442.0,
      "reward": 0.01897294819355011,
      "reward_std": 0.040863893926143646,
      "rewards/accuracy_reward_step": 0.0,
      "rewards/final_brier_reward_step": 0.0055792974308133125,
      "rewards/format_reward_step_strict": 0.03515625,
      "step": 10
    },
    {
      "aux_brier/lambda": 0.10000000000000003,
      "aux_brier/loss": -1.3934533342011926e-08,
      "aux_brier/mean_group_std": 0.07243567145754536,
      "aux_brier/mean_r": 0.4100341393936979,
      "aux_brier/n_active_tok": 36.166666666666664,
      "aux_brier/n_groups": 5.75,
      "aux_brier/n_step_records": 9.041666666666666,
      "calib/answer_extract_rate": 0.17578125,
      "calib/auroc": 0.8355555555555555,
      "calib/avg_num_step_conf": 0.85546875,
      "calib/ece": 0.6752941176470588,
      "calib/final_conf_rate": 0.1328125,
      "calib/format_rate": 0.1171875,
      "calib/frac_conf_gt_0.9": 0.8529411764705882,
      "calib/gap": 0.06195555555555554,
      "calib/mean_conf": 0.9400000000000001,
      "calib/mu_c": 0.9855555555555556,
      "calib/mu_w": 0.9236000000000001,
      "calib/nonempty_final_conf_rate": 0.1328125,
      "calib/nonempty_reasoning_rate": 0.22265625,
      "calib/nonempty_step_conf_rate": 0.17578125,
      "calib/pce": 0.6752941176470588,
      "calib/std_conf": 0.09032620621620152,
      "calib/step_conf_rate": 0.17578125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3010.0,
      "completions/max_terminated_length": 3010.0,
      "completions/mean_length": 677.8359375,
      "completions/mean_terminated_length": 735.2796630859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.011733333333333333,
      "grad_norm": 0.7488620281219482,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.0113,
      "num_tokens": 3046448.0,
      "reward": 0.1034020483493805,
      "reward_std": 0.13478590548038483,
      "rewards/accuracy_reward_step": 0.03515625,
      "rewards/final_brier_reward_step": 0.04642070084810257,
      "rewards/format_reward_step_strict": 0.11328125,
      "step": 11
    },
    {
      "aux_brier/lambda": 0.10000000000000002,
      "aux_brier/loss": 6.589414259741556e-08,
      "aux_brier/mean_group_std": 0.05026715882137679,
      "aux_brier/mean_r": 0.5114636334870325,
      "aux_brier/n_active_tok": 35.23809523809524,
      "aux_brier/n_groups": 6.285714285714286,
      "aux_brier/n_step_records": 8.80952380952381,
      "calib/answer_extract_rate": 0.16796875,
      "calib/auroc": 0.6019736842105263,
      "calib/avg_num_step_conf": 0.73828125,
      "calib/ece": 0.6296296296296297,
      "calib/final_conf_rate": 0.10546875,
      "calib/format_rate": 0.09375,
      "calib/frac_conf_gt_0.9": 0.8518518518518519,
      "calib/gap": -0.01552631578947361,
      "calib/mean_conf": 0.9259259259259258,
      "calib/mu_c": 0.915,
      "calib/mu_w": 0.9305263157894736,
      "calib/nonempty_final_conf_rate": 0.10546875,
      "calib/nonempty_reasoning_rate": 0.1953125,
      "calib/nonempty_step_conf_rate": 0.1484375,
      "calib/pce": 0.6296296296296297,
      "calib/std_conf": 0.11873591173378746,
      "calib/step_conf_rate": 0.1484375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08984375,
      "completions/max_length": 2959.0,
      "completions/max_terminated_length": 2959.0,
      "completions/mean_length": 593.83984375,
      "completions/mean_terminated_length": 652.459228515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0128,
      "grad_norm": 0.6210219860076904,
      "learning_rate": 3e-06,
      "loss": 0.0426,
      "num_tokens": 3302647.0,
      "reward": 0.08681464195251465,
      "reward_std": 0.14767587184906006,
      "rewards/accuracy_reward_step": 0.03125,
      "rewards/final_brier_reward_step": 0.03475859388709068,
      "rewards/format_reward_step_strict": 0.09375,
      "step": 12
    },
    {
      "aux_brier/lambda": 0.10000000000000002,
      "aux_brier/loss": 9.382572043747928e-08,
      "aux_brier/mean_group_std": 0.04230635742120345,
      "aux_brier/mean_r": 0.41870340991872157,
      "aux_brier/n_active_tok": 35.4,
      "aux_brier/n_groups": 6.4,
      "aux_brier/n_step_records": 8.85,
      "calib/answer_extract_rate": 0.15234375,
      "calib/auroc": 0.5178571428571429,
      "calib/avg_num_step_conf": 0.69140625,
      "calib/ece": 0.7665625,
      "calib/final_conf_rate": 0.125,
      "calib/format_rate": 0.09375,
      "calib/frac_conf_gt_0.9": 0.75,
      "calib/gap": 0.0724999999999999,
      "calib/mean_conf": 0.8915625,
      "calib/mu_c": 0.955,
      "calib/mu_w": 0.8825000000000001,
      "calib/nonempty_final_conf_rate": 0.125,
      "calib/nonempty_reasoning_rate": 0.17578125,
      "calib/nonempty_step_conf_rate": 0.125,
      "calib/pce": 0.7665625,
      "calib/std_conf": 0.23365451759756328,
      "calib/step_conf_rate": 0.125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2970.0,
      "completions/max_terminated_length": 2970.0,
      "completions/mean_length": 632.59765625,
      "completions/mean_terminated_length": 686.2076416015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.013866666666666666,
      "grad_norm": 0.1094781756401062,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.05,
      "num_tokens": 3569184.0,
      "reward": 0.06870673596858978,
      "reward_std": 0.1556394249200821,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.024826953187584877,
      "rewards/format_reward_step_strict": 0.09375,
      "step": 13
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 5.6569791104339695e-08,
      "aux_brier/mean_group_std": 0.06792149563949823,
      "aux_brier/mean_r": 0.4946154840274279,
      "aux_brier/n_active_tok": 30.344827586206897,
      "aux_brier/n_groups": 5.0,
      "aux_brier/n_step_records": 7.586206896551724,
      "calib/answer_extract_rate": 0.23046875,
      "calib/auroc": 0.46756756756756757,
      "calib/avg_num_step_conf": 0.8671875,
      "calib/ece": 0.646595744680851,
      "calib/final_conf_rate": 0.18359375,
      "calib/format_rate": 0.16015625,
      "calib/frac_conf_gt_0.9": 0.574468085106383,
      "calib/gap": 0.04943243243243245,
      "calib/mean_conf": 0.8380851063829787,
      "calib/mu_c": 0.877,
      "calib/mu_w": 0.8275675675675676,
      "calib/nonempty_final_conf_rate": 0.18359375,
      "calib/nonempty_reasoning_rate": 0.25390625,
      "calib/nonempty_step_conf_rate": 0.1953125,
      "calib/pce": 0.6359574468085106,
      "calib/std_conf": 0.25984969712764555,
      "calib/step_conf_rate": 0.1953125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2976.0,
      "completions/max_terminated_length": 2976.0,
      "completions/mean_length": 694.9609375,
      "completions/mean_terminated_length": 726.1632080078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.014933333333333333,
      "grad_norm": 0.08381670713424683,
      "learning_rate": 3.5e-06,
      "loss": 0.0793,
      "num_tokens": 3852494.0,
      "reward": 0.13864755630493164,
      "reward_std": 0.23478180170059204,
      "rewards/accuracy_reward_step": 0.04296875,
      "rewards/final_brier_reward_step": 0.06240273267030716,
      "rewards/format_reward_step_strict": 0.16015625,
      "step": 14
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.5840396259049243e-08,
      "aux_brier/mean_group_std": 0.10441008311054162,
      "aux_brier/mean_r": 0.44875280923062527,
      "aux_brier/n_active_tok": 38.3448275862069,
      "aux_brier/n_groups": 6.344827586206897,
      "aux_brier/n_step_records": 9.586206896551724,
      "calib/answer_extract_rate": 0.234375,
      "calib/auroc": 0.39,
      "calib/avg_num_step_conf": 1.1171875,
      "calib/ece": 0.5867777777777777,
      "calib/final_conf_rate": 0.17578125,
      "calib/format_rate": 0.16015625,
      "calib/frac_conf_gt_0.9": 0.7777777777777778,
      "calib/gap": 0.016833333333333145,
      "calib/mean_conf": 0.920111111111111,
      "calib/mu_c": 0.9313333333333332,
      "calib/mu_w": 0.9145000000000001,
      "calib/nonempty_final_conf_rate": 0.17578125,
      "calib/nonempty_reasoning_rate": 0.28515625,
      "calib/nonempty_step_conf_rate": 0.234375,
      "calib/pce": 0.5867777777777777,
      "calib/std_conf": 0.14971264657072184,
      "calib/step_conf_rate": 0.234375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3064.0,
      "completions/max_terminated_length": 3064.0,
      "completions/mean_length": 568.4140625,
      "completions/mean_terminated_length": 627.2155151367188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.016,
      "grad_norm": 0.012408778071403503,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.0791,
      "num_tokens": 4105888.0,
      "reward": 0.15055759251117706,
      "reward_std": 0.2482212483882904,
      "rewards/accuracy_reward_step": 0.05859375,
      "rewards/final_brier_reward_step": 0.07098037004470825,
      "rewards/format_reward_step_strict": 0.1484375,
      "step": 15
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -8.537876257327266e-08,
      "aux_brier/mean_group_std": 0.09054687516205945,
      "aux_brier/mean_r": 0.5068588152499831,
      "aux_brier/n_active_tok": 37.6551724137931,
      "aux_brier/n_groups": 5.793103448275862,
      "aux_brier/n_step_records": 9.413793103448276,
      "calib/answer_extract_rate": 0.21484375,
      "calib/auroc": 0.582175925925926,
      "calib/avg_num_step_conf": 1.0859375,
      "calib/ece": 0.6368749999999999,
      "calib/final_conf_rate": 0.1875,
      "calib/format_rate": 0.1640625,
      "calib/frac_conf_gt_0.9": 0.7291666666666666,
      "calib/gap": 0.06750000000000023,
      "calib/mean_conf": 0.8868749999999999,
      "calib/mu_c": 0.9375000000000001,
      "calib/mu_w": 0.8699999999999999,
      "calib/nonempty_final_conf_rate": 0.1875,
      "calib/nonempty_reasoning_rate": 0.2734375,
      "calib/nonempty_step_conf_rate": 0.23046875,
      "calib/pce": 0.6368749999999999,
      "calib/std_conf": 0.19325238517286145,
      "calib/step_conf_rate": 0.23046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3060.0,
      "completions/max_terminated_length": 3060.0,
      "completions/mean_length": 659.6953125,
      "completions/mean_terminated_length": 712.5822143554688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.017066666666666667,
      "grad_norm": 0.37085506319999695,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0919,
      "num_tokens": 4383618.0,
      "reward": 0.1381245106458664,
      "reward_std": 0.26648402214050293,
      "rewards/accuracy_reward_step": 0.046875,
      "rewards/final_brier_reward_step": 0.06031055003404617,
      "rewards/format_reward_step_strict": 0.15234375,
      "step": 16
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.2811494829278627e-08,
      "aux_brier/mean_group_std": 0.10850807264933728,
      "aux_brier/mean_r": 0.4261221348054745,
      "aux_brier/n_active_tok": 50.32258064516129,
      "aux_brier/n_groups": 7.741935483870968,
      "aux_brier/n_step_records": 12.580645161290322,
      "calib/answer_extract_rate": 0.296875,
      "calib/auroc": 0.5654761904761905,
      "calib/avg_num_step_conf": 1.55859375,
      "calib/ece": 0.6929032258064516,
      "calib/final_conf_rate": 0.2421875,
      "calib/format_rate": 0.20703125,
      "calib/frac_conf_gt_0.9": 0.7580645161290323,
      "calib/gap": 0.04687500000000011,
      "calib/mean_conf": 0.9187096774193547,
      "calib/mu_c": 0.9550000000000001,
      "calib/mu_w": 0.908125,
      "calib/nonempty_final_conf_rate": 0.2421875,
      "calib/nonempty_reasoning_rate": 0.359375,
      "calib/nonempty_step_conf_rate": 0.28515625,
      "calib/pce": 0.6929032258064516,
      "calib/std_conf": 0.1391040904174413,
      "calib/step_conf_rate": 0.28515625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2709.0,
      "completions/max_terminated_length": 2709.0,
      "completions/mean_length": 511.2578125,
      "completions/mean_terminated_length": 545.3416748046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.018133333333333335,
      "grad_norm": 0.1399787813425064,
      "learning_rate": 4.25e-06,
      "loss": 0.083,
      "num_tokens": 4618028.0,
      "reward": 0.1719304621219635,
      "reward_std": 0.2832571268081665,
      "rewards/accuracy_reward_step": 0.0546875,
      "rewards/final_brier_reward_step": 0.07053437829017639,
      "rewards/format_reward_step_strict": 0.19921875,
      "step": 17
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.0705595424968333e-08,
      "aux_brier/mean_group_std": 0.10074239404875009,
      "aux_brier/mean_r": 0.4890022355941549,
      "aux_brier/n_active_tok": 39.06666666666667,
      "aux_brier/n_groups": 5.566666666666666,
      "aux_brier/n_step_records": 9.766666666666667,
      "calib/answer_extract_rate": 0.26953125,
      "calib/auroc": 0.5829787234042552,
      "calib/avg_num_step_conf": 1.15234375,
      "calib/ece": 0.6591935483870968,
      "calib/final_conf_rate": 0.2421875,
      "calib/format_rate": 0.21875,
      "calib/frac_conf_gt_0.9": 0.7580645161290323,
      "calib/gap": 0.03896453900709218,
      "calib/mean_conf": 0.9011290322580646,
      "calib/mu_c": 0.9306666666666668,
      "calib/mu_w": 0.8917021276595746,
      "calib/nonempty_final_conf_rate": 0.2421875,
      "calib/nonempty_reasoning_rate": 0.296875,
      "calib/nonempty_step_conf_rate": 0.25,
      "calib/pce": 0.6591935483870968,
      "calib/std_conf": 0.17225360934192796,
      "calib/step_conf_rate": 0.25,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3057.0,
      "completions/max_terminated_length": 3057.0,
      "completions/mean_length": 642.140625,
      "completions/mean_terminated_length": 670.9713745117188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.0192,
      "grad_norm": 0.41152575612068176,
      "learning_rate": 4.5e-06,
      "loss": 0.1446,
      "num_tokens": 4893136.0,
      "reward": 0.1874864101409912,
      "reward_std": 0.30030369758605957,
      "rewards/accuracy_reward_step": 0.05859375,
      "rewards/final_brier_reward_step": 0.08588320016860962,
      "rewards/format_reward_step_strict": 0.21484375,
      "step": 18
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.464701352901001e-08,
      "aux_brier/mean_group_std": 0.16125124480125577,
      "aux_brier/mean_r": 0.5427735700486835,
      "aux_brier/n_active_tok": 73.0,
      "aux_brier/n_groups": 6.96875,
      "aux_brier/n_step_records": 18.25,
      "calib/answer_extract_rate": 0.52734375,
      "calib/auroc": 0.3931818181818182,
      "calib/avg_num_step_conf": 2.37890625,
      "calib/ece": 0.7021097345132743,
      "calib/final_conf_rate": 0.44140625,
      "calib/format_rate": 0.3671875,
      "calib/frac_conf_gt_0.9": 0.6637168141592921,
      "calib/gap": -0.11874327272727303,
      "calib/mean_conf": 0.8644566371681416,
      "calib/mu_c": 0.7719839999999999,
      "calib/mu_w": 0.8907272727272729,
      "calib/nonempty_final_conf_rate": 0.44140625,
      "calib/nonempty_reasoning_rate": 0.609375,
      "calib/nonempty_step_conf_rate": 0.48046875,
      "calib/pce": 0.6726637168141594,
      "calib/std_conf": 0.21202785961379741,
      "calib/step_conf_rate": 0.48046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2379.0,
      "completions/max_terminated_length": 2379.0,
      "completions/mean_length": 440.4296875,
      "completions/mean_terminated_length": 454.6370849609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.020266666666666665,
      "grad_norm": 0.08732657134532928,
      "learning_rate": 4.75e-06,
      "loss": 0.0775,
      "num_tokens": 5110646.0,
      "reward": 0.2961248755455017,
      "reward_std": 0.3688579201698303,
      "rewards/accuracy_reward_step": 0.1015625,
      "rewards/final_brier_reward_step": 0.12199944257736206,
      "rewards/format_reward_step_strict": 0.328125,
      "step": 19
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -6.4677526945278174e-09,
      "aux_brier/mean_group_std": 0.22795472360291566,
      "aux_brier/mean_r": 0.551406513573312,
      "aux_brier/n_active_tok": 88.125,
      "aux_brier/n_groups": 6.90625,
      "aux_brier/n_step_records": 22.03125,
      "calib/answer_extract_rate": 0.671875,
      "calib/auroc": 0.5786764705882353,
      "calib/avg_num_step_conf": 2.82421875,
      "calib/ece": 0.6226138364779874,
      "calib/final_conf_rate": 0.62109375,
      "calib/format_rate": 0.515625,
      "calib/frac_conf_gt_0.9": 0.6981132075471698,
      "calib/gap": 0.05920924369747904,
      "calib/mean_conf": 0.8741861635220127,
      "calib/mu_c": 0.9185000000000001,
      "calib/mu_w": 0.859290756302521,
      "calib/nonempty_final_conf_rate": 0.62109375,
      "calib/nonempty_reasoning_rate": 0.7265625,
      "calib/nonempty_step_conf_rate": 0.6171875,
      "calib/pce": 0.6226138364779874,
      "calib/std_conf": 0.201787693548619,
      "calib/step_conf_rate": 0.6171875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2937.0,
      "completions/max_terminated_length": 2937.0,
      "completions/mean_length": 370.1953125,
      "completions/mean_terminated_length": 385.243896484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.021333333333333333,
      "grad_norm": 0.04635249078273773,
      "learning_rate": 5e-06,
      "loss": 0.134,
      "num_tokens": 5310288.0,
      "reward": 0.4592316746711731,
      "reward_std": 0.46332481503486633,
      "rewards/accuracy_reward_step": 0.16015625,
      "rewards/final_brier_reward_step": 0.21192675828933716,
      "rewards/format_reward_step_strict": 0.4921875,
      "step": 20
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.8800755660630664e-08,
      "aux_brier/mean_group_std": 0.2036430602918916,
      "aux_brier/mean_r": 0.4884979263540808,
      "aux_brier/n_active_tok": 108.125,
      "aux_brier/n_groups": 7.9375,
      "aux_brier/n_step_records": 27.03125,
      "calib/answer_extract_rate": 0.765625,
      "calib/auroc": 0.47993197278911565,
      "calib/avg_num_step_conf": 3.4453125,
      "calib/ece": 0.6724153297682708,
      "calib/final_conf_rate": 0.73046875,
      "calib/format_rate": 0.65234375,
      "calib/frac_conf_gt_0.9": 0.6631016042780749,
      "calib/gap": 0.003720521541950017,
      "calib/mean_conf": 0.8628253119429589,
      "calib/mu_c": 0.8657499999999999,
      "calib/mu_w": 0.8620294784580499,
      "calib/nonempty_final_conf_rate": 0.73046875,
      "calib/nonempty_reasoning_rate": 0.8671875,
      "calib/nonempty_step_conf_rate": 0.77734375,
      "calib/pce": 0.6606684491978608,
      "calib/std_conf": 0.2257502058947101,
      "calib/step_conf_rate": 0.77734375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2741.0,
      "completions/max_terminated_length": 2741.0,
      "completions/mean_length": 350.2421875,
      "completions/mean_terminated_length": 354.395263671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 13.0,
      "epoch": 0.0224,
      "grad_norm": 0.024518921971321106,
      "learning_rate": 4.9722222222222224e-06,
      "loss": 0.1386,
      "num_tokens": 5502910.0,
      "reward": 0.5317217111587524,
      "reward_std": 0.463791161775589,
      "rewards/accuracy_reward_step": 0.16015625,
      "rewards/final_brier_reward_step": 0.2440744787454605,
      "rewards/format_reward_step_strict": 0.62109375,
      "step": 21
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.4348118521366757e-08,
      "aux_brier/mean_group_std": 0.21018110046635474,
      "aux_brier/mean_r": 0.5201762509289959,
      "aux_brier/n_active_tok": 118.375,
      "aux_brier/n_groups": 7.875,
      "aux_brier/n_step_records": 29.59375,
      "calib/answer_extract_rate": 0.79296875,
      "calib/auroc": 0.6036585365853657,
      "calib/avg_num_step_conf": 3.77734375,
      "calib/ece": 0.676582233502538,
      "calib/final_conf_rate": 0.76953125,
      "calib/format_rate": 0.6953125,
      "calib/frac_conf_gt_0.9": 0.7055837563451777,
      "calib/gap": 0.05874066604127581,
      "calib/mean_conf": 0.8847040609137056,
      "calib/mu_c": 0.9312195121951219,
      "calib/mu_w": 0.8724788461538461,
      "calib/nonempty_final_conf_rate": 0.76953125,
      "calib/nonempty_reasoning_rate": 0.86328125,
      "calib/nonempty_step_conf_rate": 0.796875,
      "calib/pce": 0.676582233502538,
      "calib/std_conf": 0.18038984344476477,
      "calib/step_conf_rate": 0.796875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2918.0,
      "completions/max_terminated_length": 2918.0,
      "completions/mean_length": 349.08984375,
      "completions/mean_terminated_length": 353.229248046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 20.0,
      "epoch": 0.023466666666666667,
      "grad_norm": 0.24271811544895172,
      "learning_rate": 4.944444444444445e-06,
      "loss": 0.1173,
      "num_tokens": 5694093.0,
      "reward": 0.5552133321762085,
      "reward_std": 0.5039442181587219,
      "rewards/accuracy_reward_step": 0.1640625,
      "rewards/final_brier_reward_step": 0.25991591811180115,
      "rewards/format_reward_step_strict": 0.65234375,
      "step": 22
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -9.517676952119913e-09,
      "aux_brier/mean_group_std": 0.20458716808645896,
      "aux_brier/mean_r": 0.46074778540779154,
      "aux_brier/n_active_tok": 119.875,
      "aux_brier/n_groups": 8.0625,
      "aux_brier/n_step_records": 29.96875,
      "calib/answer_extract_rate": 0.8046875,
      "calib/auroc": 0.4563636363636364,
      "calib/avg_num_step_conf": 3.7734375,
      "calib/ece": 0.6743137254901962,
      "calib/final_conf_rate": 0.796875,
      "calib/format_rate": 0.71484375,
      "calib/frac_conf_gt_0.9": 0.7647058823529411,
      "calib/gap": -0.01882597402597408,
      "calib/mean_conf": 0.8994117647058825,
      "calib/mu_c": 0.8852,
      "calib/mu_w": 0.9040259740259741,
      "calib/nonempty_final_conf_rate": 0.796875,
      "calib/nonempty_reasoning_rate": 0.8984375,
      "calib/nonempty_step_conf_rate": 0.8515625,
      "calib/pce": 0.6643137254901962,
      "calib/std_conf": 0.16784285092517265,
      "calib/step_conf_rate": 0.8515625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2896.0,
      "completions/max_terminated_length": 2896.0,
      "completions/mean_length": 327.28125,
      "completions/mean_terminated_length": 327.28125,
      "completions/min_length": 21.0,
      "completions/min_terminated_length": 21.0,
      "epoch": 0.024533333333333334,
      "grad_norm": 0.030861619859933853,
      "learning_rate": 4.9166666666666665e-06,
      "loss": 0.1793,
      "num_tokens": 5881813.0,
      "reward": 0.6187982559204102,
      "reward_std": 0.5070070624351501,
      "rewards/accuracy_reward_step": 0.203125,
      "rewards/final_brier_reward_step": 0.24863046407699585,
      "rewards/format_reward_step_strict": 0.70703125,
      "step": 23
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.1131297938948848e-08,
      "aux_brier/mean_group_std": 0.1958564425362323,
      "aux_brier/mean_r": 0.5294594716014864,
      "aux_brier/n_active_tok": 136.375,
      "aux_brier/n_groups": 8.90625,
      "aux_brier/n_step_records": 34.09375,
      "calib/answer_extract_rate": 0.85546875,
      "calib/auroc": 0.5885695187165776,
      "calib/avg_num_step_conf": 4.43359375,
      "calib/ece": 0.6848598130841121,
      "calib/final_conf_rate": 0.8359375,
      "calib/format_rate": 0.78515625,
      "calib/frac_conf_gt_0.9": 0.719626168224299,
      "calib/gap": 0.04093048128342236,
      "calib/mean_conf": 0.8811214953271028,
      "calib/mu_c": 0.9136363636363636,
      "calib/mu_w": 0.8727058823529412,
      "calib/nonempty_final_conf_rate": 0.8359375,
      "calib/nonempty_reasoning_rate": 0.921875,
      "calib/nonempty_step_conf_rate": 0.87890625,
      "calib/pce": 0.6801869158878504,
      "calib/std_conf": 0.1940051406140566,
      "calib/step_conf_rate": 0.87890625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2692.0,
      "completions/max_terminated_length": 2692.0,
      "completions/mean_length": 303.0078125,
      "completions/mean_terminated_length": 306.6007995605469,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 5.0,
      "epoch": 0.0256,
      "grad_norm": 0.3642455041408539,
      "learning_rate": 4.888888888888889e-06,
      "loss": 0.0942,
      "num_tokens": 6063895.0,
      "reward": 0.6258395910263062,
      "reward_std": 0.45120394229888916,
      "rewards/accuracy_reward_step": 0.1796875,
      "rewards/final_brier_reward_step": 0.2846086025238037,
      "rewards/format_reward_step_strict": 0.75,
      "step": 24
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -6.354167785688203e-08,
      "aux_brier/mean_group_std": 0.19292708907994796,
      "aux_brier/mean_r": 0.5279369356274024,
      "aux_brier/n_active_tok": 140.0,
      "aux_brier/n_groups": 7.71875,
      "aux_brier/n_step_records": 35.0,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5089924670433146,
      "calib/avg_num_step_conf": 4.4453125,
      "calib/ece": 0.6421603375527426,
      "calib/final_conf_rate": 0.92578125,
      "calib/format_rate": 0.8984375,
      "calib/frac_conf_gt_0.9": 0.6877637130801688,
      "calib/gap": 0.012508474576271245,
      "calib/mean_conf": 0.8953248945147679,
      "calib/mu_c": 0.9046666666666666,
      "calib/mu_w": 0.8921581920903954,
      "calib/nonempty_final_conf_rate": 0.92578125,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.95703125,
      "calib/pce": 0.6421603375527426,
      "calib/std_conf": 0.1481714707320978,
      "calib/step_conf_rate": 0.95703125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 706.0,
      "completions/max_terminated_length": 706.0,
      "completions/mean_length": 245.2578125,
      "completions/mean_terminated_length": 246.21961975097656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.02666666666666667,
      "grad_norm": 0.23752334713935852,
      "learning_rate": 4.861111111111111e-06,
      "loss": -0.0143,
      "num_tokens": 6229905.0,
      "reward": 0.7569185495376587,
      "reward_std": 0.44906532764434814,
      "rewards/accuracy_reward_step": 0.23828125,
      "rewards/final_brier_reward_step": 0.3479868769645691,
      "rewards/format_reward_step_strict": 0.86328125,
      "step": 25
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 9.23448233158819e-09,
      "aux_brier/mean_group_std": 0.1978469486302426,
      "aux_brier/mean_r": 0.5483203594789697,
      "aux_brier/n_active_tok": 149.5,
      "aux_brier/n_groups": 8.4375,
      "aux_brier/n_step_records": 37.375,
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.4829678735339113,
      "calib/avg_num_step_conf": 4.71875,
      "calib/ece": 0.6882310924369749,
      "calib/final_conf_rate": 0.9296875,
      "calib/format_rate": 0.87109375,
      "calib/frac_conf_gt_0.9": 0.7058823529411765,
      "calib/gap": -0.023430392656807686,
      "calib/mean_conf": 0.8944579831932773,
      "calib/mu_c": 0.876245283018868,
      "calib/mu_w": 0.8996756756756756,
      "calib/nonempty_final_conf_rate": 0.9296875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.953125,
      "calib/pce": 0.68,
      "calib/std_conf": 0.16985796191578595,
      "calib/step_conf_rate": 0.953125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1587.0,
      "completions/max_terminated_length": 1587.0,
      "completions/mean_length": 271.40625,
      "completions/mean_terminated_length": 272.4706115722656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 26.0,
      "epoch": 0.027733333333333332,
      "grad_norm": 0.04942494258284569,
      "learning_rate": 4.833333333333333e-06,
      "loss": 0.0014,
      "num_tokens": 6404625.0,
      "reward": 0.7099546194076538,
      "reward_std": 0.4411212205886841,
      "rewards/accuracy_reward_step": 0.21484375,
      "rewards/final_brier_reward_step": 0.29294368624687195,
      "rewards/format_reward_step_strict": 0.84375,
      "step": 26
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 8.965374892130895e-09,
      "aux_brier/mean_group_std": 0.1958536275388754,
      "aux_brier/mean_r": 0.5246195927248277,
      "aux_brier/n_active_tok": 140.0,
      "aux_brier/n_groups": 7.65625,
      "aux_brier/n_step_records": 35.0,
      "calib/answer_extract_rate": 0.9140625,
      "calib/auroc": 0.5,
      "calib/avg_num_step_conf": 4.45703125,
      "calib/ece": 0.7244303797468354,
      "calib/final_conf_rate": 0.92578125,
      "calib/format_rate": 0.87109375,
      "calib/frac_conf_gt_0.9": 0.6877637130801688,
      "calib/gap": 0.01561774461028198,
      "calib/mean_conf": 0.8681434599156116,
      "calib/mu_c": 0.8813888888888889,
      "calib/mu_w": 0.8657711442786069,
      "calib/nonempty_final_conf_rate": 0.92578125,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.7203375527426161,
      "calib/std_conf": 0.21613973369684655,
      "calib/step_conf_rate": 0.96484375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2481.0,
      "completions/max_terminated_length": 2481.0,
      "completions/mean_length": 256.06640625,
      "completions/mean_terminated_length": 256.06640625,
      "completions/min_length": 20.0,
      "completions/min_terminated_length": 20.0,
      "epoch": 0.0288,
      "grad_norm": 0.26420357823371887,
      "learning_rate": 4.805555555555556e-06,
      "loss": 0.0394,
      "num_tokens": 6575394.0,
      "reward": 0.6430661082267761,
      "reward_std": 0.4162224531173706,
      "rewards/accuracy_reward_step": 0.14453125,
      "rewards/final_brier_reward_step": 0.2832019329071045,
      "rewards/format_reward_step_strict": 0.85546875,
      "step": 27
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.8132602324093536e-08,
      "aux_brier/mean_group_std": 0.2016200024171586,
      "aux_brier/mean_r": 0.5291456947085008,
      "aux_brier/n_active_tok": 141.25,
      "aux_brier/n_groups": 8.0625,
      "aux_brier/n_step_records": 35.3125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.49616840966462317,
      "calib/avg_num_step_conf": 4.5234375,
      "calib/ece": 0.6449392712550608,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.921875,
      "calib/frac_conf_gt_0.9": 0.6275303643724697,
      "calib/gap": 0.007981428056256568,
      "calib/mean_conf": 0.8757894736842105,
      "calib/mu_c": 0.8818644067796609,
      "calib/mu_w": 0.8738829787234044,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.640931174089069,
      "calib/std_conf": 0.18055045222908706,
      "calib/step_conf_rate": 0.96484375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2386.0,
      "completions/max_terminated_length": 2386.0,
      "completions/mean_length": 249.8125,
      "completions/mean_terminated_length": 249.8125,
      "completions/min_length": 37.0,
      "completions/min_terminated_length": 37.0,
      "epoch": 0.029866666666666666,
      "grad_norm": 0.07079173624515533,
      "learning_rate": 4.777777777777778e-06,
      "loss": 0.0775,
      "num_tokens": 6746290.0,
      "reward": 0.7679669857025146,
      "reward_std": 0.4454382359981537,
      "rewards/accuracy_reward_step": 0.23046875,
      "rewards/final_brier_reward_step": 0.353117972612381,
      "rewards/format_reward_step_strict": 0.8984375,
      "step": 28
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 6.849219413851415e-09,
      "aux_brier/mean_group_std": 0.2091342825495418,
      "aux_brier/mean_r": 0.5685604715827565,
      "aux_brier/n_active_tok": 148.875,
      "aux_brier/n_groups": 7.9375,
      "aux_brier/n_step_records": 37.21875,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.4083418497700562,
      "calib/avg_num_step_conf": 4.671875,
      "calib/ece": 0.7211270491803279,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.9375,
      "calib/frac_conf_gt_0.9": 0.610655737704918,
      "calib/gap": -0.027967552376085614,
      "calib/mean_conf": 0.8657172131147541,
      "calib/mu_c": 0.8421052631578949,
      "calib/mu_w": 0.8700728155339805,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.7155532786885246,
      "calib/std_conf": 0.20098374340613906,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2379.0,
      "completions/max_terminated_length": 2379.0,
      "completions/mean_length": 268.42578125,
      "completions/mean_terminated_length": 268.42578125,
      "completions/min_length": 50.0,
      "completions/min_terminated_length": 50.0,
      "epoch": 0.030933333333333334,
      "grad_norm": 0.09916239231824875,
      "learning_rate": 4.75e-06,
      "loss": 0.0289,
      "num_tokens": 6922135.0,
      "reward": 0.6932663917541504,
      "reward_std": 0.34607529640197754,
      "rewards/accuracy_reward_step": 0.15625,
      "rewards/final_brier_reward_step": 0.2965030074119568,
      "rewards/format_reward_step_strict": 0.92578125,
      "step": 29
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 7.016452467645262e-09,
      "aux_brier/mean_group_std": 0.19812405039959854,
      "aux_brier/mean_r": 0.5731785546310743,
      "aux_brier/n_active_tok": 153.25,
      "aux_brier/n_groups": 9.8125,
      "aux_brier/n_step_records": 38.3125,
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.49194444444444446,
      "calib/avg_num_step_conf": 4.7890625,
      "calib/ece": 0.6851836734693879,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.921875,
      "calib/frac_conf_gt_0.9": 0.5795918367346938,
      "calib/gap": 0.017733333333333268,
      "calib/mean_conf": 0.8688571428571429,
      "calib/mu_c": 0.8833333333333333,
      "calib/mu_w": 0.8656,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.6851836734693879,
      "calib/std_conf": 0.17370958640260262,
      "calib/step_conf_rate": 0.9765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2526.0,
      "completions/max_terminated_length": 2526.0,
      "completions/mean_length": 263.88671875,
      "completions/mean_terminated_length": 263.88671875,
      "completions/min_length": 26.0,
      "completions/min_terminated_length": 26.0,
      "epoch": 0.032,
      "grad_norm": 0.4797416925430298,
      "learning_rate": 4.722222222222222e-06,
      "loss": 0.0018,
      "num_tokens": 7096674.0,
      "reward": 0.7211823463439941,
      "reward_std": 0.36588066816329956,
      "rewards/accuracy_reward_step": 0.1796875,
      "rewards/final_brier_reward_step": 0.3300417959690094,
      "rewards/format_reward_step_strict": 0.91796875,
      "step": 30
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.3833868207100437e-09,
      "aux_brier/mean_group_std": 0.2112627605462767,
      "aux_brier/mean_r": 0.5961313745848398,
      "aux_brier/n_active_tok": 162.0,
      "aux_brier/n_groups": 8.78125,
      "aux_brier/n_step_records": 40.5,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.4347704991087344,
      "calib/avg_num_step_conf": 5.08203125,
      "calib/ece": 0.6572983870967744,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.5443548387096774,
      "calib/gap": -0.00869429590017834,
      "calib/mean_conf": 0.8273790322580645,
      "calib/mu_c": 0.8202272727272728,
      "calib/mu_w": 0.8289215686274511,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.6536290322580646,
      "calib/std_conf": 0.2199578072715507,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1720.0,
      "completions/max_terminated_length": 1720.0,
      "completions/mean_length": 261.76171875,
      "completions/mean_terminated_length": 261.76171875,
      "completions/min_length": 54.0,
      "completions/min_terminated_length": 54.0,
      "epoch": 0.03306666666666667,
      "grad_norm": 0.6870817542076111,
      "learning_rate": 4.694444444444445e-06,
      "loss": -0.0057,
      "num_tokens": 7269597.0,
      "reward": 0.73150634765625,
      "reward_std": 0.3738311529159546,
      "rewards/accuracy_reward_step": 0.171875,
      "rewards/final_brier_reward_step": 0.363525390625,
      "rewards/format_reward_step_strict": 0.9375,
      "step": 31
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.489066333073332e-08,
      "aux_brier/mean_group_std": 0.21115774034924586,
      "aux_brier/mean_r": 0.6755294313460579,
      "aux_brier/n_active_tok": 142.875,
      "aux_brier/n_groups": 7.65625,
      "aux_brier/n_step_records": 35.71875,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.46529634581105167,
      "calib/avg_num_step_conf": 4.4921875,
      "calib/ece": 0.6478629032258065,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.4435483870967742,
      "calib/gap": -0.013204099821747084,
      "calib/mean_conf": 0.8035887096774195,
      "calib/mu_c": 0.7927272727272726,
      "calib/mu_w": 0.8059313725490197,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.6370161290322581,
      "calib/std_conf": 0.21958452381334861,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1562.0,
      "completions/max_terminated_length": 1562.0,
      "completions/mean_length": 240.34375,
      "completions/mean_terminated_length": 240.34375,
      "completions/min_length": 88.0,
      "completions/min_terminated_length": 88.0,
      "epoch": 0.034133333333333335,
      "grad_norm": 0.0818771943449974,
      "learning_rate": 4.666666666666667e-06,
      "loss": 0.0605,
      "num_tokens": 7437829.0,
      "reward": 0.7398393154144287,
      "reward_std": 0.3222612142562866,
      "rewards/accuracy_reward_step": 0.17578125,
      "rewards/final_brier_reward_step": 0.38904494047164917,
      "rewards/format_reward_step_strict": 0.93359375,
      "step": 32
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.1705792274673286e-08,
      "aux_brier/mean_group_std": 0.18984168847228203,
      "aux_brier/mean_r": 0.6854015350212486,
      "aux_brier/n_active_tok": 148.75,
      "aux_brier/n_groups": 8.125,
      "aux_brier/n_step_records": 37.1875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5320335863814124,
      "calib/avg_num_step_conf": 4.6796875,
      "calib/ece": 0.6018795180722891,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.3815261044176707,
      "calib/gap": 0.03542374051069719,
      "calib/mean_conf": 0.7703132530120481,
      "calib/mu_c": 0.7997619047619047,
      "calib/mu_w": 0.7643381642512075,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.6017590361445784,
      "calib/std_conf": 0.23741732499946133,
      "calib/step_conf_rate": 0.98046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1270.0,
      "completions/max_terminated_length": 1270.0,
      "completions/mean_length": 264.72265625,
      "completions/mean_terminated_length": 265.76080322265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 48.0,
      "epoch": 0.0352,
      "grad_norm": 1.760973334312439,
      "learning_rate": 4.638888888888889e-06,
      "loss": -0.0303,
      "num_tokens": 7612470.0,
      "reward": 0.7472243309020996,
      "reward_std": 0.33903104066848755,
      "rewards/accuracy_reward_step": 0.171875,
      "rewards/final_brier_reward_step": 0.4263971745967865,
      "rewards/format_reward_step_strict": 0.9375,
      "step": 33
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.220304423123821e-08,
      "aux_brier/mean_group_std": 0.17374843804909995,
      "aux_brier/mean_r": 0.7320428933024568,
      "aux_brier/n_active_tok": 139.125,
      "aux_brier/n_groups": 7.3125,
      "aux_brier/n_step_records": 34.78125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5317150297619047,
      "calib/avg_num_step_conf": 4.34765625,
      "calib/ece": 0.5133064516129033,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.31048387096774194,
      "calib/gap": 0.021443452380952244,
      "calib/mean_conf": 0.7316129032258064,
      "calib/mu_c": 0.7482142857142857,
      "calib/mu_w": 0.7267708333333335,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.5095564516129033,
      "calib/std_conf": 0.2545644195571614,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2056.0,
      "completions/max_terminated_length": 2056.0,
      "completions/mean_length": 237.6171875,
      "completions/mean_terminated_length": 237.6171875,
      "completions/min_length": 52.0,
      "completions/min_terminated_length": 52.0,
      "epoch": 0.03626666666666667,
      "grad_norm": 0.17196181416511536,
      "learning_rate": 4.611111111111112e-06,
      "loss": 0.031,
      "num_tokens": 7778412.0,
      "reward": 0.8209953308105469,
      "reward_std": 0.3929782509803772,
      "rewards/accuracy_reward_step": 0.22265625,
      "rewards/final_brier_reward_step": 0.48710623383522034,
      "rewards/format_reward_step_strict": 0.953125,
      "step": 34
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.531298018941918e-08,
      "aux_brier/mean_group_std": 0.16789939570729087,
      "aux_brier/mean_r": 0.7882815575863105,
      "aux_brier/n_active_tok": 151.125,
      "aux_brier/n_groups": 8.53125,
      "aux_brier/n_step_records": 37.78125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5594684385382059,
      "calib/avg_num_step_conf": 4.73828125,
      "calib/ece": 0.5158300395256918,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.2766798418972332,
      "calib/gap": 0.059986157253599015,
      "calib/mean_conf": 0.685790513833992,
      "calib/mu_c": 0.7355813953488372,
      "calib/mu_w": 0.6755952380952381,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.5158300395256918,
      "calib/std_conf": 0.2782372654379569,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 681.0,
      "completions/max_terminated_length": 681.0,
      "completions/mean_length": 244.8671875,
      "completions/mean_terminated_length": 245.8274688720703,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 74.0,
      "epoch": 0.037333333333333336,
      "grad_norm": 0.07762862741947174,
      "learning_rate": 4.583333333333333e-06,
      "loss": 0.0018,
      "num_tokens": 7950354.0,
      "reward": 0.7836238145828247,
      "reward_std": 0.34538066387176514,
      "rewards/accuracy_reward_step": 0.171875,
      "rewards/final_brier_reward_step": 0.517307698726654,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 35
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.978279810081588e-08,
      "aux_brier/mean_group_std": 0.14886058145328834,
      "aux_brier/mean_r": 0.8125586780805625,
      "aux_brier/n_active_tok": 148.625,
      "aux_brier/n_groups": 8.1875,
      "aux_brier/n_step_records": 37.15625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.45760743321718933,
      "calib/avg_num_step_conf": 4.703125,
      "calib/ece": 0.35986000000000007,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.172,
      "calib/gap": -0.03168699186991886,
      "calib/mean_conf": 0.63794,
      "calib/mu_c": 0.6166463414634146,
      "calib/mu_w": 0.6483333333333334,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.33490000000000003,
      "calib/std_conf": 0.2565218439041791,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 752.0,
      "completions/max_terminated_length": 752.0,
      "completions/mean_length": 233.54296875,
      "completions/mean_terminated_length": 234.45883178710938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 35.0,
      "epoch": 0.0384,
      "grad_norm": 0.6392661333084106,
      "learning_rate": 4.555555555555556e-06,
      "loss": -0.0299,
      "num_tokens": 8112853.0,
      "reward": 0.9495212435722351,
      "reward_std": 0.4306148886680603,
      "rewards/accuracy_reward_step": 0.32421875,
      "rewards/final_brier_reward_step": 0.58714759349823,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 36
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 6.859237547396368e-08,
      "aux_brier/mean_group_std": 0.1333417137353705,
      "aux_brier/mean_r": 0.8526910328969939,
      "aux_brier/n_active_tok": 151.375,
      "aux_brier/n_groups": 8.53125,
      "aux_brier/n_step_records": 37.84375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5208987783595114,
      "calib/avg_num_step_conf": 4.76953125,
      "calib/ece": 0.3800996015936255,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.14741035856573706,
      "calib/gap": 0.019286649214659723,
      "calib/mean_conf": 0.5515737051792828,
      "calib/mu_c": 0.56625,
      "calib/mu_w": 0.5469633507853403,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.3463147410358566,
      "calib/std_conf": 0.295522384308376,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 796.0,
      "completions/max_terminated_length": 796.0,
      "completions/mean_length": 245.25390625,
      "completions/mean_terminated_length": 246.2156982421875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 52.0,
      "epoch": 0.039466666666666664,
      "grad_norm": 0.16471506655216217,
      "learning_rate": 4.527777777777778e-06,
      "loss": 0.0028,
      "num_tokens": 8282734.0,
      "reward": 0.8762906789779663,
      "reward_std": 0.36260733008384705,
      "rewards/accuracy_reward_step": 0.23828125,
      "rewards/final_brier_reward_step": 0.62235027551651,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 37
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.332215516016479e-08,
      "aux_brier/mean_group_std": 0.09968430742331046,
      "aux_brier/mean_r": 0.8990072955013308,
      "aux_brier/n_active_tok": 154.125,
      "aux_brier/n_groups": 9.21875,
      "aux_brier/n_step_records": 38.53125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4822916666666666,
      "calib/avg_num_step_conf": 4.81640625,
      "calib/ece": 0.3063888888888889,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.12301587301587301,
      "calib/gap": -0.02189583333333328,
      "calib/mean_conf": 0.5038492063492064,
      "calib/mu_c": 0.4871666666666667,
      "calib/mu_w": 0.5090625,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.28607142857142853,
      "calib/std_conf": 0.2990710901170733,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2226.0,
      "completions/max_terminated_length": 2226.0,
      "completions/mean_length": 262.5234375,
      "completions/mean_terminated_length": 262.5234375,
      "completions/min_length": 58.0,
      "completions/min_terminated_length": 58.0,
      "epoch": 0.04053333333333333,
      "grad_norm": 0.07935165613889694,
      "learning_rate": 4.5e-06,
      "loss": 0.0893,
      "num_tokens": 8456828.0,
      "reward": 0.8824505805969238,
      "reward_std": 0.330085813999176,
      "rewards/accuracy_reward_step": 0.234375,
      "rewards/final_brier_reward_step": 0.6391773819923401,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 38
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.1815635095335786e-08,
      "aux_brier/mean_group_std": 0.09178553734832681,
      "aux_brier/mean_r": 0.907417051799585,
      "aux_brier/n_active_tok": 156.125,
      "aux_brier/n_groups": 8.75,
      "aux_brier/n_step_records": 39.03125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5073298429319373,
      "calib/avg_num_step_conf": 4.890625,
      "calib/ece": 0.25776892430278886,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.055776892430278883,
      "calib/gap": 0.0013132635253054747,
      "calib/mean_conf": 0.4381673306772908,
      "calib/mu_c": 0.4391666666666667,
      "calib/mu_w": 0.43785340314136123,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.22844621513944224,
      "calib/std_conf": 0.2740238522501975,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2102.0,
      "completions/max_terminated_length": 2102.0,
      "completions/mean_length": 266.1171875,
      "completions/mean_terminated_length": 266.1171875,
      "completions/min_length": 63.0,
      "completions/min_terminated_length": 63.0,
      "epoch": 0.0416,
      "grad_norm": 0.33297401666641235,
      "learning_rate": 4.472222222222223e-06,
      "loss": 0.0174,
      "num_tokens": 8631042.0,
      "reward": 0.8796318173408508,
      "reward_std": 0.29891446232795715,
      "rewards/accuracy_reward_step": 0.234375,
      "rewards/final_brier_reward_step": 0.6747773885726929,
      "rewards/format_reward_step_strict": 0.953125,
      "step": 39
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.216957964735645e-07,
      "aux_brier/mean_group_std": 0.07120818175989829,
      "aux_brier/mean_r": 0.9208270935628,
      "aux_brier/n_active_tok": 178.25,
      "aux_brier/n_groups": 10.59375,
      "aux_brier/n_step_records": 44.5625,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5219904389395915,
      "calib/avg_num_step_conf": 5.58203125,
      "calib/ece": 0.23673228346456693,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.05905511811023622,
      "calib/gap": 0.008357235984354616,
      "calib/mean_conf": 0.40578740157480314,
      "calib/mu_c": 0.41220338983050847,
      "calib/mu_w": 0.40384615384615385,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.20511811023622045,
      "calib/std_conf": 0.2758422500776374,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2127.0,
      "completions/max_terminated_length": 2127.0,
      "completions/mean_length": 291.62109375,
      "completions/mean_terminated_length": 291.62109375,
      "completions/min_length": 54.0,
      "completions/min_terminated_length": 54.0,
      "epoch": 0.042666666666666665,
      "grad_norm": 0.1428987979888916,
      "learning_rate": 4.444444444444444e-06,
      "loss": 0.0219,
      "num_tokens": 8812457.0,
      "reward": 0.9007915258407593,
      "reward_std": 0.303219199180603,
      "rewards/accuracy_reward_step": 0.234375,
      "rewards/final_brier_reward_step": 0.7125409841537476,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 40
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -4.718792642743885e-08,
      "aux_brier/mean_group_std": 0.08801166791896568,
      "aux_brier/mean_r": 0.9106452685143135,
      "aux_brier/n_active_tok": 167.125,
      "aux_brier/n_groups": 9.6875,
      "aux_brier/n_step_records": 41.78125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5757065319109115,
      "calib/avg_num_step_conf": 5.265625,
      "calib/ece": 0.18740157480314962,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.03937007874015748,
      "calib/gap": 0.06223469960696243,
      "calib/mean_conf": 0.35874015748031496,
      "calib/mu_c": 0.39230769230769236,
      "calib/mu_w": 0.33007299270072993,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.04275590551181101,
      "calib/std_conf": 0.25022507073667105,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 919.0,
      "completions/max_terminated_length": 919.0,
      "completions/mean_length": 271.484375,
      "completions/mean_terminated_length": 272.5490417480469,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 64.0,
      "epoch": 0.04373333333333333,
      "grad_norm": 0.3115524649620056,
      "learning_rate": 4.416666666666667e-06,
      "loss": 0.0019,
      "num_tokens": 8989205.0,
      "reward": 1.1249455213546753,
      "reward_std": 0.3867625296115875,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.6950949430465698,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 41
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.563943969358107e-07,
      "aux_brier/mean_group_std": 0.078453620524218,
      "aux_brier/mean_r": 0.9349983468055203,
      "aux_brier/n_active_tok": 154.75,
      "aux_brier/n_groups": 8.15625,
      "aux_brier/n_step_records": 38.6875,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6110141131965663,
      "calib/avg_num_step_conf": 4.90234375,
      "calib/ece": 0.1409881422924901,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.03162055335968379,
      "calib/gap": 0.08844245598719624,
      "calib/mean_conf": 0.2941106719367589,
      "calib/mu_c": 0.3549367088607595,
      "calib/mu_w": 0.26649425287356326,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.061422924901185755,
      "calib/std_conf": 0.23719811186740408,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 560.0,
      "completions/max_terminated_length": 560.0,
      "completions/mean_length": 234.99609375,
      "completions/mean_terminated_length": 235.91766357421875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 73.0,
      "epoch": 0.0448,
      "grad_norm": 0.0935867577791214,
      "learning_rate": 4.388888888888889e-06,
      "loss": -0.0048,
      "num_tokens": 9153732.0,
      "reward": 0.9800031185150146,
      "reward_std": 0.3009245991706848,
      "rewards/accuracy_reward_step": 0.30859375,
      "rewards/final_brier_reward_step": 0.7559499740600586,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 42
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.9479223564711212e-07,
      "aux_brier/mean_group_std": 0.06715368757473159,
      "aux_brier/mean_r": 0.9352575832435351,
      "aux_brier/n_active_tok": 175.75,
      "aux_brier/n_groups": 10.625,
      "aux_brier/n_step_records": 43.9375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5921327547412407,
      "calib/avg_num_step_conf": 5.609375,
      "calib/ece": 0.12836653386454183,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0199203187250996,
      "calib/gap": 0.06621423979427832,
      "calib/mean_conf": 0.28804780876494024,
      "calib/mu_c": 0.3363235294117647,
      "calib/mu_w": 0.27010928961748637,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.07274900398406374,
      "calib/std_conf": 0.23657873049429048,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1347.0,
      "completions/max_terminated_length": 1347.0,
      "completions/mean_length": 299.45703125,
      "completions/mean_terminated_length": 300.6313781738281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 61.0,
      "epoch": 0.04586666666666667,
      "grad_norm": 0.37313738465309143,
      "learning_rate": 4.361111111111112e-06,
      "loss": 0.0321,
      "num_tokens": 9335617.0,
      "reward": 0.9451814889907837,
      "reward_std": 0.29758742451667786,
      "rewards/accuracy_reward_step": 0.26953125,
      "rewards/final_brier_reward_step": 0.7572882771492004,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 43
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.963976862835338e-07,
      "aux_brier/mean_group_std": 0.07284111904645876,
      "aux_brier/mean_r": 0.939259584647508,
      "aux_brier/n_active_tok": 179.375,
      "aux_brier/n_groups": 10.1875,
      "aux_brier/n_step_records": 44.84375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5067740901074235,
      "calib/avg_num_step_conf": 5.6171875,
      "calib/ece": 0.2019607843137255,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0392156862745098,
      "calib/gap": -0.005894660894660841,
      "calib/mean_conf": 0.2748235294117647,
      "calib/mu_c": 0.2704545454545455,
      "calib/mu_w": 0.27634920634920634,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.10898039215686275,
      "calib/std_conf": 0.24381901885655005,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 855.0,
      "completions/max_terminated_length": 855.0,
      "completions/mean_length": 309.26171875,
      "completions/mean_terminated_length": 310.4745178222656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 58.0,
      "epoch": 0.046933333333333334,
      "grad_norm": 0.22018788754940033,
      "learning_rate": 4.333333333333334e-06,
      "loss": 0.0311,
      "num_tokens": 9521108.0,
      "reward": 0.9387511610984802,
      "reward_std": 0.23437447845935822,
      "rewards/accuracy_reward_step": 0.2578125,
      "rewards/final_brier_reward_step": 0.7393797039985657,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 44
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.7874278649142603e-07,
      "aux_brier/mean_group_std": 0.07398399706363962,
      "aux_brier/mean_r": 0.9395140010316367,
      "aux_brier/n_active_tok": 180.875,
      "aux_brier/n_groups": 11.15625,
      "aux_brier/n_step_records": 45.21875,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5308208366219416,
      "calib/avg_num_step_conf": 5.765625,
      "calib/ece": 0.188804780876494,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.01593625498007968,
      "calib/gap": 0.0098382004735596,
      "calib/mean_conf": 0.22219123505976093,
      "calib/mu_c": 0.2292857142857143,
      "calib/mu_w": 0.2194475138121547,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.06605577689243029,
      "calib/std_conf": 0.21994652393095757,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 946.0,
      "completions/max_terminated_length": 946.0,
      "completions/mean_length": 302.66796875,
      "completions/mean_terminated_length": 303.85491943359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 49.0,
      "epoch": 0.048,
      "grad_norm": 0.12611983716487885,
      "learning_rate": 4.305555555555556e-06,
      "loss": -0.0078,
      "num_tokens": 9703639.0,
      "reward": 0.9419686794281006,
      "reward_std": 0.3299618065357208,
      "rewards/accuracy_reward_step": 0.28125,
      "rewards/final_brier_reward_step": 0.7288120985031128,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 45
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.4474342063808265e-07,
      "aux_brier/mean_group_std": 0.06889973066903073,
      "aux_brier/mean_r": 0.9353062366776169,
      "aux_brier/n_active_tok": 200.875,
      "aux_brier/n_groups": 12.78125,
      "aux_brier/n_step_records": 50.21875,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.512075807334428,
      "calib/avg_num_step_conf": 6.27734375,
      "calib/ece": 0.2322352941176471,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.00392156862745098,
      "calib/gap": 0.01679597701149424,
      "calib/mean_conf": 0.20698039215686276,
      "calib/mu_c": 0.21804597701149425,
      "calib/mu_w": 0.20125,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.049019607843137275,
      "calib/std_conf": 0.21204674754625277,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2091.0,
      "completions/max_terminated_length": 2091.0,
      "completions/mean_length": 337.0390625,
      "completions/mean_terminated_length": 337.0390625,
      "completions/min_length": 80.0,
      "completions/min_terminated_length": 80.0,
      "epoch": 0.04906666666666667,
      "grad_norm": 0.09090101718902588,
      "learning_rate": 4.277777777777778e-06,
      "loss": 0.04,
      "num_tokens": 9894689.0,
      "reward": 1.0143649578094482,
      "reward_std": 0.27545514702796936,
      "rewards/accuracy_reward_step": 0.33984375,
      "rewards/final_brier_reward_step": 0.7137101888656616,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 46
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -8.303458840508426e-07,
      "aux_brier/mean_group_std": 0.05258738284857578,
      "aux_brier/mean_r": 0.9464171112388273,
      "aux_brier/n_active_tok": 210.625,
      "aux_brier/n_groups": 13.8125,
      "aux_brier/n_step_records": 52.65625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4995444350995234,
      "calib/avg_num_step_conf": 6.59375,
      "calib/ece": 0.25390438247011954,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.00796812749003984,
      "calib/gap": -0.026028174936921777,
      "calib/mean_conf": 0.17286852589641435,
      "calib/mu_c": 0.15586206896551724,
      "calib/mu_w": 0.18189024390243902,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0400796812749004,
      "calib/std_conf": 0.17991072001609987,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2961.0,
      "completions/max_terminated_length": 2961.0,
      "completions/mean_length": 360.734375,
      "completions/mean_terminated_length": 360.734375,
      "completions/min_length": 71.0,
      "completions/min_terminated_length": 71.0,
      "epoch": 0.050133333333333335,
      "grad_norm": 0.08796198666095734,
      "learning_rate": 4.25e-06,
      "loss": 0.09,
      "num_tokens": 10093013.0,
      "reward": 0.994473934173584,
      "reward_std": 0.28420084714889526,
      "rewards/accuracy_reward_step": 0.33984375,
      "rewards/final_brier_reward_step": 0.6810207366943359,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 47
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 7.772790537874386e-07,
      "aux_brier/mean_group_std": 0.05807757377194304,
      "aux_brier/mean_r": 0.9566157598240606,
      "aux_brier/n_active_tok": 167.625,
      "aux_brier/n_groups": 9.28125,
      "aux_brier/n_step_records": 41.90625,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.4553491572067424,
      "calib/avg_num_step_conf": 5.2421875,
      "calib/ece": 0.24847656250000003,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0257784657722738,
      "calib/mean_conf": 0.1291015625,
      "calib/mu_c": 0.11188235294117649,
      "calib/mu_w": 0.1376608187134503,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.022773437500000004,
      "calib/std_conf": 0.1452704019580678,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1096.0,
      "completions/max_terminated_length": 1096.0,
      "completions/mean_length": 289.5078125,
      "completions/mean_terminated_length": 290.6431579589844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 68.0,
      "epoch": 0.0512,
      "grad_norm": 0.02429216168820858,
      "learning_rate": 4.222222222222223e-06,
      "loss": -0.0063,
      "num_tokens": 10270815.0,
      "reward": 1.0027838945388794,
      "reward_std": 0.26423102617263794,
      "rewards/accuracy_reward_step": 0.33203125,
      "rewards/final_brier_reward_step": 0.6986355185508728,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 48
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.090116954540534e-07,
      "aux_brier/mean_group_std": 0.06381136482834818,
      "aux_brier/mean_r": 0.9547274216868711,
      "aux_brier/n_active_tok": 192.125,
      "aux_brier/n_groups": 10.71875,
      "aux_brier/n_step_records": 48.03125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5651565622918054,
      "calib/avg_num_step_conf": 6.015625,
      "calib/ece": 0.2537154150197628,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.003952569169960474,
      "calib/gap": 0.03445436375749497,
      "calib/mean_conf": 0.13711462450592884,
      "calib/mu_c": 0.1586315789473684,
      "calib/mu_w": 0.12417721518987343,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.007667984189723321,
      "calib/std_conf": 0.15052731654829482,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2251.0,
      "completions/max_terminated_length": 2251.0,
      "completions/mean_length": 330.765625,
      "completions/mean_terminated_length": 330.765625,
      "completions/min_length": 87.0,
      "completions/min_terminated_length": 87.0,
      "epoch": 0.05226666666666667,
      "grad_norm": 0.10576039552688599,
      "learning_rate": 4.194444444444445e-06,
      "loss": 0.0535,
      "num_tokens": 10460027.0,
      "reward": 1.0348154306411743,
      "reward_std": 0.24864214658737183,
      "rewards/accuracy_reward_step": 0.37109375,
      "rewards/final_brier_reward_step": 0.6939488649368286,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 49
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.1354192910872918e-07,
      "aux_brier/mean_group_std": 0.053774767738826584,
      "aux_brier/mean_r": 0.9551484619635056,
      "aux_brier/n_active_tok": 218.25,
      "aux_brier/n_groups": 16.40625,
      "aux_brier/n_step_records": 54.5625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5377824858757062,
      "calib/avg_num_step_conf": 6.8203125,
      "calib/ece": 0.40528000000000003,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0036312275295326407,
      "calib/mean_conf": 0.10768000000000001,
      "calib/mu_c": 0.10576271186440676,
      "calib/mu_w": 0.1093939393939394,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.020479999999999998,
      "calib/std_conf": 0.15038689304590344,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2664.0,
      "completions/max_terminated_length": 2664.0,
      "completions/mean_length": 406.77734375,
      "completions/mean_terminated_length": 406.77734375,
      "completions/min_length": 89.0,
      "completions/min_terminated_length": 89.0,
      "epoch": 0.05333333333333334,
      "grad_norm": 0.2878992557525635,
      "learning_rate": 4.166666666666667e-06,
      "loss": 0.1021,
      "num_tokens": 10669522.0,
      "reward": 1.0921945571899414,
      "reward_std": 0.2993980646133423,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.5797156095504761,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 50
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 5.598089822672625e-07,
      "aux_brier/mean_group_std": 0.051267485836996016,
      "aux_brier/mean_r": 0.964020473786176,
      "aux_brier/n_active_tok": 185.375,
      "aux_brier/n_groups": 11.09375,
      "aux_brier/n_step_records": 46.34375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5456861413043478,
      "calib/avg_num_step_conf": 5.79296875,
      "calib/ece": 0.2904365079365079,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.007936507936507936,
      "calib/gap": 0.004682065217391285,
      "calib/mean_conf": 0.0928968253968254,
      "calib/mu_c": 0.09586956521739129,
      "calib/mu_w": 0.0911875,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.009126984126984128,
      "calib/std_conf": 0.13054199162424127,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1443.0,
      "completions/max_terminated_length": 1443.0,
      "completions/mean_length": 324.91015625,
      "completions/mean_terminated_length": 327.468505859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 54.0,
      "epoch": 0.0544,
      "grad_norm": 0.20882609486579895,
      "learning_rate": 4.138888888888889e-06,
      "loss": -0.0007,
      "num_tokens": 10861995.0,
      "reward": 1.016768455505371,
      "reward_std": 0.27805376052856445,
      "rewards/accuracy_reward_step": 0.359375,
      "rewards/final_brier_reward_step": 0.6686363220214844,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 51
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 6.149199491822799e-07,
      "aux_brier/mean_group_std": 0.07735836416445827,
      "aux_brier/mean_r": 0.9498911083880336,
      "aux_brier/n_active_tok": 158.125,
      "aux_brier/n_groups": 8.8125,
      "aux_brier/n_step_records": 39.53125,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5679389312977099,
      "calib/avg_num_step_conf": 4.9453125,
      "calib/ece": 0.421328125,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.020738931297709942,
      "calib/mean_conf": 0.0721875,
      "calib/mu_c": 0.08280000000000001,
      "calib/mu_w": 0.06206106870229007,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0026171874999999997,
      "calib/std_conf": 0.1045409421889338,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 827.0,
      "completions/max_terminated_length": 827.0,
      "completions/mean_length": 302.4609375,
      "completions/mean_terminated_length": 303.6470642089844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 52.0,
      "epoch": 0.055466666666666664,
      "grad_norm": 0.09126332402229309,
      "learning_rate": 4.111111111111111e-06,
      "loss": -0.0247,
      "num_tokens": 11047377.0,
      "reward": 1.126570701599121,
      "reward_std": 0.3329008221626282,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.5687824487686157,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 52
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.0911899479836293e-06,
      "aux_brier/mean_group_std": 0.07043554288810829,
      "aux_brier/mean_r": 0.9518235448680586,
      "aux_brier/n_active_tok": 189.25,
      "aux_brier/n_groups": 10.59375,
      "aux_brier/n_step_records": 47.3125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.48905018145413587,
      "calib/avg_num_step_conf": 5.92578125,
      "calib/ece": 0.40984189723320164,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0008803654110874637,
      "calib/mean_conf": 0.08430830039525691,
      "calib/mu_c": 0.08385245901639345,
      "calib/mu_w": 0.08473282442748091,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.005968379446640316,
      "calib/std_conf": 0.1234133123383727,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1142.0,
      "completions/max_terminated_length": 1142.0,
      "completions/mean_length": 351.15234375,
      "completions/mean_terminated_length": 352.5294189453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 66.0,
      "epoch": 0.05653333333333333,
      "grad_norm": 0.18907392024993896,
      "learning_rate": 4.083333333333334e-06,
      "loss": -0.0111,
      "num_tokens": 11243096.0,
      "reward": 1.1107137203216553,
      "reward_std": 0.2957611083984375,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.567854642868042,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 53
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.94792701869018e-07,
      "aux_brier/mean_group_std": 0.06574302311170417,
      "aux_brier/mean_r": 0.9539206846222074,
      "aux_brier/n_active_tok": 173.25,
      "aux_brier/n_groups": 10.40625,
      "aux_brier/n_step_records": 43.3125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5529012345679012,
      "calib/avg_num_step_conf": 5.4140625,
      "calib/ece": 0.46313725490196084,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.02050925925925927,
      "calib/mean_conf": 0.06627450980392156,
      "calib/mu_c": 0.07592592592592594,
      "calib/mu_w": 0.05541666666666667,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.09302668359161163,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1197.0,
      "completions/max_terminated_length": 1197.0,
      "completions/mean_length": 310.359375,
      "completions/mean_terminated_length": 310.359375,
      "completions/min_length": 63.0,
      "completions/min_terminated_length": 63.0,
      "epoch": 0.0576,
      "grad_norm": 0.07939843833446503,
      "learning_rate": 4.055555555555556e-06,
      "loss": -0.0116,
      "num_tokens": 11428780.0,
      "reward": 1.156419038772583,
      "reward_std": 0.26543140411376953,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.5319265723228455,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 54
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.9486765046039523e-07,
      "aux_brier/mean_group_std": 0.049335625154782355,
      "aux_brier/mean_r": 0.9636705985066784,
      "aux_brier/n_active_tok": 181.75,
      "aux_brier/n_groups": 10.59375,
      "aux_brier/n_step_records": 45.4375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5019778746228629,
      "calib/avg_num_step_conf": 5.78125,
      "calib/ece": 0.33662698412698405,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.002472008045591688,
      "calib/mean_conf": 0.045119047619047614,
      "calib/mu_c": 0.04357894736842105,
      "calib/mu_w": 0.04605095541401274,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0023809523809523807,
      "calib/std_conf": 0.06821116921327605,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1660.0,
      "completions/max_terminated_length": 1660.0,
      "completions/mean_length": 339.85546875,
      "completions/mean_terminated_length": 341.1882629394531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 55.0,
      "epoch": 0.058666666666666666,
      "grad_norm": 0.6068412661552429,
      "learning_rate": 4.027777777777779e-06,
      "loss": 0.0517,
      "num_tokens": 11623607.0,
      "reward": 1.0230414867401123,
      "reward_std": 0.25731363892555237,
      "rewards/accuracy_reward_step": 0.37109375,
      "rewards/final_brier_reward_step": 0.639041006565094,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 55
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.337756533770218e-06,
      "aux_brier/mean_group_std": 0.059995697663266526,
      "aux_brier/mean_r": 0.9583495937558031,
      "aux_brier/n_active_tok": 213.5,
      "aux_brier/n_groups": 13.28125,
      "aux_brier/n_step_records": 53.375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5153618421052631,
      "calib/avg_num_step_conf": 6.69140625,
      "calib/ece": 0.36038888888888887,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.003968253968253968,
      "calib/gap": 0.010719999999999986,
      "calib/mean_conf": 0.046753968253968256,
      "calib/mu_c": 0.05321999999999999,
      "calib/mu_w": 0.0425,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.005158730158730159,
      "calib/std_conf": 0.09279754260121541,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2089.0,
      "completions/max_terminated_length": 2089.0,
      "completions/mean_length": 401.078125,
      "completions/mean_terminated_length": 402.6510009765625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 101.0,
      "epoch": 0.05973333333333333,
      "grad_norm": 0.4605119824409485,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0343,
      "num_tokens": 11833123.0,
      "reward": 1.034104585647583,
      "reward_std": 0.27225345373153687,
      "rewards/accuracy_reward_step": 0.390625,
      "rewards/final_brier_reward_step": 0.6207932233810425,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 56
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.5314033576796504e-06,
      "aux_brier/mean_group_std": 0.049375711026349946,
      "aux_brier/mean_r": 0.968344623238407,
      "aux_brier/n_active_tok": 188.875,
      "aux_brier/n_groups": 11.40625,
      "aux_brier/n_step_records": 47.21875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5041346153846153,
      "calib/avg_num_step_conf": 6.1484375,
      "calib/ece": 0.48672,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.006641025641025637,
      "calib/mean_conf": 0.038880000000000005,
      "calib/mu_c": 0.0356923076923077,
      "calib/mu_w": 0.042333333333333334,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0028,
      "calib/std_conf": 0.07396989657962218,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2406.0,
      "completions/max_terminated_length": 2406.0,
      "completions/mean_length": 365.08203125,
      "completions/mean_terminated_length": 367.9566955566406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 109.0,
      "epoch": 0.0608,
      "grad_norm": 1.6192066669464111,
      "learning_rate": 3.972222222222223e-06,
      "loss": 0.0277,
      "num_tokens": 12033376.0,
      "reward": 1.1186858415603638,
      "reward_std": 0.276128351688385,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.49818047881126404,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 57
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.279837858720768e-06,
      "aux_brier/mean_group_std": 0.03885044724419651,
      "aux_brier/mean_r": 0.9713866398942863,
      "aux_brier/n_active_tok": 206.75,
      "aux_brier/n_groups": 14.71875,
      "aux_brier/n_step_records": 51.6875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5091141564487575,
      "calib/avg_num_step_conf": 6.49609375,
      "calib/ece": 0.3799801587301588,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.003968253968253968,
      "calib/gap": 0.008191921841190747,
      "calib/mean_conf": 0.031130952380952384,
      "calib/mu_c": 0.036039603960396044,
      "calib/mu_w": 0.027847682119205297,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.005158730158730159,
      "calib/std_conf": 0.08191174786790666,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2650.0,
      "completions/max_terminated_length": 2650.0,
      "completions/mean_length": 410.66015625,
      "completions/mean_terminated_length": 412.2705993652344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 65.0,
      "epoch": 0.06186666666666667,
      "grad_norm": 2.518641948699951,
      "learning_rate": 3.944444444444445e-06,
      "loss": 0.022,
      "num_tokens": 12244825.0,
      "reward": 1.0325732231140137,
      "reward_std": 0.307841032743454,
      "rewards/accuracy_reward_step": 0.39453125,
      "rewards/final_brier_reward_step": 0.6068553924560547,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 58
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.8966270745046643e-06,
      "aux_brier/mean_group_std": 0.045400346765770226,
      "aux_brier/mean_r": 0.9650495315775414,
      "aux_brier/n_active_tok": 203.5,
      "aux_brier/n_groups": 13.09375,
      "aux_brier/n_step_records": 50.875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5511583011583011,
      "calib/avg_num_step_conf": 6.359375,
      "calib/ece": 0.4228685258964143,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.014387387387387386,
      "calib/mean_conf": 0.019362549800796814,
      "calib/mu_c": 0.02738738738738739,
      "calib/mu_w": 0.013000000000000003,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.06028147812927871,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2176.0,
      "completions/max_terminated_length": 2176.0,
      "completions/mean_length": 383.21484375,
      "completions/mean_terminated_length": 386.2322692871094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 83.0,
      "epoch": 0.06293333333333333,
      "grad_norm": 0.35520368814468384,
      "learning_rate": 3.916666666666667e-06,
      "loss": 0.0206,
      "num_tokens": 12449176.0,
      "reward": 1.0694079399108887,
      "reward_std": 0.27380746603012085,
      "rewards/accuracy_reward_step": 0.4375,
      "rewards/final_brier_reward_step": 0.5666944980621338,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 59
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.7207030169608117e-06,
      "aux_brier/mean_group_std": 0.05149790445141064,
      "aux_brier/mean_r": 0.9679355183235356,
      "aux_brier/n_active_tok": 182.125,
      "aux_brier/n_groups": 10.375,
      "aux_brier/n_step_records": 45.53125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4766156462585034,
      "calib/avg_num_step_conf": 5.69140625,
      "calib/ece": 0.40848605577689245,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.016946624803767664,
      "calib/mean_conf": 0.023386454183266934,
      "calib/mu_c": 0.013461538461538462,
      "calib/mu_w": 0.030408163265306126,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.008764940239043825,
      "calib/std_conf": 0.07845168823060447,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2509.0,
      "completions/max_terminated_length": 2509.0,
      "completions/mean_length": 365.5078125,
      "completions/mean_terminated_length": 365.5078125,
      "completions/min_length": 92.0,
      "completions/min_terminated_length": 92.0,
      "epoch": 0.064,
      "grad_norm": 1.4717086553573608,
      "learning_rate": 3.88888888888889e-06,
      "loss": 0.0392,
      "num_tokens": 12651602.0,
      "reward": 1.0382107496261597,
      "reward_std": 0.25028032064437866,
      "rewards/accuracy_reward_step": 0.40625,
      "rewards/final_brier_reward_step": 0.5747183561325073,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 60
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 4.028031238334506e-06,
      "aux_brier/mean_group_std": 0.05582616345583301,
      "aux_brier/mean_r": 0.9608498901411248,
      "aux_brier/n_active_tok": 178.0,
      "aux_brier/n_groups": 11.1875,
      "aux_brier/n_step_records": 44.5,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4845317992969,
      "calib/avg_num_step_conf": 5.5703125,
      "calib/ece": 0.5752362204724408,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.005680409076382231,
      "calib/mean_conf": 0.018858267716535433,
      "calib/mu_c": 0.01651006711409396,
      "calib/mu_w": 0.02219047619047619,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0037401574803149606,
      "calib/std_conf": 0.06028047625451905,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1030.0,
      "completions/max_terminated_length": 1030.0,
      "completions/mean_length": 311.8671875,
      "completions/mean_terminated_length": 313.0902099609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 57.0,
      "epoch": 0.06506666666666666,
      "grad_norm": 0.7951279878616333,
      "learning_rate": 3.861111111111112e-06,
      "loss": 0.0155,
      "num_tokens": 12835504.0,
      "reward": 1.1844792366027832,
      "reward_std": 0.1937452107667923,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.4254167675971985,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 61
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.7647091771211265e-06,
      "aux_brier/mean_group_std": 0.05182514719913517,
      "aux_brier/mean_r": 0.9683787715187064,
      "aux_brier/n_active_tok": 216.5,
      "aux_brier/n_groups": 13.40625,
      "aux_brier/n_step_records": 54.125,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.52943717190511,
      "calib/avg_num_step_conf": 6.796875,
      "calib/ece": 0.39504000000000006,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003007508804571731,
      "calib/mean_conf": 0.012960000000000001,
      "calib/mu_c": 0.014752475247524752,
      "calib/mu_w": 0.01174496644295302,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.002,
      "calib/std_conf": 0.03341913224486836,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2200.0,
      "completions/max_terminated_length": 2200.0,
      "completions/mean_length": 422.99609375,
      "completions/mean_terminated_length": 424.6549377441406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 90.0,
      "epoch": 0.06613333333333334,
      "grad_norm": 0.06681760400533676,
      "learning_rate": 3.833333333333334e-06,
      "loss": 0.0156,
      "num_tokens": 13050871.0,
      "reward": 1.0221278667449951,
      "reward_std": 0.28524404764175415,
      "rewards/accuracy_reward_step": 0.39453125,
      "rewards/final_brier_reward_step": 0.5806984305381775,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 62
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.99389647912779e-06,
      "aux_brier/mean_group_std": 0.05313575891634672,
      "aux_brier/mean_r": 0.9664477263550149,
      "aux_brier/n_active_tok": 204.875,
      "aux_brier/n_groups": 13.5,
      "aux_brier/n_step_records": 51.21875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.48502690724912945,
      "calib/avg_num_step_conf": 6.5,
      "calib/ece": 0.44896825396825385,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0008319088319088327,
      "calib/mean_conf": 0.018015873015873016,
      "calib/mu_c": 0.018461538461538463,
      "calib/mu_w": 0.01762962962962963,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0013492063492063493,
      "calib/std_conf": 0.061133731125956145,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2593.0,
      "completions/max_terminated_length": 2593.0,
      "completions/mean_length": 421.80859375,
      "completions/mean_terminated_length": 425.1299133300781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 100.0,
      "epoch": 0.0672,
      "grad_norm": 0.08206135034561157,
      "learning_rate": 3.8055555555555556e-06,
      "loss": 0.022,
      "num_tokens": 13267494.0,
      "reward": 1.0754882097244263,
      "reward_std": 0.3273842930793762,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.528515636920929,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 63
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.847803295320418e-06,
      "aux_brier/mean_group_std": 0.04095292747955301,
      "aux_brier/mean_r": 0.9677550824873439,
      "aux_brier/n_active_tok": 215.0,
      "aux_brier/n_groups": 15.1875,
      "aux_brier/n_step_records": 53.75,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5263414634146342,
      "calib/avg_num_step_conf": 6.71875,
      "calib/ece": 0.4842338709677419,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003051056910569108,
      "calib/mean_conf": 0.014153225806451614,
      "calib/mu_c": 0.015691056910569108,
      "calib/mu_w": 0.01264,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0012096774193548385,
      "calib/std_conf": 0.03741776884152827,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2435.0,
      "completions/max_terminated_length": 2435.0,
      "completions/mean_length": 412.7734375,
      "completions/mean_terminated_length": 412.7734375,
      "completions/min_length": 105.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.06826666666666667,
      "grad_norm": 0.052142687141895294,
      "learning_rate": 3.777777777777778e-06,
      "loss": 0.1098,
      "num_tokens": 13476940.0,
      "reward": 1.0854132175445557,
      "reward_std": 0.32813769578933716,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.4979027211666107,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 64
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.4041878437787148e-07,
      "aux_brier/mean_group_std": 0.05316383888763907,
      "aux_brier/mean_r": 0.9674401294258691,
      "aux_brier/n_active_tok": 175.875,
      "aux_brier/n_groups": 9.9375,
      "aux_brier/n_step_records": 43.96875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.4761935483870968,
      "calib/avg_num_step_conf": 5.54296875,
      "calib/ece": 0.48520080321285136,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.005476451612903227,
      "calib/mean_conf": 0.016807228915662654,
      "calib/mu_c": 0.01408,
      "calib/mu_w": 0.019556451612903227,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.03152653927513252,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1922.0,
      "completions/max_terminated_length": 1922.0,
      "completions/mean_length": 326.38671875,
      "completions/mean_terminated_length": 328.9566955566406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 116.0,
      "epoch": 0.06933333333333333,
      "grad_norm": 0.11075668781995773,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.0136,
      "num_tokens": 13665519.0,
      "reward": 1.100783348083496,
      "reward_std": 0.23851731419563293,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.4968835115432739,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 65
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.95382818737977e-06,
      "aux_brier/mean_group_std": 0.04170639229763923,
      "aux_brier/mean_r": 0.9752149906375521,
      "aux_brier/n_active_tok": 217.625,
      "aux_brier/n_groups": 14.8125,
      "aux_brier/n_step_records": 54.40625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5184893267651888,
      "calib/avg_num_step_conf": 6.80078125,
      "calib/ece": 0.39896000000000004,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.004,
      "calib/gap": 0.018896551724137935,
      "calib/mean_conf": 0.02104,
      "calib/mu_c": 0.032,
      "calib/mu_w": 0.013103448275862068,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.07690330552063417,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2430.0,
      "completions/max_terminated_length": 2430.0,
      "completions/mean_length": 429.39453125,
      "completions/mean_terminated_length": 429.39453125,
      "completions/min_length": 107.0,
      "completions/min_terminated_length": 107.0,
      "epoch": 0.0704,
      "grad_norm": 0.019984150305390358,
      "learning_rate": 3.7222222222222225e-06,
      "loss": 0.0783,
      "num_tokens": 13881796.0,
      "reward": 1.0489559173583984,
      "reward_std": 0.24764403700828552,
      "rewards/accuracy_reward_step": 0.4140625,
      "rewards/final_brier_reward_step": 0.5864484310150146,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 66
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.49393127735409e-07,
      "aux_brier/mean_group_std": 0.06469581027960165,
      "aux_brier/mean_r": 0.9589652238567143,
      "aux_brier/n_active_tok": 187.625,
      "aux_brier/n_groups": 11.78125,
      "aux_brier/n_step_records": 46.90625,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.43666605177396545,
      "calib/avg_num_step_conf": 5.89453125,
      "calib/ece": 0.5223828125000001,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.006274980015987211,
      "calib/mean_conf": 0.0221484375,
      "calib/mu_c": 0.019280575539568346,
      "calib/mu_w": 0.025555555555555557,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.00078125,
      "calib/std_conf": 0.05436128991579021,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1291.0,
      "completions/max_terminated_length": 1291.0,
      "completions/mean_length": 382.62890625,
      "completions/mean_terminated_length": 384.1294250488281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 114.0,
      "epoch": 0.07146666666666666,
      "grad_norm": 0.11290935426950455,
      "learning_rate": 3.694444444444445e-06,
      "loss": -0.0363,
      "num_tokens": 14084757.0,
      "reward": 1.1530730724334717,
      "reward_std": 0.1719469279050827,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.46385470032691956,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 67
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.7287329072710484e-06,
      "aux_brier/mean_group_std": 0.03377295447585385,
      "aux_brier/mean_r": 0.9799369483727938,
      "aux_brier/n_active_tok": 186.125,
      "aux_brier/n_groups": 11.46875,
      "aux_brier/n_step_records": 46.53125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.49965473948524797,
      "calib/avg_num_step_conf": 5.87109375,
      "calib/ece": 0.45184189723320156,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0010290018832391692,
      "calib/mean_conf": 0.01614229249011858,
      "calib/mu_c": 0.015593220338983053,
      "calib/mu_w": 0.016622222222222222,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0007905138339920949,
      "calib/std_conf": 0.03126141704297812,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1688.0,
      "completions/max_terminated_length": 1688.0,
      "completions/mean_length": 371.45703125,
      "completions/mean_terminated_length": 372.91375732421875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 90.0,
      "epoch": 0.07253333333333334,
      "grad_norm": 0.18110527098178864,
      "learning_rate": 3.6666666666666666e-06,
      "loss": 0.0146,
      "num_tokens": 14283938.0,
      "reward": 1.0882487297058105,
      "reward_std": 0.241624116897583,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.5404953956604004,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 68
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.227063449240397e-06,
      "aux_brier/mean_group_std": 0.04779126202984317,
      "aux_brier/mean_r": 0.9652886564493792,
      "aux_brier/n_active_tok": 201.25,
      "aux_brier/n_groups": 13.1875,
      "aux_brier/n_step_records": 50.3125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4581994623303389,
      "calib/avg_num_step_conf": 6.2890625,
      "calib/ece": 0.38337301587301587,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004948527965379323,
      "calib/mean_conf": 0.017420634920634923,
      "calib/mu_c": 0.014455445544554454,
      "calib/mu_w": 0.019403973509933777,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0,
      "calib/std_conf": 0.029935341213440846,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2844.0,
      "completions/max_terminated_length": 2844.0,
      "completions/mean_length": 452.0546875,
      "completions/mean_terminated_length": 452.0546875,
      "completions/min_length": 75.0,
      "completions/min_terminated_length": 75.0,
      "epoch": 0.0736,
      "grad_norm": 0.014755764044821262,
      "learning_rate": 3.638888888888889e-06,
      "loss": 0.0802,
      "num_tokens": 14504160.0,
      "reward": 1.030881643295288,
      "reward_std": 0.258872926235199,
      "rewards/accuracy_reward_step": 0.39453125,
      "rewards/final_brier_reward_step": 0.5922762155532837,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 69
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 4.408656835153124e-06,
      "aux_brier/mean_group_std": 0.04061929032654197,
      "aux_brier/mean_r": 0.9700416407596402,
      "aux_brier/n_active_tok": 212.625,
      "aux_brier/n_groups": 15.0,
      "aux_brier/n_step_records": 53.15625,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.6369284954811271,
      "calib/avg_num_step_conf": 6.64453125,
      "calib/ece": 0.4481300813008131,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.018680223285486443,
      "calib/mean_conf": 0.025853658536585365,
      "calib/mu_c": 0.03587719298245614,
      "calib/mu_w": 0.017196969696969697,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.005284552845528456,
      "calib/std_conf": 0.08678472411820899,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2556.0,
      "completions/max_terminated_length": 2556.0,
      "completions/mean_length": 447.1796875,
      "completions/mean_terminated_length": 448.933349609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 88.0,
      "epoch": 0.07466666666666667,
      "grad_norm": 0.028785645961761475,
      "learning_rate": 3.6111111111111115e-06,
      "loss": 0.0586,
      "num_tokens": 14725630.0,
      "reward": 1.0579323768615723,
      "reward_std": 0.2267770767211914,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.5364171266555786,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 70
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 4.029969959545854e-06,
      "aux_brier/mean_group_std": 0.048995203603129825,
      "aux_brier/mean_r": 0.9659952706389444,
      "aux_brier/n_active_tok": 227.625,
      "aux_brier/n_groups": 14.6875,
      "aux_brier/n_step_records": 56.90625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.534349173553719,
      "calib/avg_num_step_conf": 7.46875,
      "calib/ece": 0.4715662650602409,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00507941632231405,
      "calib/mean_conf": 0.01598393574297189,
      "calib/mu_c": 0.01859504132231405,
      "calib/mu_w": 0.013515625,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0008032128514056224,
      "calib/std_conf": 0.04218832970889454,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1952.0,
      "completions/max_terminated_length": 1952.0,
      "completions/mean_length": 447.92578125,
      "completions/mean_terminated_length": 451.4527587890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.07573333333333333,
      "grad_norm": 0.346962034702301,
      "learning_rate": 3.5833333333333335e-06,
      "loss": 0.0354,
      "num_tokens": 14944707.0,
      "reward": 1.0917901992797852,
      "reward_std": 0.28287506103515625,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.5155984163284302,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 71
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.3916517893619584e-06,
      "aux_brier/mean_group_std": 0.033061452835828356,
      "aux_brier/mean_r": 0.976291224687943,
      "aux_brier/n_active_tok": 225.875,
      "aux_brier/n_groups": 16.25,
      "aux_brier/n_step_records": 56.46875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5433564652702616,
      "calib/avg_num_step_conf": 7.05859375,
      "calib/ece": 0.43376984126984125,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0021041573820589562,
      "calib/mean_conf": 0.016626984126984126,
      "calib/mu_c": 0.017787610619469027,
      "calib/mu_w": 0.01568345323741007,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.000992063492063492,
      "calib/std_conf": 0.029869619891987435,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2809.0,
      "completions/max_terminated_length": 2809.0,
      "completions/mean_length": 427.97265625,
      "completions/mean_terminated_length": 427.97265625,
      "completions/min_length": 122.0,
      "completions/min_terminated_length": 122.0,
      "epoch": 0.0768,
      "grad_norm": 0.05484065040946007,
      "learning_rate": 3.555555555555556e-06,
      "loss": 0.1103,
      "num_tokens": 15158676.0,
      "reward": 1.0729742050170898,
      "reward_std": 0.21148447692394257,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.5575214624404907,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 72
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -4.360505024725292e-06,
      "aux_brier/mean_group_std": 0.04531713227609749,
      "aux_brier/mean_r": 0.972327532595925,
      "aux_brier/n_active_tok": 198.75,
      "aux_brier/n_groups": 11.3125,
      "aux_brier/n_step_records": 49.6875,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5806263073704934,
      "calib/avg_num_step_conf": 6.25,
      "calib/ece": 0.4771764705882352,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005575858250276854,
      "calib/mean_conf": 0.016941176470588237,
      "calib/mu_c": 0.019761904761904762,
      "calib/mu_w": 0.014186046511627907,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.02482994062733726,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1189.0,
      "completions/max_terminated_length": 1189.0,
      "completions/mean_length": 391.87890625,
      "completions/mean_terminated_length": 393.41571044921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 136.0,
      "epoch": 0.07786666666666667,
      "grad_norm": 0.04808742180466652,
      "learning_rate": 3.5277777777777784e-06,
      "loss": 0.0091,
      "num_tokens": 15366029.0,
      "reward": 1.1169428825378418,
      "reward_std": 0.25438421964645386,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.5224593877792358,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 73
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.7314468178385454e-06,
      "aux_brier/mean_group_std": 0.03765971292555707,
      "aux_brier/mean_r": 0.9796891334346636,
      "aux_brier/n_active_tok": 221.125,
      "aux_brier/n_groups": 13.15625,
      "aux_brier/n_step_records": 55.28125,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5207613606165293,
      "calib/avg_num_step_conf": 6.93359375,
      "calib/ece": 0.40467741935483875,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0060855700239170855,
      "calib/mean_conf": 0.02274193548387097,
      "calib/mu_c": 0.026226415094339622,
      "calib/mu_w": 0.020140845070422537,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.036935836031565285,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2389.0,
      "completions/max_terminated_length": 2389.0,
      "completions/mean_length": 453.0078125,
      "completions/mean_terminated_length": 454.7843322753906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 109.0,
      "epoch": 0.07893333333333333,
      "grad_norm": 0.05028804391622543,
      "learning_rate": 3.5e-06,
      "loss": 0.08,
      "num_tokens": 15585927.0,
      "reward": 1.0399447679519653,
      "reward_std": 0.2993197441101074,
      "rewards/accuracy_reward_step": 0.4140625,
      "rewards/final_brier_reward_step": 0.573841392993927,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 74
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.2505642699187653e-06,
      "aux_brier/mean_group_std": 0.06429856926379476,
      "aux_brier/mean_r": 0.9510935457446775,
      "aux_brier/n_active_tok": 198.375,
      "aux_brier/n_groups": 11.40625,
      "aux_brier/n_step_records": 49.59375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.513403425706726,
      "calib/avg_num_step_conf": 6.19921875,
      "calib/ece": 0.6353359683794466,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00225316808243977,
      "calib/mean_conf": 0.02474308300395257,
      "calib/mu_c": 0.025508982035928142,
      "calib/mu_w": 0.023255813953488372,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.029762452165010417,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2159.0,
      "completions/max_terminated_length": 2159.0,
      "completions/mean_length": 395.53515625,
      "completions/mean_terminated_length": 395.53515625,
      "completions/min_length": 82.0,
      "completions/min_terminated_length": 82.0,
      "epoch": 0.08,
      "grad_norm": 0.007172565907239914,
      "learning_rate": 3.4722222222222224e-06,
      "loss": 0.0416,
      "num_tokens": 15791936.0,
      "reward": 1.2384189367294312,
      "reward_std": 0.2703310549259186,
      "rewards/accuracy_reward_step": 0.65234375,
      "rewards/final_brier_reward_step": 0.3677383065223694,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 75
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 7.346894645987589e-07,
      "aux_brier/mean_group_std": 0.05394622553427382,
      "aux_brier/mean_r": 0.9652903386673772,
      "aux_brier/n_active_tok": 205.625,
      "aux_brier/n_groups": 13.1875,
      "aux_brier/n_step_records": 51.40625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.47307063851181497,
      "calib/avg_num_step_conf": 6.42578125,
      "calib/ece": 0.5091699604743082,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0025377073906485623,
      "calib/mean_conf": 0.028379446640316205,
      "calib/mu_c": 0.027205882352941177,
      "calib/mu_w": 0.02974358974358974,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.023543381296148785,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3031.0,
      "completions/max_terminated_length": 3031.0,
      "completions/mean_length": 412.6171875,
      "completions/mean_terminated_length": 412.6171875,
      "completions/min_length": 58.0,
      "completions/min_terminated_length": 58.0,
      "epoch": 0.08106666666666666,
      "grad_norm": 0.006502797827124596,
      "learning_rate": 3.444444444444445e-06,
      "loss": 0.0562,
      "num_tokens": 16000622.0,
      "reward": 1.1465389728546143,
      "reward_std": 0.2662979066371918,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.4845937490463257,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 76
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.569284396319871e-06,
      "aux_brier/mean_group_std": 0.0676945337044517,
      "aux_brier/mean_r": 0.9563684466021387,
      "aux_brier/n_active_tok": 205.75,
      "aux_brier/n_groups": 14.0625,
      "aux_brier/n_step_records": 51.4375,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5506756756756757,
      "calib/avg_num_step_conf": 6.4296875,
      "calib/ece": 0.5623199999999999,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0032684154742978304,
      "calib/mean_conf": 0.029679999999999998,
      "calib/mu_c": 0.031013513513513512,
      "calib/mu_w": 0.027745098039215682,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0,
      "calib/std_conf": 0.025375137438051445,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2751.0,
      "completions/max_terminated_length": 2751.0,
      "completions/mean_length": 425.17578125,
      "completions/mean_terminated_length": 426.8431701660156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 94.0,
      "epoch": 0.08213333333333334,
      "grad_norm": 0.01168504822999239,
      "learning_rate": 3.416666666666667e-06,
      "loss": 0.0847,
      "num_tokens": 16214131.0,
      "reward": 1.172635555267334,
      "reward_std": 0.28494516015052795,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.4327300786972046,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 77
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 5.763075152098729e-07,
      "aux_brier/mean_group_std": 0.05472211652629763,
      "aux_brier/mean_r": 0.9655651246092493,
      "aux_brier/n_active_tok": 220.125,
      "aux_brier/n_groups": 14.40625,
      "aux_brier/n_step_records": 55.03125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5013574660633484,
      "calib/avg_num_step_conf": 6.91796875,
      "calib/ece": 0.4926746987951807,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.007744214608920497,
      "calib/mean_conf": 0.03857028112449799,
      "calib/mu_c": 0.034869230769230763,
      "calib/mu_w": 0.04261344537815126,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.004578313253012048,
      "calib/std_conf": 0.054301738420637125,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2512.0,
      "completions/max_terminated_length": 2512.0,
      "completions/mean_length": 497.328125,
      "completions/mean_terminated_length": 497.328125,
      "completions/min_length": 163.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.0832,
      "grad_norm": 0.006679881364107132,
      "learning_rate": 3.3888888888888893e-06,
      "loss": 0.098,
      "num_tokens": 16449471.0,
      "reward": 1.1172987222671509,
      "reward_std": 0.3159688115119934,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.4926322102546692,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 78
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 7.383033536478845e-07,
      "aux_brier/mean_group_std": 0.05742701776696072,
      "aux_brier/mean_r": 0.9586039216655239,
      "aux_brier/n_active_tok": 224.5,
      "aux_brier/n_groups": 14.1875,
      "aux_brier/n_step_records": 56.125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5595282089184528,
      "calib/avg_num_step_conf": 7.01953125,
      "calib/ece": 0.48423529411764704,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.00392156862745098,
      "calib/gap": 0.0005728011825572865,
      "calib/mean_conf": 0.04078431372549019,
      "calib/mu_c": 0.04106060606060606,
      "calib/mu_w": 0.040487804878048775,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003686274509803921,
      "calib/std_conf": 0.06920396501320307,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2365.0,
      "completions/max_terminated_length": 2365.0,
      "completions/mean_length": 465.0546875,
      "completions/mean_terminated_length": 465.0546875,
      "completions/min_length": 87.0,
      "completions/min_terminated_length": 87.0,
      "epoch": 0.08426666666666667,
      "grad_norm": 0.006277347914874554,
      "learning_rate": 3.3611111111111117e-06,
      "loss": 0.0475,
      "num_tokens": 16674901.0,
      "reward": 1.1378858089447021,
      "reward_std": 0.2390265166759491,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.5124804973602295,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 79
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.460258663630979e-06,
      "aux_brier/mean_group_std": 0.06699746246123675,
      "aux_brier/mean_r": 0.9554511037314575,
      "aux_brier/n_active_tok": 219.75,
      "aux_brier/n_groups": 13.09375,
      "aux_brier/n_step_records": 54.9375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5270473574045003,
      "calib/avg_num_step_conf": 6.9765625,
      "calib/ece": 0.5416334661354582,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0019453165881737344,
      "calib/mean_conf": 0.04402390438247012,
      "calib/mu_c": 0.04482993197278912,
      "calib/mu_w": 0.04288461538461538,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.029527207830841,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2527.0,
      "completions/max_terminated_length": 2527.0,
      "completions/mean_length": 441.390625,
      "completions/mean_terminated_length": 444.86614990234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.08533333333333333,
      "grad_norm": 0.12150449305772781,
      "learning_rate": 3.3333333333333333e-06,
      "loss": -0.005,
      "num_tokens": 16890057.0,
      "reward": 1.1750924587249756,
      "reward_std": 0.2941782474517822,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.4503699243068695,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 80
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.3112985259032826e-06,
      "aux_brier/mean_group_std": 0.05724071403940858,
      "aux_brier/mean_r": 0.9653533296743093,
      "aux_brier/n_active_tok": 222.875,
      "aux_brier/n_groups": 15.84375,
      "aux_brier/n_step_records": 55.71875,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.4776960784313725,
      "calib/avg_num_step_conf": 7.1328125,
      "calib/ece": 0.5206970954356845,
      "calib/final_conf_rate": 0.94140625,
      "calib/format_rate": 0.9296875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00030238095238095675,
      "calib/mean_conf": 0.04361825726141079,
      "calib/mu_c": 0.043750000000000004,
      "calib/mu_w": 0.04344761904761905,
      "calib/nonempty_final_conf_rate": 0.94140625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.0361915649871068,
      "calib/step_conf_rate": 0.98046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2729.0,
      "completions/max_terminated_length": 2729.0,
      "completions/mean_length": 470.4609375,
      "completions/mean_terminated_length": 474.16534423828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 126.0,
      "epoch": 0.0864,
      "grad_norm": 0.022036930546164513,
      "learning_rate": 3.3055555555555558e-06,
      "loss": 0.0753,
      "num_tokens": 17116743.0,
      "reward": 1.110398769378662,
      "reward_std": 0.315701425075531,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.4494074881076813,
      "rewards/format_reward_step_strict": 0.92578125,
      "step": 81
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.6606594932522256e-08,
      "aux_brier/mean_group_std": 0.057497007152077374,
      "aux_brier/mean_r": 0.963235195565649,
      "aux_brier/n_active_tok": 194.625,
      "aux_brier/n_groups": 11.71875,
      "aux_brier/n_step_records": 48.65625,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.4693302387267905,
      "calib/avg_num_step_conf": 6.18359375,
      "calib/ece": 0.5404016064257028,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0028474801061007943,
      "calib/mean_conf": 0.04353413654618474,
      "calib/mu_c": 0.0423448275862069,
      "calib/mu_w": 0.04519230769230769,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.0008032128514056224,
      "calib/std_conf": 0.025085393103278526,
      "calib/step_conf_rate": 0.9765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2273.0,
      "completions/max_terminated_length": 2273.0,
      "completions/mean_length": 408.19921875,
      "completions/mean_terminated_length": 413.03955078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 73.0,
      "epoch": 0.08746666666666666,
      "grad_norm": 0.1904134601354599,
      "learning_rate": 3.277777777777778e-06,
      "loss": 0.0472,
      "num_tokens": 17326794.0,
      "reward": 1.151876449584961,
      "reward_std": 0.29784828424453735,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.4356308579444885,
      "rewards/format_reward_step_strict": 0.953125,
      "step": 82
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.218257767805376e-06,
      "aux_brier/mean_group_std": 0.06747656495244791,
      "aux_brier/mean_r": 0.9456673557669775,
      "aux_brier/n_active_tok": 245.5,
      "aux_brier/n_groups": 19.6875,
      "aux_brier/n_step_records": 61.375,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5392962072649573,
      "calib/avg_num_step_conf": 7.82421875,
      "calib/ece": 0.48138775510204085,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.9375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0004119925213675324,
      "calib/mean_conf": 0.042693877551020415,
      "calib/mu_c": 0.04289062500000001,
      "calib/mu_w": 0.042478632478632476,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.000816326530612245,
      "calib/std_conf": 0.027878895596104156,
      "calib/step_conf_rate": 0.97265625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2879.0,
      "completions/max_terminated_length": 2879.0,
      "completions/mean_length": 536.4609375,
      "completions/mean_terminated_length": 542.8221435546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 116.0,
      "epoch": 0.08853333333333334,
      "grad_norm": 0.13237819075584412,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.0673,
      "num_tokens": 17571392.0,
      "reward": 1.0950324535369873,
      "reward_std": 0.2291625440120697,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.4895046651363373,
      "rewards/format_reward_step_strict": 0.9375,
      "step": 83
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 5.204469538666601e-07,
      "aux_brier/mean_group_std": 0.0631882685984794,
      "aux_brier/mean_r": 0.9567116047566644,
      "aux_brier/n_active_tok": 198.375,
      "aux_brier/n_groups": 12.09375,
      "aux_brier/n_step_records": 49.59375,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5419250645994833,
      "calib/avg_num_step_conf": 6.34375,
      "calib/ece": 0.4406024096385542,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0035542635658914815,
      "calib/mean_conf": 0.04132530120481928,
      "calib/mu_c": 0.04316666666666667,
      "calib/mu_w": 0.03961240310077519,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.02680156521551764,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2960.0,
      "completions/max_terminated_length": 2960.0,
      "completions/mean_length": 428.19921875,
      "completions/mean_terminated_length": 431.57086181640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 64.0,
      "epoch": 0.0896,
      "grad_norm": 0.0071330503560602665,
      "learning_rate": 3.2222222222222227e-06,
      "loss": 0.0068,
      "num_tokens": 17786931.0,
      "reward": 1.0847227573394775,
      "reward_std": 0.23574988543987274,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.5342034697532654,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 84
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.4682940519559224e-08,
      "aux_brier/mean_group_std": 0.07396974394006378,
      "aux_brier/mean_r": 0.9495267329428589,
      "aux_brier/n_active_tok": 202.75,
      "aux_brier/n_groups": 12.46875,
      "aux_brier/n_step_records": 50.6875,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5214947089947091,
      "calib/avg_num_step_conf": 6.3359375,
      "calib/ece": 0.4397967479674797,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0009603174603174669,
      "calib/mean_conf": 0.04800813008130081,
      "calib/mu_c": 0.0485,
      "calib/mu_w": 0.047539682539682535,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0,
      "calib/std_conf": 0.028720972177920225,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2823.0,
      "completions/max_terminated_length": 2823.0,
      "completions/mean_length": 485.11328125,
      "completions/mean_terminated_length": 488.9330749511719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 122.0,
      "epoch": 0.09066666666666667,
      "grad_norm": 0.011264825239777565,
      "learning_rate": 3.1944444444444443e-06,
      "loss": 0.0655,
      "num_tokens": 18018944.0,
      "reward": 1.0750786066055298,
      "reward_std": 0.27728450298309326,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5268769264221191,
      "rewards/format_reward_step_strict": 0.94921875,
      "step": 85
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.8439633884061455e-06,
      "aux_brier/mean_group_std": 0.061103717383855255,
      "aux_brier/mean_r": 0.960463549216192,
      "aux_brier/n_active_tok": 206.625,
      "aux_brier/n_groups": 13.1875,
      "aux_brier/n_step_records": 51.65625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5642167314687068,
      "calib/avg_num_step_conf": 6.484375,
      "calib/ece": 0.42150197628458497,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0011526401605418402,
      "calib/mean_conf": 0.05501976284584981,
      "calib/mu_c": 0.05563025210084034,
      "calib/mu_w": 0.0544776119402985,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0030830039525691698,
      "calib/std_conf": 0.0402818578013471,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2693.0,
      "completions/max_terminated_length": 2693.0,
      "completions/mean_length": 455.29296875,
      "completions/mean_terminated_length": 457.0784606933594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 100.0,
      "epoch": 0.09173333333333333,
      "grad_norm": 0.04765058308839798,
      "learning_rate": 3.1666666666666667e-06,
      "loss": 0.01,
      "num_tokens": 18241011.0,
      "reward": 1.0938220024108887,
      "reward_std": 0.24476972222328186,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5627878904342651,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 86
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.4273639888928713e-06,
      "aux_brier/mean_group_std": 0.07217300149636746,
      "aux_brier/mean_r": 0.9548417209835198,
      "aux_brier/n_active_tok": 210.0,
      "aux_brier/n_groups": 15.28125,
      "aux_brier/n_step_records": 52.5,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5291983657368273,
      "calib/avg_num_step_conf": 6.8515625,
      "calib/ece": 0.5777327935222673,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.004048582995951417,
      "calib/gap": 0.007298534798534803,
      "calib/mean_conf": 0.05570850202429149,
      "calib/mu_c": 0.058397435897435894,
      "calib/mu_w": 0.05109890109890109,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.000931174089068826,
      "calib/std_conf": 0.06676598239708885,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2618.0,
      "completions/max_terminated_length": 2618.0,
      "completions/mean_length": 445.9609375,
      "completions/mean_terminated_length": 453.0397033691406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 88.0,
      "epoch": 0.0928,
      "grad_norm": 0.07679977267980576,
      "learning_rate": 3.138888888888889e-06,
      "loss": 0.0397,
      "num_tokens": 18460673.0,
      "reward": 1.1927268505096436,
      "reward_std": 0.2760070264339447,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.41934531927108765,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 87
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.4337857021938802e-06,
      "aux_brier/mean_group_std": 0.06245186857284721,
      "aux_brier/mean_r": 0.9612910237050545,
      "aux_brier/n_active_tok": 230.625,
      "aux_brier/n_groups": 14.65625,
      "aux_brier/n_step_records": 57.65625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.576890945311998,
      "calib/avg_num_step_conf": 7.21484375,
      "calib/ece": 0.47091999999999995,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00030139451192083555,
      "calib/mean_conf": 0.06828000000000001,
      "calib/mu_c": 0.06842105263157895,
      "calib/mu_w": 0.06811965811965812,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0036000000000000003,
      "calib/std_conf": 0.06230763677110535,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2893.0,
      "completions/max_terminated_length": 2893.0,
      "completions/mean_length": 505.69140625,
      "completions/mean_terminated_length": 505.69140625,
      "completions/min_length": 145.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.09386666666666667,
      "grad_norm": 0.006025264505296946,
      "learning_rate": 3.1111111111111116e-06,
      "loss": 0.0901,
      "num_tokens": 18699978.0,
      "reward": 1.1336658000946045,
      "reward_std": 0.24178946018218994,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.5190386772155762,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 88
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.287836486733699e-06,
      "aux_brier/mean_group_std": 0.062280394224279924,
      "aux_brier/mean_r": 0.9556622252107899,
      "aux_brier/n_active_tok": 206.25,
      "aux_brier/n_groups": 15.34375,
      "aux_brier/n_step_records": 51.5625,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.4993795715778474,
      "calib/avg_num_step_conf": 6.4609375,
      "calib/ece": 0.47518145161290326,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.003880616509926868,
      "calib/mean_conf": 0.059012096774193556,
      "calib/mu_c": 0.05719696969696969,
      "calib/mu_w": 0.06107758620689656,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.9609375,
      "calib/pce": 0.0009677419354838709,
      "calib/std_conf": 0.03091239988372433,
      "calib/step_conf_rate": 0.9609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2629.0,
      "completions/max_terminated_length": 2629.0,
      "completions/mean_length": 488.9765625,
      "completions/mean_terminated_length": 490.8941345214844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.09493333333333333,
      "grad_norm": 0.006578805856406689,
      "learning_rate": 3.0833333333333336e-06,
      "loss": 0.0908,
      "num_tokens": 18934044.0,
      "reward": 1.1073076725006104,
      "reward_std": 0.2533990442752838,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.4917307496070862,
      "rewards/format_reward_step_strict": 0.9375,
      "step": 89
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 6.406338775022391e-07,
      "aux_brier/mean_group_std": 0.08968575593167973,
      "aux_brier/mean_r": 0.9227656757281388,
      "aux_brier/n_active_tok": 233.25,
      "aux_brier/n_groups": 15.5,
      "aux_brier/n_step_records": 58.3125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.48957646127457455,
      "calib/avg_num_step_conf": 7.2890625,
      "calib/ece": 0.5078313253012049,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 8.840216387386024e-05,
      "calib/mean_conf": 0.06646586345381526,
      "calib/mu_c": 0.0665034965034965,
      "calib/mu_w": 0.06641509433962264,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.03282483386447449,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2614.0,
      "completions/max_terminated_length": 2614.0,
      "completions/mean_length": 481.171875,
      "completions/mean_terminated_length": 481.171875,
      "completions/min_length": 100.0,
      "completions/min_terminated_length": 100.0,
      "epoch": 0.096,
      "grad_norm": 0.006786589976400137,
      "learning_rate": 3.055555555555556e-06,
      "loss": 0.0124,
      "num_tokens": 19160544.0,
      "reward": 1.165675401687622,
      "reward_std": 0.2502981722354889,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.48301443457603455,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 90
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.2981149173363349e-06,
      "aux_brier/mean_group_std": 0.05019438075960755,
      "aux_brier/mean_r": 0.970150317796923,
      "aux_brier/n_active_tok": 221.0,
      "aux_brier/n_groups": 14.6875,
      "aux_brier/n_step_records": 55.25,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.48671215074723845,
      "calib/avg_num_step_conf": 6.90625,
      "calib/ece": 0.472289156626506,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -6.042884990255237e-05,
      "calib/mean_conf": 0.06987951807228916,
      "calib/mu_c": 0.06985185185185185,
      "calib/mu_w": 0.0699122807017544,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.027249876425950875,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2833.0,
      "completions/max_terminated_length": 2833.0,
      "completions/mean_length": 488.1796875,
      "completions/mean_terminated_length": 490.0941467285156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 203.0,
      "epoch": 0.09706666666666666,
      "grad_norm": 0.009234710596501827,
      "learning_rate": 3.0277777777777776e-06,
      "loss": 0.0621,
      "num_tokens": 19393230.0,
      "reward": 1.130971908569336,
      "reward_std": 0.26993629336357117,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.5004504323005676,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 91
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.5199346654615198e-06,
      "aux_brier/mean_group_std": 0.0317642007873023,
      "aux_brier/mean_r": 0.9793554388894053,
      "aux_brier/n_active_tok": 214.0,
      "aux_brier/n_groups": 13.71875,
      "aux_brier/n_step_records": 53.5,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6035353535353535,
      "calib/avg_num_step_conf": 6.71875,
      "calib/ece": 0.5366798418972333,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.013600288600288613,
      "calib/mean_conf": 0.07201581027667983,
      "calib/mu_c": 0.07733766233766234,
      "calib/mu_w": 0.06373737373737373,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04199905592571995,
      "calib/step_conf_rate": 0.98046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1933.0,
      "completions/max_terminated_length": 1933.0,
      "completions/mean_length": 435.40234375,
      "completions/mean_terminated_length": 435.40234375,
      "completions/min_length": 132.0,
      "completions/min_terminated_length": 132.0,
      "epoch": 0.09813333333333334,
      "grad_norm": 0.007359606679528952,
      "learning_rate": 3e-06,
      "loss": 0.009,
      "num_tokens": 19611413.0,
      "reward": 1.2064805030822754,
      "reward_std": 0.22201576828956604,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.46654728055000305,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 92
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.9570681371039456e-06,
      "aux_brier/mean_group_std": 0.04964086610747109,
      "aux_brier/mean_r": 0.9658705880275835,
      "aux_brier/n_active_tok": 250.5,
      "aux_brier/n_groups": 16.71875,
      "aux_brier/n_step_records": 62.625,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.40741106719367587,
      "calib/avg_num_step_conf": 7.8359375,
      "calib/ece": 0.4591902834008097,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.004048582995951417,
      "calib/gap": -0.020566534914361,
      "calib/mean_conf": 0.08987854251012146,
      "calib/mu_c": 0.08030303030303032,
      "calib/mu_w": 0.10086956521739132,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.0073279352226720655,
      "calib/std_conf": 0.0665243663701703,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2702.0,
      "completions/max_terminated_length": 2702.0,
      "completions/mean_length": 524.30078125,
      "completions/mean_terminated_length": 524.30078125,
      "completions/min_length": 135.0,
      "completions/min_terminated_length": 135.0,
      "epoch": 0.0992,
      "grad_norm": 0.005457157269120216,
      "learning_rate": 2.9722222222222225e-06,
      "loss": 0.04,
      "num_tokens": 19851410.0,
      "reward": 1.124093770980835,
      "reward_std": 0.25590553879737854,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.5198125243186951,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 93
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.2300667046327707e-06,
      "aux_brier/mean_group_std": 0.06045114795667379,
      "aux_brier/mean_r": 0.9547563204393912,
      "aux_brier/n_active_tok": 214.75,
      "aux_brier/n_groups": 14.71875,
      "aux_brier/n_step_records": 53.6875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5974820143884892,
      "calib/avg_num_step_conf": 6.7109375,
      "calib/ece": 0.48429718875502015,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009275343361674293,
      "calib/mean_conf": 0.07899598393574297,
      "calib/mu_c": 0.08309352517985612,
      "calib/mu_w": 0.07381818181818182,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0025301204819277107,
      "calib/std_conf": 0.04541857960603195,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2859.0,
      "completions/max_terminated_length": 2859.0,
      "completions/mean_length": 497.46484375,
      "completions/mean_terminated_length": 497.46484375,
      "completions/min_length": 127.0,
      "completions/min_terminated_length": 127.0,
      "epoch": 0.10026666666666667,
      "grad_norm": 0.006252513267099857,
      "learning_rate": 2.944444444444445e-06,
      "loss": 0.0681,
      "num_tokens": 20087441.0,
      "reward": 1.1524341106414795,
      "reward_std": 0.21631355583667755,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.5081738233566284,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 94
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.0933209314853993e-07,
      "aux_brier/mean_group_std": 0.07138223201401671,
      "aux_brier/mean_r": 0.9512899567899513,
      "aux_brier/n_active_tok": 218.75,
      "aux_brier/n_groups": 14.5,
      "aux_brier/n_step_records": 54.6875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5237669645812644,
      "calib/avg_num_step_conf": 6.8359375,
      "calib/ece": 0.5448425196850394,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003000993048659417,
      "calib/mean_conf": 0.08114173228346458,
      "calib/mu_c": 0.08226415094339624,
      "calib/mu_w": 0.07926315789473683,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.036064732791605464,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2831.0,
      "completions/max_terminated_length": 2831.0,
      "completions/mean_length": 467.0234375,
      "completions/mean_terminated_length": 467.0234375,
      "completions/min_length": 89.0,
      "completions/min_terminated_length": 89.0,
      "epoch": 0.10133333333333333,
      "grad_norm": 0.00672187888994813,
      "learning_rate": 2.916666666666667e-06,
      "loss": 0.0163,
      "num_tokens": 20313127.0,
      "reward": 1.228598952293396,
      "reward_std": 0.22712072730064392,
      "rewards/accuracy_reward_step": 0.62109375,
      "rewards/final_brier_reward_step": 0.46127110719680786,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 95
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.233655969074988e-06,
      "aux_brier/mean_group_std": 0.06829223796434122,
      "aux_brier/mean_r": 0.947683284225488,
      "aux_brier/n_active_tok": 212.625,
      "aux_brier/n_groups": 13.375,
      "aux_brier/n_step_records": 53.15625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.48457300275482096,
      "calib/avg_num_step_conf": 6.64453125,
      "calib/ece": 0.5717391304347826,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.005257575757575753,
      "calib/mean_conf": 0.08043478260869566,
      "calib/mu_c": 0.07860606060606061,
      "calib/mu_w": 0.08386363636363636,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.034242345667469226,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2602.0,
      "completions/max_terminated_length": 2602.0,
      "completions/mean_length": 431.0234375,
      "completions/mean_terminated_length": 431.0234375,
      "completions/min_length": 141.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.1024,
      "grad_norm": 0.006833321414887905,
      "learning_rate": 2.888888888888889e-06,
      "loss": 0.0386,
      "num_tokens": 20529285.0,
      "reward": 1.245131492614746,
      "reward_std": 0.22330452501773834,
      "rewards/accuracy_reward_step": 0.64453125,
      "rewards/final_brier_reward_step": 0.43365079164505005,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 96
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.4035530246925276e-06,
      "aux_brier/mean_group_std": 0.034228858761962536,
      "aux_brier/mean_r": 0.9738376620580753,
      "aux_brier/n_active_tok": 231.5,
      "aux_brier/n_groups": 15.34375,
      "aux_brier/n_step_records": 57.875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5391290527654163,
      "calib/avg_num_step_conf": 7.23828125,
      "calib/ece": 0.4316733067729083,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0016357279084551896,
      "calib/mean_conf": 0.08944223107569722,
      "calib/mu_c": 0.09023076923076924,
      "calib/mu_w": 0.08859504132231405,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0015936254980079682,
      "calib/std_conf": 0.04328732469770397,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2969.0,
      "completions/max_terminated_length": 2969.0,
      "completions/mean_length": 476.63671875,
      "completions/mean_terminated_length": 476.63671875,
      "completions/min_length": 92.0,
      "completions/min_terminated_length": 92.0,
      "epoch": 0.10346666666666667,
      "grad_norm": 0.006699815392494202,
      "learning_rate": 2.861111111111111e-06,
      "loss": 0.0761,
      "num_tokens": 20756376.0,
      "reward": 1.1296970844268799,
      "reward_std": 0.26899388432502747,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5500386953353882,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 97
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -4.5524656547279996e-07,
      "aux_brier/mean_group_std": 0.08662462380554938,
      "aux_brier/mean_r": 0.936917259113745,
      "aux_brier/n_active_tok": 211.875,
      "aux_brier/n_groups": 12.6875,
      "aux_brier/n_step_records": 52.96875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5064566115702479,
      "calib/avg_num_step_conf": 6.62109375,
      "calib/ece": 0.43232931726907625,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0014811466942148688,
      "calib/mean_conf": 0.08775100401606427,
      "calib/mu_c": 0.08703125000000002,
      "calib/mu_w": 0.08851239669421489,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0030120481927710845,
      "calib/std_conf": 0.03406470115021999,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 479.90625,
      "completions/mean_terminated_length": 479.90625,
      "completions/min_length": 58.0,
      "completions/min_terminated_length": 58.0,
      "epoch": 0.10453333333333334,
      "grad_norm": 0.006359133403748274,
      "learning_rate": 2.8333333333333335e-06,
      "loss": 0.0563,
      "num_tokens": 20985416.0,
      "reward": 1.1240954399108887,
      "reward_std": 0.2709805369377136,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.5510691404342651,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 98
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.861180691844268e-07,
      "aux_brier/mean_group_std": 0.05330940022385636,
      "aux_brier/mean_r": 0.967148880055956,
      "aux_brier/n_active_tok": 247.5,
      "aux_brier/n_groups": 16.46875,
      "aux_brier/n_step_records": 61.875,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5039298245614036,
      "calib/avg_num_step_conf": 7.8828125,
      "calib/ece": 0.2914285714285714,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0027333333333333515,
      "calib/mean_conf": 0.0963265306122449,
      "calib/mu_c": 0.09800000000000002,
      "calib/mu_w": 0.09526666666666667,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.040329959368405346,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2812.0,
      "completions/max_terminated_length": 2812.0,
      "completions/mean_length": 539.828125,
      "completions/mean_terminated_length": 548.3968505859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 122.0,
      "epoch": 0.1056,
      "grad_norm": 0.03724941611289978,
      "learning_rate": 2.805555555555556e-06,
      "loss": 0.0629,
      "num_tokens": 21229412.0,
      "reward": 1.0067994594573975,
      "reward_std": 0.3061428666114807,
      "rewards/accuracy_reward_step": 0.37109375,
      "rewards/final_brier_reward_step": 0.6443851590156555,
      "rewards/format_reward_step_strict": 0.94921875,
      "step": 99
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.1245468185410177e-06,
      "aux_brier/mean_group_std": 0.07750098171818107,
      "aux_brier/mean_r": 0.9422343609838857,
      "aux_brier/n_active_tok": 244.75,
      "aux_brier/n_groups": 18.90625,
      "aux_brier/n_step_records": 61.1875,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.557468220338983,
      "calib/avg_num_step_conf": 7.6484375,
      "calib/ece": 0.39268292682926825,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005009269067796598,
      "calib/mean_conf": 0.08951219512195123,
      "calib/mu_c": 0.0921186440677966,
      "calib/mu_w": 0.087109375,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0012601626016260164,
      "calib/std_conf": 0.034871509990285446,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3013.0,
      "completions/max_terminated_length": 3013.0,
      "completions/mean_length": 536.05859375,
      "completions/mean_terminated_length": 536.05859375,
      "completions/min_length": 114.0,
      "completions/min_terminated_length": 114.0,
      "epoch": 0.10666666666666667,
      "grad_norm": 0.006149076856672764,
      "learning_rate": 2.7777777777777783e-06,
      "loss": 0.2026,
      "num_tokens": 21474051.0,
      "reward": 1.0854196548461914,
      "reward_std": 0.27186161279678345,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.5760538578033447,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 100
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.7394725093655694e-06,
      "aux_brier/mean_group_std": 0.04651070172327584,
      "aux_brier/mean_r": 0.9688452527490977,
      "aux_brier/n_active_tok": 271.375,
      "aux_brier/n_groups": 18.25,
      "aux_brier/n_step_records": 67.84375,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5565034122842231,
      "calib/avg_num_step_conf": 8.5546875,
      "calib/ece": 0.3412550607287449,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 2.1410410812258696e-05,
      "calib/mean_conf": 0.10093117408906883,
      "calib/mu_c": 0.1009433962264151,
      "calib/mu_w": 0.10092198581560284,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.006518218623481782,
      "calib/std_conf": 0.048486645348632076,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2964.0,
      "completions/max_terminated_length": 2964.0,
      "completions/mean_length": 529.94921875,
      "completions/mean_terminated_length": 532.0274658203125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 107.0,
      "epoch": 0.10773333333333333,
      "grad_norm": 0.015220129862427711,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.1063,
      "num_tokens": 21716710.0,
      "reward": 1.0520538091659546,
      "reward_std": 0.25686049461364746,
      "rewards/accuracy_reward_step": 0.4140625,
      "rewards/final_brier_reward_step": 0.6222777366638184,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 101
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.7109194594411292e-07,
      "aux_brier/mean_group_std": 0.057648837572275936,
      "aux_brier/mean_r": 0.9555693066286415,
      "aux_brier/n_active_tok": 236.25,
      "aux_brier/n_groups": 15.1875,
      "aux_brier/n_step_records": 59.0625,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.47846601390662685,
      "calib/avg_num_step_conf": 7.6953125,
      "calib/ece": 0.5576706827309238,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0022137079608344212,
      "calib/mean_conf": 0.0953413654618474,
      "calib/mu_c": 0.0945679012345679,
      "calib/mu_w": 0.09678160919540232,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0012048192771084336,
      "calib/std_conf": 0.038495604711073925,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2461.0,
      "completions/max_terminated_length": 2461.0,
      "completions/mean_length": 433.44140625,
      "completions/mean_terminated_length": 438.5810546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 52.0,
      "epoch": 0.1088,
      "grad_norm": 0.16762180626392365,
      "learning_rate": 2.7222222222222224e-06,
      "loss": -0.0229,
      "num_tokens": 21934367.0,
      "reward": 1.2294995784759521,
      "reward_std": 0.21669785678386688,
      "rewards/accuracy_reward_step": 0.6328125,
      "rewards/final_brier_reward_step": 0.44924843311309814,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 102
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.797633503748287e-07,
      "aux_brier/mean_group_std": 0.05695505996071269,
      "aux_brier/mean_r": 0.961754498336844,
      "aux_brier/n_active_tok": 238.75,
      "aux_brier/n_groups": 14.90625,
      "aux_brier/n_step_records": 59.6875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.48091553836234685,
      "calib/avg_num_step_conf": 7.4609375,
      "calib/ece": 0.4811952191235059,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.011633784655061277,
      "calib/mean_conf": 0.09928286852589642,
      "calib/mu_c": 0.09418439716312055,
      "calib/mu_w": 0.10581818181818183,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.009362549800796814,
      "calib/std_conf": 0.04089004380785392,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2426.0,
      "completions/max_terminated_length": 2426.0,
      "completions/mean_length": 501.71484375,
      "completions/mean_terminated_length": 505.66534423828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 130.0,
      "epoch": 0.10986666666666667,
      "grad_norm": 0.01944098249077797,
      "learning_rate": 2.6944444444444444e-06,
      "loss": 0.0345,
      "num_tokens": 22167358.0,
      "reward": 1.1715490818023682,
      "reward_std": 0.21027979254722595,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.5221335887908936,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 103
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.3577439715594117e-06,
      "aux_brier/mean_group_std": 0.047835558453891036,
      "aux_brier/mean_r": 0.9681153416870975,
      "aux_brier/n_active_tok": 240.5,
      "aux_brier/n_groups": 14.0,
      "aux_brier/n_step_records": 60.125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.618741196055833,
      "calib/avg_num_step_conf": 7.62890625,
      "calib/ece": 0.35426294820717125,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.013605455243949321,
      "calib/mean_conf": 0.10231075697211155,
      "calib/mu_c": 0.10973684210526319,
      "calib/mu_w": 0.09613138686131387,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.001195219123505976,
      "calib/std_conf": 0.041479595183155624,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2482.0,
      "completions/max_terminated_length": 2482.0,
      "completions/mean_length": 482.046875,
      "completions/mean_terminated_length": 483.9372863769531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 114.0,
      "epoch": 0.11093333333333333,
      "grad_norm": 0.05234473571181297,
      "learning_rate": 2.666666666666667e-06,
      "loss": 0.0765,
      "num_tokens": 22397442.0,
      "reward": 1.0830793380737305,
      "reward_std": 0.25311779975891113,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.613567590713501,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 104
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.8971290646763173e-06,
      "aux_brier/mean_group_std": 0.04680792486013509,
      "aux_brier/mean_r": 0.9673522841144343,
      "aux_brier/n_active_tok": 258.0,
      "aux_brier/n_groups": 17.8125,
      "aux_brier/n_step_records": 64.5,
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.4296706989247312,
      "calib/avg_num_step_conf": 8.19921875,
      "calib/ece": 0.392172131147541,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.009276881720430105,
      "calib/mean_conf": 0.09963114754098362,
      "calib/mu_c": 0.09491666666666669,
      "calib/mu_w": 0.1041935483870968,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.0401465962911567,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2904.0,
      "completions/max_terminated_length": 2904.0,
      "completions/mean_length": 526.13671875,
      "completions/mean_terminated_length": 528.2000122070312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 137.0,
      "epoch": 0.112,
      "grad_norm": 0.008446063846349716,
      "learning_rate": 2.6388888888888893e-06,
      "loss": 0.1038,
      "num_tokens": 22637893.0,
      "reward": 1.0859030485153198,
      "reward_std": 0.3250919580459595,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5623620748519897,
      "rewards/format_reward_step_strict": 0.953125,
      "step": 105
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.0323029602531406e-06,
      "aux_brier/mean_group_std": 0.0635028179003855,
      "aux_brier/mean_r": 0.9553198005022946,
      "aux_brier/n_active_tok": 238.375,
      "aux_brier/n_groups": 14.75,
      "aux_brier/n_step_records": 59.59375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.49169165417291355,
      "calib/avg_num_step_conf": 7.44921875,
      "calib/ece": 0.359251968503937,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0051586706646676445,
      "calib/mean_conf": 0.10271653543307087,
      "calib/mu_c": 0.0999137931034483,
      "calib/mu_w": 0.10507246376811595,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.002637795275590551,
      "calib/std_conf": 0.041289528450326476,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3003.0,
      "completions/max_terminated_length": 3003.0,
      "completions/mean_length": 467.91796875,
      "completions/mean_terminated_length": 467.91796875,
      "completions/min_length": 138.0,
      "completions/min_terminated_length": 138.0,
      "epoch": 0.11306666666666666,
      "grad_norm": 0.007161939051002264,
      "learning_rate": 2.6111111111111113e-06,
      "loss": 0.0688,
      "num_tokens": 22862264.0,
      "reward": 1.103581190109253,
      "reward_std": 0.21279966831207275,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6174496412277222,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 106
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.6140966701569965e-06,
      "aux_brier/mean_group_std": 0.05417005835351656,
      "aux_brier/mean_r": 0.9619935437258145,
      "aux_brier/n_active_tok": 244.375,
      "aux_brier/n_groups": 14.0,
      "aux_brier/n_step_records": 61.09375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5812308868501529,
      "calib/avg_num_step_conf": 7.63671875,
      "calib/ece": 0.4644268774703557,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.010589959225280304,
      "calib/mean_conf": 0.10474308300395258,
      "calib/mu_c": 0.10930555555555554,
      "calib/mu_w": 0.09871559633027524,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.035302660907854076,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2515.0,
      "completions/max_terminated_length": 2515.0,
      "completions/mean_length": 463.1796875,
      "completions/mean_terminated_length": 464.99609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 127.0,
      "epoch": 0.11413333333333334,
      "grad_norm": 0.00802396610379219,
      "learning_rate": 2.5833333333333337e-06,
      "loss": 0.0051,
      "num_tokens": 23085454.0,
      "reward": 1.1886709928512573,
      "reward_std": 0.2658335566520691,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.5359336137771606,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 107
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.3157730937039e-06,
      "aux_brier/mean_group_std": 0.07684176567918097,
      "aux_brier/mean_r": 0.9511278745127235,
      "aux_brier/n_active_tok": 269.5,
      "aux_brier/n_groups": 17.90625,
      "aux_brier/n_step_records": 67.375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5467066295582673,
      "calib/avg_num_step_conf": 8.4296875,
      "calib/ece": 0.5717637795275591,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003560408192392797,
      "calib/mean_conf": 0.10933858267716537,
      "calib/mu_c": 0.11047398843930635,
      "calib/mu_w": 0.10691358024691355,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.03944261664038043,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2982.0,
      "completions/max_terminated_length": 2982.0,
      "completions/mean_length": 507.61328125,
      "completions/mean_terminated_length": 507.61328125,
      "completions/min_length": 109.0,
      "completions/min_terminated_length": 109.0,
      "epoch": 0.1152,
      "grad_norm": 0.005916001740843058,
      "learning_rate": 2.5555555555555557e-06,
      "loss": -0.0125,
      "num_tokens": 23318635.0,
      "reward": 1.2800846099853516,
      "reward_std": 0.25666752457618713,
      "rewards/accuracy_reward_step": 0.67578125,
      "rewards/final_brier_reward_step": 0.44846364855766296,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 108
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.2602374175885167e-06,
      "aux_brier/mean_group_std": 0.03728945680070545,
      "aux_brier/mean_r": 0.9738470179694128,
      "aux_brier/n_active_tok": 266.625,
      "aux_brier/n_groups": 18.40625,
      "aux_brier/n_step_records": 66.65625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4810134620269241,
      "calib/avg_num_step_conf": 8.4140625,
      "calib/ece": 0.4011418326693227,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0035152654305308367,
      "calib/mean_conf": 0.10483426294820719,
      "calib/mu_c": 0.10309763779527562,
      "calib/mu_w": 0.10661290322580645,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04087527844312951,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3026.0,
      "completions/max_terminated_length": 3026.0,
      "completions/mean_length": 490.609375,
      "completions/mean_terminated_length": 494.4724426269531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 170.0,
      "epoch": 0.11626666666666667,
      "grad_norm": 0.020067334175109863,
      "learning_rate": 2.5277777777777778e-06,
      "loss": 0.0151,
      "num_tokens": 23548831.0,
      "reward": 1.1201258897781372,
      "reward_std": 0.1723223328590393,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5742534399032593,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 109
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.8512170134044048e-06,
      "aux_brier/mean_group_std": 0.0374497605813675,
      "aux_brier/mean_r": 0.9703611325446959,
      "aux_brier/n_active_tok": 232.125,
      "aux_brier/n_groups": 15.34375,
      "aux_brier/n_step_records": 58.03125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4751144164759725,
      "calib/avg_num_step_conf": 7.328125,
      "calib/ece": 0.3577777777777778,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.006224256292906169,
      "calib/mean_conf": 0.09761904761904762,
      "calib/mu_c": 0.09421052631578948,
      "calib/mu_w": 0.10043478260869565,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0015079365079365076,
      "calib/std_conf": 0.03454761084309574,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2616.0,
      "completions/max_terminated_length": 2616.0,
      "completions/mean_length": 445.58203125,
      "completions/mean_terminated_length": 447.3294372558594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 89.0,
      "epoch": 0.11733333333333333,
      "grad_norm": 0.008129195310175419,
      "learning_rate": 2.5e-06,
      "loss": 0.0431,
      "num_tokens": 23767820.0,
      "reward": 1.0906033515930176,
      "reward_std": 0.27544230222702026,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.6124132871627808,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 110
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.165272429631825e-06,
      "aux_brier/mean_group_std": 0.060017062456761895,
      "aux_brier/mean_r": 0.9527771586354886,
      "aux_brier/n_active_tok": 257.75,
      "aux_brier/n_groups": 18.09375,
      "aux_brier/n_step_records": 64.4375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.3436869959677419,
      "calib/avg_num_step_conf": 8.0546875,
      "calib/ece": 0.4167857142857143,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.02026965725806451,
      "calib/mean_conf": 0.10583333333333333,
      "calib/mu_c": 0.09585937500000002,
      "calib/mu_w": 0.11612903225806454,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00734126984126984,
      "calib/std_conf": 0.04434375996258152,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2782.0,
      "completions/max_terminated_length": 2782.0,
      "completions/mean_length": 504.375,
      "completions/mean_terminated_length": 504.375,
      "completions/min_length": 105.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.1184,
      "grad_norm": 0.007123699877411127,
      "learning_rate": 2.4722222222222226e-06,
      "loss": 0.076,
      "num_tokens": 24004348.0,
      "reward": 1.1340057849884033,
      "reward_std": 0.23329788446426392,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.5672730207443237,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 111
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 6.877805763805256e-07,
      "aux_brier/mean_group_std": 0.08182111997655978,
      "aux_brier/mean_r": 0.935912907754222,
      "aux_brier/n_active_tok": 246.0,
      "aux_brier/n_groups": 16.0625,
      "aux_brier/n_step_records": 61.5,
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.47402510683760685,
      "calib/avg_num_step_conf": 7.91015625,
      "calib/ece": 0.4161224489795919,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004250133547008539,
      "calib/mean_conf": 0.10632653061224491,
      "calib/mu_c": 0.10429687500000001,
      "calib/mu_w": 0.10854700854700855,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.03697179716922796,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2955.0,
      "completions/max_terminated_length": 2955.0,
      "completions/mean_length": 525.2421875,
      "completions/mean_terminated_length": 537.8480224609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 133.0,
      "epoch": 0.11946666666666667,
      "grad_norm": 0.009797528386116028,
      "learning_rate": 2.4444444444444447e-06,
      "loss": -0.0052,
      "num_tokens": 24246730.0,
      "reward": 1.1099687814712524,
      "reward_std": 0.2601466178894043,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.5414378643035889,
      "rewards/format_reward_step_strict": 0.94921875,
      "step": 112
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 4.8922507378534874e-08,
      "aux_brier/mean_group_std": 0.07131799118428014,
      "aux_brier/mean_r": 0.9567913860997896,
      "aux_brier/n_active_tok": 249.5,
      "aux_brier/n_groups": 15.71875,
      "aux_brier/n_step_records": 62.375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.536098554784491,
      "calib/avg_num_step_conf": 7.9921875,
      "calib/ece": 0.44539682539682535,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005447252817215226,
      "calib/mean_conf": 0.1061904761904762,
      "calib/mu_c": 0.10863309352517984,
      "calib/mu_w": 0.10318584070796462,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04030355340703298,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2731.0,
      "completions/max_terminated_length": 2731.0,
      "completions/mean_length": 451.68359375,
      "completions/mean_terminated_length": 453.4549255371094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 101.0,
      "epoch": 0.12053333333333334,
      "grad_norm": 0.03129152953624725,
      "learning_rate": 2.4166666666666667e-06,
      "loss": 0.0157,
      "num_tokens": 24467561.0,
      "reward": 1.1718251705169678,
      "reward_std": 0.26482635736465454,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.5466758012771606,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 113
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.0491426657165412e-06,
      "aux_brier/mean_group_std": 0.03986719759002069,
      "aux_brier/mean_r": 0.9700618380518691,
      "aux_brier/n_active_tok": 254.875,
      "aux_brier/n_groups": 16.75,
      "aux_brier/n_step_records": 63.71875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.49836076322864076,
      "calib/avg_num_step_conf": 8.68359375,
      "calib/ece": 0.5009126984126984,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004325617992262845,
      "calib/mean_conf": 0.10107142857142858,
      "calib/mu_c": 0.09933774834437084,
      "calib/mu_w": 0.10366336633663369,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0013888888888888887,
      "calib/std_conf": 0.034469374161740174,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2687.0,
      "completions/max_terminated_length": 2687.0,
      "completions/mean_length": 451.46875,
      "completions/mean_terminated_length": 456.8221435546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.1216,
      "grad_norm": 0.044061869382858276,
      "learning_rate": 2.388888888888889e-06,
      "loss": 0.0078,
      "num_tokens": 24688161.0,
      "reward": 1.2050158977508545,
      "reward_std": 0.209433913230896,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.49975115060806274,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 114
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.494877079274981e-06,
      "aux_brier/mean_group_std": 0.04775773662434711,
      "aux_brier/mean_r": 0.967077989380165,
      "aux_brier/n_active_tok": 239.25,
      "aux_brier/n_groups": 13.21875,
      "aux_brier/n_step_records": 59.8125,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5518178410794602,
      "calib/avg_num_step_conf": 7.53125,
      "calib/ece": 0.43562992125984246,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0017541229385307083,
      "calib/mean_conf": 0.10767716535433071,
      "calib/mu_c": 0.1084782608695652,
      "calib/mu_w": 0.1067241379310345,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.036978900536362214,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1323.0,
      "completions/max_terminated_length": 1323.0,
      "completions/mean_length": 447.9140625,
      "completions/mean_terminated_length": 449.6706237792969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.12266666666666666,
      "grad_norm": 0.02731587551534176,
      "learning_rate": 2.361111111111111e-06,
      "loss": 0.0108,
      "num_tokens": 24908091.0,
      "reward": 1.1723220348358154,
      "reward_std": 0.26217222213745117,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5564754009246826,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 115
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.5188563396459926e-07,
      "aux_brier/mean_group_std": 0.032075458344831345,
      "aux_brier/mean_r": 0.9742124829395283,
      "aux_brier/n_active_tok": 253.375,
      "aux_brier/n_groups": 18.03125,
      "aux_brier/n_step_records": 63.34375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5113386742473188,
      "calib/avg_num_step_conf": 7.93359375,
      "calib/ece": 0.4633466135458168,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": -0.008309859154929586,
      "calib/mean_conf": 0.11529880478087651,
      "calib/mu_c": 0.11169014084507044,
      "calib/mu_w": 0.12000000000000002,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.006454183266932271,
      "calib/std_conf": 0.06931229028266792,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2718.0,
      "completions/max_terminated_length": 2718.0,
      "completions/mean_length": 525.6953125,
      "completions/mean_terminated_length": 529.8346557617188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 90.0,
      "epoch": 0.12373333333333333,
      "grad_norm": 0.006843527778983116,
      "learning_rate": 2.3333333333333336e-06,
      "loss": 0.0429,
      "num_tokens": 25147189.0,
      "reward": 1.1779075860977173,
      "reward_std": 0.20672045648097992,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.5319429636001587,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 116
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.3370292073489054e-07,
      "aux_brier/mean_group_std": 0.03460002135801631,
      "aux_brier/mean_r": 0.973786716451061,
      "aux_brier/n_active_tok": 264.125,
      "aux_brier/n_groups": 15.25,
      "aux_brier/n_step_records": 66.03125,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.44812527294279114,
      "calib/avg_num_step_conf": 8.37109375,
      "calib/ece": 0.3456692913385827,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.009207062199762917,
      "calib/mean_conf": 0.1188976377952756,
      "calib/mu_c": 0.11393162393162395,
      "calib/mu_w": 0.12313868613138687,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.001968503937007874,
      "calib/std_conf": 0.04697508715604077,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1738.0,
      "completions/max_terminated_length": 1738.0,
      "completions/mean_length": 473.359375,
      "completions/mean_terminated_length": 475.2157287597656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 154.0,
      "epoch": 0.1248,
      "grad_norm": 0.04809485003352165,
      "learning_rate": 2.305555555555556e-06,
      "loss": 0.0031,
      "num_tokens": 25374969.0,
      "reward": 1.1069035530090332,
      "reward_std": 0.25470712780952454,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.6229265928268433,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 117
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.5123708263509172e-07,
      "aux_brier/mean_group_std": 0.040068772319806684,
      "aux_brier/mean_r": 0.9707969090019554,
      "aux_brier/n_active_tok": 265.375,
      "aux_brier/n_groups": 15.25,
      "aux_brier/n_step_records": 66.34375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.3718217645563183,
      "calib/avg_num_step_conf": 8.41015625,
      "calib/ece": 0.4444047619047619,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.02241418764302064,
      "calib/mean_conf": 0.11869047619047621,
      "calib/mu_c": 0.10855072463768115,
      "calib/mu_w": 0.13096491228070178,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00773809523809524,
      "calib/std_conf": 0.04176026901286929,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2097.0,
      "completions/max_terminated_length": 2097.0,
      "completions/mean_length": 479.99609375,
      "completions/mean_terminated_length": 483.77557373046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 102.0,
      "epoch": 0.12586666666666665,
      "grad_norm": 0.008244968950748444,
      "learning_rate": 2.277777777777778e-06,
      "loss": 0.02,
      "num_tokens": 25601856.0,
      "reward": 1.1679399013519287,
      "reward_std": 0.18270915746688843,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5467597246170044,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 118
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 9.787899220015106e-07,
      "aux_brier/mean_group_std": 0.06009293258223417,
      "aux_brier/mean_r": 0.9545985269931022,
      "aux_brier/n_active_tok": 296.625,
      "aux_brier/n_groups": 20.0625,
      "aux_brier/n_step_records": 74.15625,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.4995509654243377,
      "calib/avg_num_step_conf": 9.26953125,
      "calib/ece": 0.40492,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.007665020206555892,
      "calib/mean_conf": 0.12716,
      "calib/mu_c": 0.12351145038167939,
      "calib/mu_w": 0.13117647058823528,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00404,
      "calib/std_conf": 0.04804512878534097,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2787.0,
      "completions/max_terminated_length": 2787.0,
      "completions/mean_length": 570.80859375,
      "completions/mean_terminated_length": 570.80859375,
      "completions/min_length": 136.0,
      "completions/min_terminated_length": 136.0,
      "epoch": 0.12693333333333334,
      "grad_norm": 0.0063473815098404884,
      "learning_rate": 2.25e-06,
      "loss": 0.1005,
      "num_tokens": 25853047.0,
      "reward": 1.1375501155853271,
      "reward_std": 0.2780386507511139,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.5658254027366638,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 119
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 6.697449643766618e-07,
      "aux_brier/mean_group_std": 0.07134942448460872,
      "aux_brier/mean_r": 0.9538399584539912,
      "aux_brier/n_active_tok": 260.5,
      "aux_brier/n_groups": 17.4375,
      "aux_brier/n_step_records": 65.125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.466510067114094,
      "calib/avg_num_step_conf": 8.140625,
      "calib/ece": 0.4835341365461847,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004579865771812103,
      "calib/mean_conf": 0.11485943775100402,
      "calib/mu_c": 0.1130201342281879,
      "calib/mu_w": 0.1176,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.034988241301939595,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2618.0,
      "completions/max_terminated_length": 2618.0,
      "completions/mean_length": 493.08984375,
      "completions/mean_terminated_length": 495.0235595703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 159.0,
      "epoch": 0.128,
      "grad_norm": 0.009857230819761753,
      "learning_rate": 2.222222222222222e-06,
      "loss": 0.0882,
      "num_tokens": 26085966.0,
      "reward": 1.1971681118011475,
      "reward_std": 0.28364866971969604,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.507422685623169,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 120
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.9463558026687764e-07,
      "aux_brier/mean_group_std": 0.04828574374391937,
      "aux_brier/mean_r": 0.963195378810623,
      "aux_brier/n_active_tok": 265.875,
      "aux_brier/n_groups": 16.3125,
      "aux_brier/n_step_records": 66.46875,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5067996077149395,
      "calib/avg_num_step_conf": 8.57421875,
      "calib/ece": 0.34354838709677415,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004456358287021872,
      "calib/mean_conf": 0.12612903225806454,
      "calib/mu_c": 0.12373913043478263,
      "calib/mu_w": 0.1281954887218045,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0029838709677419356,
      "calib/std_conf": 0.04520936381417936,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2846.0,
      "completions/max_terminated_length": 2846.0,
      "completions/mean_length": 549.9609375,
      "completions/mean_terminated_length": 556.4822387695312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.12906666666666666,
      "grad_norm": 0.02184765599668026,
      "learning_rate": 2.1944444444444445e-06,
      "loss": 0.0568,
      "num_tokens": 26331812.0,
      "reward": 1.081617832183838,
      "reward_std": 0.28432029485702515,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.6077219247817993,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 121
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 5.333253980666086e-07,
      "aux_brier/mean_group_std": 0.04972777550477182,
      "aux_brier/mean_r": 0.9600536819027194,
      "aux_brier/n_active_tok": 267.5,
      "aux_brier/n_groups": 18.34375,
      "aux_brier/n_step_records": 66.875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.53106489403712,
      "calib/avg_num_step_conf": 8.6796875,
      "calib/ece": 0.453293172690763,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 6.778991707252768e-05,
      "calib/mean_conf": 0.11947791164658635,
      "calib/mu_c": 0.11950704225352111,
      "calib/mu_w": 0.11943925233644859,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0012449799196787147,
      "calib/std_conf": 0.0331984496973194,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3028.0,
      "completions/max_terminated_length": 3028.0,
      "completions/mean_length": 481.359375,
      "completions/mean_terminated_length": 485.14959716796875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 143.0,
      "epoch": 0.13013333333333332,
      "grad_norm": 0.012736966833472252,
      "learning_rate": 2.166666666666667e-06,
      "loss": 0.0392,
      "num_tokens": 26562384.0,
      "reward": 1.1788194179534912,
      "reward_std": 0.24727436900138855,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.5355902910232544,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 122
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -8.735640291690494e-07,
      "aux_brier/mean_group_std": 0.07085951931389238,
      "aux_brier/mean_r": 0.9459146764310977,
      "aux_brier/n_active_tok": 281.875,
      "aux_brier/n_groups": 18.0,
      "aux_brier/n_step_records": 70.46875,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4719491035280509,
      "calib/avg_num_step_conf": 8.85546875,
      "calib/ece": 0.34584,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004656513077565705,
      "calib/mean_conf": 0.12752000000000002,
      "calib/mu_c": 0.12504273504273505,
      "calib/mu_w": 0.12969924812030076,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0026799999999999997,
      "calib/std_conf": 0.04530176155515368,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2819.0,
      "completions/max_terminated_length": 2819.0,
      "completions/mean_length": 558.65625,
      "completions/mean_terminated_length": 563.0551147460938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 134.0,
      "epoch": 0.1312,
      "grad_norm": 0.02399447374045849,
      "learning_rate": 2.138888888888889e-06,
      "loss": -0.0337,
      "num_tokens": 26810688.0,
      "reward": 1.0906023979187012,
      "reward_std": 0.2960851788520813,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.604597270488739,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 123
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.1318501275114912e-07,
      "aux_brier/mean_group_std": 0.059568114978127334,
      "aux_brier/mean_r": 0.9554816319526316,
      "aux_brier/n_active_tok": 259.75,
      "aux_brier/n_groups": 16.46875,
      "aux_brier/n_step_records": 64.9375,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.4354725415070243,
      "calib/avg_num_step_conf": 8.34765625,
      "calib/ece": 0.4167330677290837,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.016418263090676885,
      "calib/mean_conf": 0.12677290836653388,
      "calib/mu_c": 0.11918518518518519,
      "calib/mu_w": 0.13560344827586207,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.002828685258964143,
      "calib/std_conf": 0.044564562403536456,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2628.0,
      "completions/max_terminated_length": 2628.0,
      "completions/mean_length": 505.5625,
      "completions/mean_terminated_length": 509.5433044433594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.13226666666666667,
      "grad_norm": 0.05152898281812668,
      "learning_rate": 2.1111111111111114e-06,
      "loss": 0.0187,
      "num_tokens": 27046928.0,
      "reward": 1.1520527601242065,
      "reward_std": 0.19784283638000488,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.5535234212875366,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 124
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 4.2310242481802263e-07,
      "aux_brier/mean_group_std": 0.06881672184727762,
      "aux_brier/mean_r": 0.9462462088289638,
      "aux_brier/n_active_tok": 270.75,
      "aux_brier/n_groups": 18.46875,
      "aux_brier/n_step_records": 67.6875,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.43086204671570516,
      "calib/avg_num_step_conf": 8.55078125,
      "calib/ece": 0.39590240963855416,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.01998019744483155,
      "calib/mean_conf": 0.12915783132530123,
      "calib/mu_c": 0.11928809523809523,
      "calib/mu_w": 0.1392682926829268,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00951807228915663,
      "calib/std_conf": 0.05250532530761328,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3014.0,
      "completions/max_terminated_length": 3014.0,
      "completions/mean_length": 546.67578125,
      "completions/mean_terminated_length": 548.8196411132812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 153.0,
      "epoch": 0.13333333333333333,
      "grad_norm": 0.026436125859618187,
      "learning_rate": 2.0833333333333334e-06,
      "loss": 0.0631,
      "num_tokens": 27291685.0,
      "reward": 1.1183891296386719,
      "reward_std": 0.259613573551178,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.5751187801361084,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 125
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.0745253512522979e-06,
      "aux_brier/mean_group_std": 0.06444103998597421,
      "aux_brier/mean_r": 0.9498170013024771,
      "aux_brier/n_active_tok": 279.5,
      "aux_brier/n_groups": 18.1875,
      "aux_brier/n_step_records": 69.875,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5129921259842519,
      "calib/avg_num_step_conf": 8.78515625,
      "calib/ece": 0.3904858299595142,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.004048582995951417,
      "calib/gap": -0.008677165354330746,
      "calib/mean_conf": 0.13153846153846152,
      "calib/mu_c": 0.12732283464566926,
      "calib/mu_w": 0.136,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.003927125506072875,
      "calib/std_conf": 0.06646101364315203,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2899.0,
      "completions/max_terminated_length": 2899.0,
      "completions/mean_length": 535.74609375,
      "completions/mean_terminated_length": 539.9645385742188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.1344,
      "grad_norm": 0.02762473188340664,
      "learning_rate": 2.0555555555555555e-06,
      "loss": 0.0681,
      "num_tokens": 27534300.0,
      "reward": 1.113364577293396,
      "reward_std": 0.27454179525375366,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5628331899642944,
      "rewards/format_reward_step_strict": 0.953125,
      "step": 126
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.333506252476635e-06,
      "aux_brier/mean_group_std": 0.0493533218266499,
      "aux_brier/mean_r": 0.9629633992998795,
      "aux_brier/n_active_tok": 270.625,
      "aux_brier/n_groups": 16.28125,
      "aux_brier/n_step_records": 67.65625,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.43862990702479343,
      "calib/avg_num_step_conf": 8.5234375,
      "calib/ece": 0.387710843373494,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.007109374999999987,
      "calib/mean_conf": 0.12634538152610442,
      "calib/mu_c": 0.12289062500000002,
      "calib/mu_w": 0.13,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.03911543503255671,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2598.0,
      "completions/max_terminated_length": 2598.0,
      "completions/mean_length": 523.8359375,
      "completions/mean_terminated_length": 525.8901977539062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 109.0,
      "epoch": 0.13546666666666668,
      "grad_norm": 0.007691785227507353,
      "learning_rate": 2.027777777777778e-06,
      "loss": 0.075,
      "num_tokens": 27772074.0,
      "reward": 1.1309611797332764,
      "reward_std": 0.24632525444030762,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.5785320401191711,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 127
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.216938798678015e-07,
      "aux_brier/mean_group_std": 0.06190802379627205,
      "aux_brier/mean_r": 0.9577295168062887,
      "aux_brier/n_active_tok": 237.5,
      "aux_brier/n_groups": 15.5,
      "aux_brier/n_step_records": 59.375,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.6191532258064515,
      "calib/avg_num_step_conf": 7.56640625,
      "calib/ece": 0.37377049180327876,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.019284946236559145,
      "calib/mean_conf": 0.11803278688524592,
      "calib/mu_c": 0.12783333333333335,
      "calib/mu_w": 0.10854838709677421,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05938398724751983,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2934.0,
      "completions/max_terminated_length": 2934.0,
      "completions/mean_length": 515.67578125,
      "completions/mean_terminated_length": 523.8611450195312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 151.0,
      "epoch": 0.13653333333333334,
      "grad_norm": 0.03205028548836708,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0908,
      "num_tokens": 28010751.0,
      "reward": 1.087334156036377,
      "reward_std": 0.29081636667251587,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5837117433547974,
      "rewards/format_reward_step_strict": 0.9453125,
      "step": 128
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.2448030372324013e-06,
      "aux_brier/mean_group_std": 0.07537504328134718,
      "aux_brier/mean_r": 0.9467429778290359,
      "aux_brier/n_active_tok": 258.375,
      "aux_brier/n_groups": 18.375,
      "aux_brier/n_step_records": 64.59375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.589994098747623,
      "calib/avg_num_step_conf": 8.1484375,
      "calib/ece": 0.47690476190476194,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005471116648088636,
      "calib/mean_conf": 0.12555555555555556,
      "calib/mu_c": 0.1277483443708609,
      "calib/mu_w": 0.12227722772277227,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0016269841269841267,
      "calib/std_conf": 0.04094238219268626,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2865.0,
      "completions/max_terminated_length": 2865.0,
      "completions/mean_length": 480.96484375,
      "completions/mean_terminated_length": 482.85101318359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 156.0,
      "epoch": 0.1376,
      "grad_norm": 0.008966000750660896,
      "learning_rate": 1.9722222222222224e-06,
      "loss": 0.0879,
      "num_tokens": 28236262.0,
      "reward": 1.214047908782959,
      "reward_std": 0.2387232780456543,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.5280663967132568,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 129
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.5649928951931855e-06,
      "aux_brier/mean_group_std": 0.05652320397212803,
      "aux_brier/mean_r": 0.9549481129361472,
      "aux_brier/n_active_tok": 242.5,
      "aux_brier/n_groups": 13.28125,
      "aux_brier/n_step_records": 60.625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4777851679414699,
      "calib/avg_num_step_conf": 7.578125,
      "calib/ece": 0.4907539682539682,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.006510808114399719,
      "calib/mean_conf": 0.13011904761904763,
      "calib/mu_c": 0.12761290322580646,
      "calib/mu_w": 0.13412371134020618,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0028968253968253968,
      "calib/std_conf": 0.043365209364419705,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2437.0,
      "completions/max_terminated_length": 2437.0,
      "completions/mean_length": 464.40234375,
      "completions/mean_terminated_length": 464.40234375,
      "completions/min_length": 142.0,
      "completions/min_terminated_length": 142.0,
      "epoch": 0.13866666666666666,
      "grad_norm": 0.0086931511759758,
      "learning_rate": 1.944444444444445e-06,
      "loss": 0.0319,
      "num_tokens": 28460437.0,
      "reward": 1.2212738990783691,
      "reward_std": 0.18101811408996582,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.5100957155227661,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 130
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.616686539114177e-07,
      "aux_brier/mean_group_std": 0.05609984517634906,
      "aux_brier/mean_r": 0.9611126952734949,
      "aux_brier/n_active_tok": 234.75,
      "aux_brier/n_groups": 13.34375,
      "aux_brier/n_step_records": 58.6875,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5641874681609782,
      "calib/avg_num_step_conf": 7.4375,
      "calib/ece": 0.2830980392156862,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.011147478349465098,
      "calib/mean_conf": 0.1247450980392157,
      "calib/mu_c": 0.13134615384615383,
      "calib/mu_w": 0.12019867549668874,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.042202055120135286,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1240.0,
      "completions/max_terminated_length": 1240.0,
      "completions/mean_length": 440.3359375,
      "completions/mean_terminated_length": 442.0627746582031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.13973333333333332,
      "grad_norm": 0.012002948671579361,
      "learning_rate": 1.916666666666667e-06,
      "loss": 0.0208,
      "num_tokens": 28679371.0,
      "reward": 1.0712454319000244,
      "reward_std": 0.20326244831085205,
      "rewards/accuracy_reward_step": 0.40625,
      "rewards/final_brier_reward_step": 0.6756066083908081,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 131
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.3460213122161733e-07,
      "aux_brier/mean_group_std": 0.09031914227052233,
      "aux_brier/mean_r": 0.9348492362155827,
      "aux_brier/n_active_tok": 263.125,
      "aux_brier/n_groups": 15.375,
      "aux_brier/n_step_records": 65.78125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4654584491033089,
      "calib/avg_num_step_conf": 8.2890625,
      "calib/ece": 0.43913725490196087,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0034737307400858897,
      "calib/mean_conf": 0.14125490196078433,
      "calib/mu_c": 0.1397972972972973,
      "calib/mu_w": 0.1432710280373832,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05363633650444295,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1533.0,
      "completions/max_terminated_length": 1533.0,
      "completions/mean_length": 493.8046875,
      "completions/mean_terminated_length": 495.7412109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 143.0,
      "epoch": 0.1408,
      "grad_norm": 0.05939352884888649,
      "learning_rate": 1.888888888888889e-06,
      "loss": -0.0142,
      "num_tokens": 28911377.0,
      "reward": 1.2067489624023438,
      "reward_std": 0.2673734426498413,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.5457457304000854,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 132
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.0812276646232633e-06,
      "aux_brier/mean_group_std": 0.05434612503982542,
      "aux_brier/mean_r": 0.9616093050296004,
      "aux_brier/n_active_tok": 285.25,
      "aux_brier/n_groups": 16.90625,
      "aux_brier/n_step_records": 71.3125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.45672076570952974,
      "calib/avg_num_step_conf": 9.12109375,
      "calib/ece": 0.21940239043824702,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.010070744902205608,
      "calib/mean_conf": 0.14346613545816736,
      "calib/mu_c": 0.13696629213483144,
      "calib/mu_w": 0.14703703703703705,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.004143426294820717,
      "calib/std_conf": 0.05152976404250724,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2577.0,
      "completions/max_terminated_length": 2577.0,
      "completions/mean_length": 579.0390625,
      "completions/mean_terminated_length": 581.309814453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 107.0,
      "epoch": 0.14186666666666667,
      "grad_norm": 0.02066822350025177,
      "learning_rate": 1.8611111111111113e-06,
      "loss": 0.0487,
      "num_tokens": 29165955.0,
      "reward": 1.0142062902450562,
      "reward_std": 0.2862761914730072,
      "rewards/accuracy_reward_step": 0.34765625,
      "rewards/final_brier_reward_step": 0.7052628993988037,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 133
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -7.850857843827441e-07,
      "aux_brier/mean_group_std": 0.053333546563048284,
      "aux_brier/mean_r": 0.9595509305591835,
      "aux_brier/n_active_tok": 279.125,
      "aux_brier/n_groups": 16.96875,
      "aux_brier/n_step_records": 69.78125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5063343519225871,
      "calib/avg_num_step_conf": 8.91015625,
      "calib/ece": 0.33087649402390445,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00042653425006375367,
      "calib/mean_conf": 0.14641434262948208,
      "calib/mu_c": 0.14663865546218494,
      "calib/mu_w": 0.14621212121212118,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0015936254980079682,
      "calib/std_conf": 0.0563912382066595,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2657.0,
      "completions/max_terminated_length": 2657.0,
      "completions/mean_length": 564.0625,
      "completions/mean_terminated_length": 568.50390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 187.0,
      "epoch": 0.14293333333333333,
      "grad_norm": 0.03287725895643234,
      "learning_rate": 1.8333333333333333e-06,
      "loss": -0.0112,
      "num_tokens": 29419307.0,
      "reward": 1.1120322942733765,
      "reward_std": 0.29269343614578247,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.6278167963027954,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 134
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.5976514672620965e-07,
      "aux_brier/mean_group_std": 0.027995954789728373,
      "aux_brier/mean_r": 0.9751566209296453,
      "aux_brier/n_active_tok": 270.375,
      "aux_brier/n_groups": 17.03125,
      "aux_brier/n_step_records": 67.59375,
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5068783068783069,
      "calib/avg_num_step_conf": 8.72265625,
      "calib/ece": 0.38109756097560976,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0036071428571428366,
      "calib/mean_conf": 0.13109756097560976,
      "calib/mu_c": 0.13285714285714284,
      "calib/mu_w": 0.12925,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04556794527217112,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3023.0,
      "completions/max_terminated_length": 3023.0,
      "completions/mean_length": 534.2265625,
      "completions/mean_terminated_length": 540.561279296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 134.0,
      "epoch": 0.144,
      "grad_norm": 0.026167117059230804,
      "learning_rate": 1.8055555555555557e-06,
      "loss": 0.0611,
      "num_tokens": 29661949.0,
      "reward": 1.115006685256958,
      "reward_std": 0.31411486864089966,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.5772144794464111,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 135
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 7.419795309826505e-07,
      "aux_brier/mean_group_std": 0.03866174836164761,
      "aux_brier/mean_r": 0.9682558963961708,
      "aux_brier/n_active_tok": 283.0,
      "aux_brier/n_groups": 17.8125,
      "aux_brier/n_step_records": 70.75,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5845853658536586,
      "calib/avg_num_step_conf": 9.01171875,
      "calib/ece": 0.35963709677419353,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.011139512195121881,
      "calib/mean_conf": 0.1443951612903226,
      "calib/mu_c": 0.14991999999999997,
      "calib/mu_w": 0.1387804878048781,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05304856547699979,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3041.0,
      "completions/max_terminated_length": 3041.0,
      "completions/mean_length": 515.52734375,
      "completions/mean_terminated_length": 521.6403198242188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 171.0,
      "epoch": 0.14506666666666668,
      "grad_norm": 0.09230382740497589,
      "learning_rate": 1.777777777777778e-06,
      "loss": 0.0336,
      "num_tokens": 29902412.0,
      "reward": 1.1248645782470703,
      "reward_std": 0.2411046326160431,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.6010206937789917,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 136
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.368074245948318e-07,
      "aux_brier/mean_group_std": 0.06382341056104424,
      "aux_brier/mean_r": 0.9450567836904806,
      "aux_brier/n_active_tok": 277.25,
      "aux_brier/n_groups": 17.03125,
      "aux_brier/n_step_records": 69.3125,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.537962962962963,
      "calib/avg_num_step_conf": 8.99609375,
      "calib/ece": 0.343739837398374,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004063492063492075,
      "calib/mean_conf": 0.14658536585365856,
      "calib/mu_c": 0.1486666666666667,
      "calib/mu_w": 0.14460317460317462,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0012601626016260164,
      "calib/std_conf": 0.04126711063290114,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2812.0,
      "completions/max_terminated_length": 2812.0,
      "completions/mean_length": 491.6484375,
      "completions/mean_terminated_length": 499.4524230957031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 193.0,
      "epoch": 0.14613333333333334,
      "grad_norm": 0.07573703676462173,
      "learning_rate": 1.75e-06,
      "loss": 0.0318,
      "num_tokens": 30135258.0,
      "reward": 1.1015383005142212,
      "reward_std": 0.272910475730896,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.60927814245224,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 137
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.365847257019716e-07,
      "aux_brier/mean_group_std": 0.03673627551638543,
      "aux_brier/mean_r": 0.9593570070056029,
      "aux_brier/n_active_tok": 288.125,
      "aux_brier/n_groups": 20.25,
      "aux_brier/n_step_records": 72.03125,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5379941284904759,
      "calib/avg_num_step_conf": 9.21875,
      "calib/ece": 0.46395161290322584,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005709018911722569,
      "calib/mean_conf": 0.14491935483870966,
      "calib/mu_c": 0.14715231788079472,
      "calib/mu_w": 0.14144329896907215,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04999993496353731,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2899.0,
      "completions/max_terminated_length": 2899.0,
      "completions/mean_length": 521.3984375,
      "completions/mean_terminated_length": 523.4431762695312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 122.0,
      "epoch": 0.1472,
      "grad_norm": 0.009818018414080143,
      "learning_rate": 1.7222222222222224e-06,
      "loss": 0.0436,
      "num_tokens": 30373072.0,
      "reward": 1.2066519260406494,
      "reward_std": 0.2799428701400757,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.5297328233718872,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 138
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.0452663158395694e-08,
      "aux_brier/mean_group_std": 0.05374665811390989,
      "aux_brier/mean_r": 0.9614916759713343,
      "aux_brier/n_active_tok": 250.375,
      "aux_brier/n_groups": 13.8125,
      "aux_brier/n_step_records": 62.59375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5967126124377143,
      "calib/avg_num_step_conf": 7.82421875,
      "calib/ece": 0.4651181102362204,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.01449168446256388,
      "calib/mean_conf": 0.13724409448818897,
      "calib/mu_c": 0.1430065359477124,
      "calib/mu_w": 0.12851485148514852,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04110055204630173,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2799.0,
      "completions/max_terminated_length": 2799.0,
      "completions/mean_length": 455.1171875,
      "completions/mean_terminated_length": 455.1171875,
      "completions/min_length": 130.0,
      "completions/min_terminated_length": 130.0,
      "epoch": 0.14826666666666666,
      "grad_norm": 0.008437794633209705,
      "learning_rate": 1.6944444444444446e-06,
      "loss": 0.0459,
      "num_tokens": 30592678.0,
      "reward": 1.2212949991226196,
      "reward_std": 0.2543236017227173,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.5336176156997681,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 139
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.9473522832047507e-06,
      "aux_brier/mean_group_std": 0.042736197051407236,
      "aux_brier/mean_r": 0.9649691936762791,
      "aux_brier/n_active_tok": 235.125,
      "aux_brier/n_groups": 12.90625,
      "aux_brier/n_step_records": 58.78125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5880708929788684,
      "calib/avg_num_step_conf": 7.34765625,
      "calib/ece": 0.5006719367588933,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.018517382413087924,
      "calib/mean_conf": 0.14359683794466402,
      "calib/mu_c": 0.15018404907975458,
      "calib/mu_w": 0.13166666666666665,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05844584255308957,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 492.02734375,
      "completions/mean_terminated_length": 492.02734375,
      "completions/min_length": 150.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.14933333333333335,
      "grad_norm": 0.007128539029508829,
      "learning_rate": 1.6666666666666667e-06,
      "loss": 0.0045,
      "num_tokens": 30823653.0,
      "reward": 1.2577022314071655,
      "reward_std": 0.23854640126228333,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.5151835680007935,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 140
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.102467689963031e-07,
      "aux_brier/mean_group_std": 0.06433482313202449,
      "aux_brier/mean_r": 0.9463781406931842,
      "aux_brier/n_active_tok": 245.0,
      "aux_brier/n_groups": 13.875,
      "aux_brier/n_step_records": 61.25,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5789543524416136,
      "calib/avg_num_step_conf": 7.66015625,
      "calib/ece": 0.4796442687747035,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004854697452229301,
      "calib/mean_conf": 0.14707509881422925,
      "calib/mu_c": 0.1489171974522293,
      "calib/mu_w": 0.1440625,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0030830039525691698,
      "calib/std_conf": 0.06790005364555529,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1896.0,
      "completions/max_terminated_length": 1896.0,
      "completions/mean_length": 504.4765625,
      "completions/mean_terminated_length": 506.4549255371094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 191.0,
      "epoch": 0.1504,
      "grad_norm": 0.013903039507567883,
      "learning_rate": 1.638888888888889e-06,
      "loss": -0.0088,
      "num_tokens": 31059895.0,
      "reward": 1.2403523921966553,
      "reward_std": 0.23382900655269623,
      "rewards/accuracy_reward_step": 0.61328125,
      "rewards/final_brier_reward_step": 0.5317223072052002,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 141
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.0547654013190466e-06,
      "aux_brier/mean_group_std": 0.05228157119585481,
      "aux_brier/mean_r": 0.9612131276230842,
      "aux_brier/n_active_tok": 253.75,
      "aux_brier/n_groups": 13.5,
      "aux_brier/n_step_records": 63.4375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5245987654320988,
      "calib/avg_num_step_conf": 7.953125,
      "calib/ece": 0.33168627450980387,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0026759259259259427,
      "calib/mean_conf": 0.14666666666666667,
      "calib/mu_c": 0.14808333333333332,
      "calib/mu_w": 0.14540740740740737,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003882352941176471,
      "calib/std_conf": 0.0520507545166411,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1383.0,
      "completions/max_terminated_length": 1383.0,
      "completions/mean_length": 497.546875,
      "completions/mean_terminated_length": 499.4980773925781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 146.0,
      "epoch": 0.15146666666666667,
      "grad_norm": 0.014494812116026878,
      "learning_rate": 1.6111111111111113e-06,
      "loss": 0.035,
      "num_tokens": 31292427.0,
      "reward": 1.1244139671325684,
      "reward_std": 0.20455797016620636,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6382808685302734,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 142
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 6.215460009867391e-07,
      "aux_brier/mean_group_std": 0.04666321619680647,
      "aux_brier/mean_r": 0.9616438497821584,
      "aux_brier/n_active_tok": 267.25,
      "aux_brier/n_groups": 17.0,
      "aux_brier/n_step_records": 66.8125,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5603998158742685,
      "calib/avg_num_step_conf": 8.49609375,
      "calib/ece": 0.4161290322580645,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0043348457946998376,
      "calib/mean_conf": 0.1403225806451613,
      "calib/mu_c": 0.14226277372262774,
      "calib/mu_w": 0.1379279279279279,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0020161290322580645,
      "calib/std_conf": 0.05181286868223189,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3030.0,
      "completions/max_terminated_length": 3030.0,
      "completions/mean_length": 533.7109375,
      "completions/mean_terminated_length": 535.803955078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 131.0,
      "epoch": 0.15253333333333333,
      "grad_norm": 0.012073071673512459,
      "learning_rate": 1.5833333333333333e-06,
      "loss": 0.0206,
      "num_tokens": 31536393.0,
      "reward": 1.1518467664718628,
      "reward_std": 0.21789219975471497,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.5605117082595825,
      "rewards/format_reward_step_strict": 0.953125,
      "step": 143
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.017924849440078e-08,
      "aux_brier/mean_group_std": 0.02846353088420098,
      "aux_brier/mean_r": 0.9751172320338265,
      "aux_brier/n_active_tok": 264.75,
      "aux_brier/n_groups": 15.4375,
      "aux_brier/n_step_records": 66.1875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5559006211180124,
      "calib/avg_num_step_conf": 8.3671875,
      "calib/ece": 0.49158730158730146,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.007300525561395127,
      "calib/mean_conf": 0.1473015873015873,
      "calib/mu_c": 0.14993788819875778,
      "calib/mu_w": 0.14263736263736265,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04959218925975856,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2553.0,
      "completions/max_terminated_length": 2553.0,
      "completions/mean_length": 507.02734375,
      "completions/mean_terminated_length": 509.0157165527344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 117.0,
      "epoch": 0.1536,
      "grad_norm": 0.061257150024175644,
      "learning_rate": 1.5555555555555558e-06,
      "loss": 0.0094,
      "num_tokens": 31770320.0,
      "reward": 1.249025821685791,
      "reward_std": 0.22299563884735107,
      "rewards/accuracy_reward_step": 0.62890625,
      "rewards/final_brier_reward_step": 0.5195406675338745,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 144
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.059240721495991e-06,
      "aux_brier/mean_group_std": 0.06043944022073585,
      "aux_brier/mean_r": 0.9429568436991247,
      "aux_brier/n_active_tok": 275.5,
      "aux_brier/n_groups": 17.53125,
      "aux_brier/n_step_records": 68.875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.4821592271348987,
      "calib/avg_num_step_conf": 8.61328125,
      "calib/ece": 0.48932000000000003,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0031175516536210146,
      "calib/mean_conf": 0.16268000000000002,
      "calib/mu_c": 0.16159509202453987,
      "calib/mu_w": 0.16471264367816088,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05991007928554259,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2742.0,
      "completions/max_terminated_length": 2742.0,
      "completions/mean_length": 514.234375,
      "completions/mean_terminated_length": 514.234375,
      "completions/min_length": 141.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.15466666666666667,
      "grad_norm": 0.007433212827891111,
      "learning_rate": 1.527777777777778e-06,
      "loss": 0.0847,
      "num_tokens": 32004668.0,
      "reward": 1.2540688514709473,
      "reward_std": 0.26964327692985535,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.5162754058837891,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 145
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 7.627102511231776e-07,
      "aux_brier/mean_group_std": 0.027543769427134806,
      "aux_brier/mean_r": 0.9743984052820087,
      "aux_brier/n_active_tok": 253.125,
      "aux_brier/n_groups": 15.1875,
      "aux_brier/n_step_records": 63.28125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6105024927840462,
      "calib/avg_num_step_conf": 7.94140625,
      "calib/ece": 0.26549800796812756,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.014973104172133322,
      "calib/mean_conf": 0.14486055776892431,
      "calib/mu_c": 0.15368932038834954,
      "calib/mu_w": 0.13871621621621621,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04644622129085265,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2520.0,
      "completions/max_terminated_length": 2520.0,
      "completions/mean_length": 501.9375,
      "completions/mean_terminated_length": 503.9059143066406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 126.0,
      "epoch": 0.15573333333333333,
      "grad_norm": 0.024848168715834618,
      "learning_rate": 1.5e-06,
      "loss": 0.0479,
      "num_tokens": 32240380.0,
      "reward": 1.0565755367279053,
      "reward_std": 0.24422100186347961,
      "rewards/accuracy_reward_step": 0.40234375,
      "rewards/final_brier_reward_step": 0.6716148853302002,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 146
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 5.208359558639408e-07,
      "aux_brier/mean_group_std": 0.042396414436749914,
      "aux_brier/mean_r": 0.961728043766565,
      "aux_brier/n_active_tok": 262.25,
      "aux_brier/n_groups": 14.8125,
      "aux_brier/n_step_records": 65.5625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6199466056445461,
      "calib/avg_num_step_conf": 8.25,
      "calib/ece": 0.2928174603174603,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.018577421815408085,
      "calib/mean_conf": 0.15956349206349207,
      "calib/mu_c": 0.16973684210526316,
      "calib/mu_w": 0.15115942028985507,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.0482574109830772,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2741.0,
      "completions/max_terminated_length": 2741.0,
      "completions/mean_length": 506.20703125,
      "completions/mean_terminated_length": 512.2095336914062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 87.0,
      "epoch": 0.1568,
      "grad_norm": 0.01770801469683647,
      "learning_rate": 1.4722222222222225e-06,
      "loss": 0.0015,
      "num_tokens": 32473649.0,
      "reward": 1.0973799228668213,
      "reward_std": 0.2255202680826187,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.6551449298858643,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 147
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.9145170915280119e-07,
      "aux_brier/mean_group_std": 0.04804261957314319,
      "aux_brier/mean_r": 0.9610529469897078,
      "aux_brier/n_active_tok": 250.25,
      "aux_brier/n_groups": 14.34375,
      "aux_brier/n_step_records": 62.5625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6382622368061253,
      "calib/avg_num_step_conf": 7.828125,
      "calib/ece": 0.4787649402390438,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.01874624008750342,
      "calib/mean_conf": 0.1547011952191235,
      "calib/mu_c": 0.16157232704402513,
      "calib/mu_w": 0.14282608695652171,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05005385902557692,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2682.0,
      "completions/max_terminated_length": 2682.0,
      "completions/mean_length": 496.4140625,
      "completions/mean_terminated_length": 498.3608093261719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 80.0,
      "epoch": 0.15786666666666666,
      "grad_norm": 0.029627349227666855,
      "learning_rate": 1.4444444444444445e-06,
      "loss": 0.0217,
      "num_tokens": 32705843.0,
      "reward": 1.2390427589416504,
      "reward_std": 0.31594008207321167,
      "rewards/accuracy_reward_step": 0.62109375,
      "rewards/final_brier_reward_step": 0.5264832377433777,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 148
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 4.4705874335804197e-07,
      "aux_brier/mean_group_std": 0.05452732319415469,
      "aux_brier/mean_r": 0.9559296365957464,
      "aux_brier/n_active_tok": 282.625,
      "aux_brier/n_groups": 16.34375,
      "aux_brier/n_step_records": 70.65625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5492105933282403,
      "calib/avg_num_step_conf": 8.921875,
      "calib/ece": 0.3125498007968127,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.008390628978864195,
      "calib/mean_conf": 0.16155378486055777,
      "calib/mu_c": 0.16596638655462181,
      "calib/mu_w": 0.15757575757575762,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.048650883463411516,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3023.0,
      "completions/max_terminated_length": 3023.0,
      "completions/mean_length": 572.5,
      "completions/mean_terminated_length": 574.7451171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 98.0,
      "epoch": 0.15893333333333334,
      "grad_norm": 0.01007721945643425,
      "learning_rate": 1.4166666666666667e-06,
      "loss": 0.0684,
      "num_tokens": 32956859.0,
      "reward": 1.1155809164047241,
      "reward_std": 0.25350266695022583,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.6420113444328308,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 149
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -4.075938205838625e-07,
      "aux_brier/mean_group_std": 0.06084920416650598,
      "aux_brier/mean_r": 0.9450912648550661,
      "aux_brier/n_active_tok": 255.0,
      "aux_brier/n_groups": 16.03125,
      "aux_brier/n_step_records": 63.75,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5419854947102268,
      "calib/avg_num_step_conf": 8.390625,
      "calib/ece": 0.38699186991869916,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005154035531306195,
      "calib/mean_conf": 0.1595121951219512,
      "calib/mu_c": 0.16187969924812032,
      "calib/mu_w": 0.15672566371681412,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0029268292682926834,
      "calib/std_conf": 0.05698172091070668,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2674.0,
      "completions/max_terminated_length": 2674.0,
      "completions/mean_length": 478.140625,
      "completions/mean_terminated_length": 483.810302734375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 144.0,
      "epoch": 0.16,
      "grad_norm": 0.04251770302653313,
      "learning_rate": 1.3888888888888892e-06,
      "loss": -0.0007,
      "num_tokens": 33184223.0,
      "reward": 1.1404184103012085,
      "reward_std": 0.24159850180149078,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.5772988796234131,
      "rewards/format_reward_step_strict": 0.953125,
      "step": 150
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.0426429335602672e-06,
      "aux_brier/mean_group_std": 0.060192081828735935,
      "aux_brier/mean_r": 0.9468632518120125,
      "aux_brier/n_active_tok": 281.375,
      "aux_brier/n_groups": 19.21875,
      "aux_brier/n_step_records": 70.34375,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5349037823490378,
      "calib/avg_num_step_conf": 8.97265625,
      "calib/ece": 0.28433198380566804,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.000797611147976135,
      "calib/mean_conf": 0.16101214574898787,
      "calib/mu_c": 0.1614545454545455,
      "calib/mu_w": 0.16065693430656935,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.059527388435393944,
      "calib/step_conf_rate": 0.98046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3053.0,
      "completions/max_terminated_length": 3053.0,
      "completions/mean_length": 566.61328125,
      "completions/mean_terminated_length": 571.0748291015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 177.0,
      "epoch": 0.16106666666666666,
      "grad_norm": 0.024479515850543976,
      "learning_rate": 1.3611111111111112e-06,
      "loss": 0.0408,
      "num_tokens": 33436300.0,
      "reward": 1.0677103996276855,
      "reward_std": 0.26120448112487793,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.6380289196968079,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 151
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 5.760061240756009e-07,
      "aux_brier/mean_group_std": 0.03143787467963727,
      "aux_brier/mean_r": 0.972342106949101,
      "aux_brier/n_active_tok": 260.25,
      "aux_brier/n_groups": 15.46875,
      "aux_brier/n_step_records": 65.0625,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5108595800524934,
      "calib/avg_num_step_conf": 8.171875,
      "calib/ece": 0.3569230769230769,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0024251968503936905,
      "calib/mean_conf": 0.15724696356275306,
      "calib/mu_c": 0.1584251968503937,
      "calib/mu_w": 0.156,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04700451934610316,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2543.0,
      "completions/max_terminated_length": 2543.0,
      "completions/mean_length": 495.3984375,
      "completions/mean_terminated_length": 499.2992248535156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.16213333333333332,
      "grad_norm": 0.02353908307850361,
      "learning_rate": 1.3333333333333334e-06,
      "loss": 0.0081,
      "num_tokens": 33668514.0,
      "reward": 1.1226801872253418,
      "reward_std": 0.33606499433517456,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5922832489013672,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 152
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.107078837607304e-08,
      "aux_brier/mean_group_std": 0.06549836477067371,
      "aux_brier/mean_r": 0.9423625538808893,
      "aux_brier/n_active_tok": 251.375,
      "aux_brier/n_groups": 14.28125,
      "aux_brier/n_step_records": 62.84375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4867467581998475,
      "calib/avg_num_step_conf": 7.9609375,
      "calib/ece": 0.3948809523809523,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.008867276887871839,
      "calib/mean_conf": 0.1632142857142857,
      "calib/mu_c": 0.15920289855072464,
      "calib/mu_w": 0.16807017543859648,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.005238095238095236,
      "calib/std_conf": 0.06122740686550364,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2310.0,
      "completions/max_terminated_length": 2310.0,
      "completions/mean_length": 514.62890625,
      "completions/mean_terminated_length": 516.6470947265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.1632,
      "grad_norm": 0.027405643835663795,
      "learning_rate": 1.3055555555555556e-06,
      "loss": 0.0451,
      "num_tokens": 33907579.0,
      "reward": 1.175753116607666,
      "reward_std": 0.24725493788719177,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5858250260353088,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 153
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.40159171024834e-06,
      "aux_brier/mean_group_std": 0.049404520346966,
      "aux_brier/mean_r": 0.9571009186646207,
      "aux_brier/n_active_tok": 245.375,
      "aux_brier/n_groups": 12.875,
      "aux_brier/n_step_records": 61.34375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5975159554498811,
      "calib/avg_num_step_conf": 7.94140625,
      "calib/ece": 0.32177865612648215,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.015465523714178503,
      "calib/mean_conf": 0.16043478260869568,
      "calib/mu_c": 0.16844262295081971,
      "calib/mu_w": 0.1529770992366412,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.0448472230430833,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1195.0,
      "completions/max_terminated_length": 1195.0,
      "completions/mean_length": 457.48046875,
      "completions/mean_terminated_length": 461.0826721191406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 183.0,
      "epoch": 0.16426666666666667,
      "grad_norm": 0.029376503080129623,
      "learning_rate": 1.2777777777777779e-06,
      "loss": -0.0002,
      "num_tokens": 34129134.0,
      "reward": 1.1298280954360962,
      "reward_std": 0.2767148017883301,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.6443125009536743,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 154
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.8444394972449984e-07,
      "aux_brier/mean_group_std": 0.06876151737155475,
      "aux_brier/mean_r": 0.9484107735269428,
      "aux_brier/n_active_tok": 236.875,
      "aux_brier/n_groups": 12.6875,
      "aux_brier/n_step_records": 59.21875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6465105779153766,
      "calib/avg_num_step_conf": 7.46484375,
      "calib/ece": 0.3039599999999999,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.032853457172342615,
      "calib/mean_conf": 0.15204,
      "calib/mu_c": 0.16991228070175438,
      "calib/mu_w": 0.13705882352941176,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.06402060918173147,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1101.0,
      "completions/max_terminated_length": 1101.0,
      "completions/mean_length": 443.41796875,
      "completions/mean_terminated_length": 445.1568908691406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 190.0,
      "epoch": 0.16533333333333333,
      "grad_norm": 0.012148459441959858,
      "learning_rate": 1.25e-06,
      "loss": -0.0043,
      "num_tokens": 34349865.0,
      "reward": 1.1015002727508545,
      "reward_std": 0.23070210218429565,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.6560012102127075,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 155
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.743661991517655e-08,
      "aux_brier/mean_group_std": 0.04543403264020979,
      "aux_brier/mean_r": 0.9614790204493904,
      "aux_brier/n_active_tok": 259.75,
      "aux_brier/n_groups": 14.40625,
      "aux_brier/n_step_records": 64.9375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5598499061913697,
      "calib/avg_num_step_conf": 8.21484375,
      "calib/ece": 0.3351383399209486,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.01092432770481555,
      "calib/mean_conf": 0.15869565217391307,
      "calib/mu_c": 0.16430894308943092,
      "calib/mu_w": 0.15338461538461537,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0038339920948616594,
      "calib/std_conf": 0.05606565762823203,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2092.0,
      "completions/max_terminated_length": 2092.0,
      "completions/mean_length": 496.07421875,
      "completions/mean_terminated_length": 499.9803161621094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 154.0,
      "epoch": 0.1664,
      "grad_norm": 0.016778623685240746,
      "learning_rate": 1.2222222222222223e-06,
      "loss": -0.0275,
      "num_tokens": 34581620.0,
      "reward": 1.1311582326889038,
      "reward_std": 0.2194337248802185,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.6340078115463257,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 156
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.2148936446298997e-06,
      "aux_brier/mean_group_std": 0.07013445319637046,
      "aux_brier/mean_r": 0.9377018502358946,
      "aux_brier/n_active_tok": 285.875,
      "aux_brier/n_groups": 18.96875,
      "aux_brier/n_step_records": 71.46875,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5923857216230097,
      "calib/avg_num_step_conf": 9.10546875,
      "calib/ece": 0.36272,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.008,
      "calib/gap": 0.016565228556753964,
      "calib/mean_conf": 0.17256,
      "calib/mu_c": 0.1803787878787879,
      "calib/mu_w": 0.16381355932203392,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0036399999999999996,
      "calib/std_conf": 0.08917312599656917,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2753.0,
      "completions/max_terminated_length": 2753.0,
      "completions/mean_length": 517.30078125,
      "completions/mean_terminated_length": 521.3740234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 129.0,
      "epoch": 0.16746666666666668,
      "grad_norm": 0.012949691154062748,
      "learning_rate": 1.1944444444444446e-06,
      "loss": 0.0313,
      "num_tokens": 34817777.0,
      "reward": 1.1535099744796753,
      "reward_std": 0.25031453371047974,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.6062273979187012,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 157
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.4456085095648774e-07,
      "aux_brier/mean_group_std": 0.049263813028915936,
      "aux_brier/mean_r": 0.9545683363001688,
      "aux_brier/n_active_tok": 250.375,
      "aux_brier/n_groups": 13.6875,
      "aux_brier/n_step_records": 62.59375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5115191017036654,
      "calib/avg_num_step_conf": 8.00390625,
      "calib/ece": 0.4301185770750989,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0010886680433660756,
      "calib/mean_conf": 0.15881422924901187,
      "calib/mu_c": 0.15926174496644296,
      "calib/mu_w": 0.1581730769230769,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05077440879067236,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2420.0,
      "completions/max_terminated_length": 2420.0,
      "completions/mean_length": 466.625,
      "completions/mean_terminated_length": 468.4549255371094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 148.0,
      "epoch": 0.16853333333333334,
      "grad_norm": 0.01752115599811077,
      "learning_rate": 1.1666666666666668e-06,
      "loss": 0.0593,
      "num_tokens": 35042473.0,
      "reward": 1.2172133922576904,
      "reward_std": 0.2381628155708313,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.5641664266586304,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 158
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.8741938578870077e-07,
      "aux_brier/mean_group_std": 0.0356345215848251,
      "aux_brier/mean_r": 0.9655458395128108,
      "aux_brier/n_active_tok": 270.25,
      "aux_brier/n_groups": 17.3125,
      "aux_brier/n_step_records": 67.5625,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4752791068580542,
      "calib/avg_num_step_conf": 8.69140625,
      "calib/ece": 0.39800813008130076,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.007448165869218509,
      "calib/mean_conf": 0.1445121951219512,
      "calib/mu_c": 0.14106060606060605,
      "calib/mu_w": 0.14850877192982456,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0029674796747967483,
      "calib/std_conf": 0.05210571896481207,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3057.0,
      "completions/max_terminated_length": 3057.0,
      "completions/mean_length": 508.8125,
      "completions/mean_terminated_length": 510.807861328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.1696,
      "grad_norm": 0.017778778448700905,
      "learning_rate": 1.138888888888889e-06,
      "loss": 0.1077,
      "num_tokens": 35277513.0,
      "reward": 1.1381198167800903,
      "reward_std": 0.29481813311576843,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.5681042671203613,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 159
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.1538541261634236e-07,
      "aux_brier/mean_group_std": 0.03308633600277228,
      "aux_brier/mean_r": 0.9712904034975596,
      "aux_brier/n_active_tok": 282.5,
      "aux_brier/n_groups": 16.03125,
      "aux_brier/n_step_records": 70.625,
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5847333333333333,
      "calib/avg_num_step_conf": 9.5625,
      "calib/ece": 0.36269387755102034,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009526666666666628,
      "calib/mean_conf": 0.1586938775510204,
      "calib/mu_c": 0.16335999999999998,
      "calib/mu_w": 0.15383333333333335,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.005591836734693879,
      "calib/std_conf": 0.05835786993134501,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2456.0,
      "completions/max_terminated_length": 2456.0,
      "completions/mean_length": 486.5390625,
      "completions/mean_terminated_length": 498.2160339355469,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 172.0,
      "epoch": 0.17066666666666666,
      "grad_norm": 0.0668088048696518,
      "learning_rate": 1.111111111111111e-06,
      "loss": -0.0019,
      "num_tokens": 35506907.0,
      "reward": 1.1119287014007568,
      "reward_std": 0.29088109731674194,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.5961523652076721,
      "rewards/format_reward_step_strict": 0.94921875,
      "step": 160
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.257408504042658e-07,
      "aux_brier/mean_group_std": 0.060142646147707056,
      "aux_brier/mean_r": 0.9541991361035498,
      "aux_brier/n_active_tok": 252.25,
      "aux_brier/n_groups": 14.59375,
      "aux_brier/n_step_records": 63.0625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5834820101100209,
      "calib/avg_num_step_conf": 7.9921875,
      "calib/ece": 0.5429249011857709,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009742789176330674,
      "calib/mean_conf": 0.16023715415019762,
      "calib/mu_c": 0.16316384180790963,
      "calib/mu_w": 0.15342105263157896,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0017786561264822134,
      "calib/std_conf": 0.05992042716721184,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2286.0,
      "completions/max_terminated_length": 2286.0,
      "completions/mean_length": 474.49609375,
      "completions/mean_terminated_length": 476.3569030761719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.17173333333333332,
      "grad_norm": 0.012965511530637741,
      "learning_rate": 1.0833333333333335e-06,
      "loss": 0.0217,
      "num_tokens": 35732298.0,
      "reward": 1.3089410066604614,
      "reward_std": 0.20046359300613403,
      "rewards/accuracy_reward_step": 0.69140625,
      "rewards/final_brier_reward_step": 0.4935765564441681,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 161
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.7168893418073594e-07,
      "aux_brier/mean_group_std": 0.04964035382205659,
      "aux_brier/mean_r": 0.9612997269895767,
      "aux_brier/n_active_tok": 244.25,
      "aux_brier/n_groups": 12.75,
      "aux_brier/n_step_records": 61.0625,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.49456446475557403,
      "calib/avg_num_step_conf": 7.9375,
      "calib/ece": 0.5091304347826087,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004753496745603103,
      "calib/mean_conf": 0.14699604743083003,
      "calib/mu_c": 0.14536144578313254,
      "calib/mu_w": 0.15011494252873564,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04871133801057367,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2237.0,
      "completions/max_terminated_length": 2237.0,
      "completions/mean_length": 468.48046875,
      "completions/mean_terminated_length": 470.31768798828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 154.0,
      "epoch": 0.1728,
      "grad_norm": 0.034487757831811905,
      "learning_rate": 1.0555555555555557e-06,
      "loss": -0.0076,
      "num_tokens": 35956373.0,
      "reward": 1.2667899131774902,
      "reward_std": 0.23064537346363068,
      "rewards/accuracy_reward_step": 0.6484375,
      "rewards/final_brier_reward_step": 0.5046597719192505,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 162
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 4.471214591905692e-07,
      "aux_brier/mean_group_std": 0.05555076010570242,
      "aux_brier/mean_r": 0.9525882588155883,
      "aux_brier/n_active_tok": 308.0,
      "aux_brier/n_groups": 22.03125,
      "aux_brier/n_step_records": 77.0,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.518890881147541,
      "calib/avg_num_step_conf": 9.6484375,
      "calib/ece": 0.35147999999999996,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0071426741803278515,
      "calib/mean_conf": 0.16627999999999998,
      "calib/mu_c": 0.169765625,
      "calib/mu_w": 0.16262295081967215,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0028799999999999997,
      "calib/std_conf": 0.06980087105473684,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3003.0,
      "completions/max_terminated_length": 3003.0,
      "completions/mean_length": 577.58984375,
      "completions/mean_terminated_length": 579.8549194335938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 156.0,
      "epoch": 0.17386666666666667,
      "grad_norm": 0.08797299116849899,
      "learning_rate": 1.0277777777777777e-06,
      "loss": 0.1231,
      "num_tokens": 36209068.0,
      "reward": 1.1419235467910767,
      "reward_std": 0.2853378355503082,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.6145691871643066,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 163
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.9866855716088594e-07,
      "aux_brier/mean_group_std": 0.03667347612396784,
      "aux_brier/mean_r": 0.9707665289826557,
      "aux_brier/n_active_tok": 275.25,
      "aux_brier/n_groups": 15.625,
      "aux_brier/n_step_records": 68.8125,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5839753466872111,
      "calib/avg_num_step_conf": 8.73046875,
      "calib/ece": 0.3771599999999999,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.012505136106830989,
      "calib/mean_conf": 0.15084,
      "calib/mu_c": 0.15674242424242424,
      "calib/mu_w": 0.14423728813559325,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05185069334155523,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2610.0,
      "completions/max_terminated_length": 2610.0,
      "completions/mean_length": 547.1640625,
      "completions/mean_terminated_length": 551.472412109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 209.0,
      "epoch": 0.17493333333333333,
      "grad_norm": 0.057448893785476685,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0145,
      "num_tokens": 36455278.0,
      "reward": 1.1483054161071777,
      "reward_std": 0.26751455664634705,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.5932214856147766,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 164
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.2906746038598982e-07,
      "aux_brier/mean_group_std": 0.03551182583994825,
      "aux_brier/mean_r": 0.9717435124993495,
      "aux_brier/n_active_tok": 271.5,
      "aux_brier/n_groups": 16.6875,
      "aux_brier/n_step_records": 67.875,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5469710272168569,
      "calib/avg_num_step_conf": 8.51953125,
      "calib/ece": 0.3275494071146245,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.003952569169960474,
      "calib/gap": 0.000681048538818535,
      "calib/mean_conf": 0.15972332015810276,
      "calib/mu_c": 0.1600840336134454,
      "calib/mu_w": 0.15940298507462686,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.008458498023715413,
      "calib/std_conf": 0.07853602764329508,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1706.0,
      "completions/max_terminated_length": 1706.0,
      "completions/mean_length": 508.46875,
      "completions/mean_terminated_length": 510.4627685546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.176,
      "grad_norm": 0.050327152013778687,
      "learning_rate": 9.722222222222224e-07,
      "loss": -0.0101,
      "num_tokens": 36691022.0,
      "reward": 1.123129963874817,
      "reward_std": 0.2538028359413147,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6409574151039124,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 165
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -4.193520997120004e-08,
      "aux_brier/mean_group_std": 0.05831487297588019,
      "aux_brier/mean_r": 0.9546354778008198,
      "aux_brier/n_active_tok": 282.5,
      "aux_brier/n_groups": 17.5,
      "aux_brier/n_step_records": 70.625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5058467741935484,
      "calib/avg_num_step_conf": 8.90234375,
      "calib/ece": 0.4602390438247012,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0003360215053763438,
      "calib/mean_conf": 0.15729083665338647,
      "calib/mu_c": 0.15741935483870967,
      "calib/mu_w": 0.15708333333333332,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.052162700146282756,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2893.0,
      "completions/max_terminated_length": 2893.0,
      "completions/mean_length": 551.71484375,
      "completions/mean_terminated_length": 553.8784790039062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 181.0,
      "epoch": 0.17706666666666668,
      "grad_norm": 0.03957146778702736,
      "learning_rate": 9.444444444444445e-07,
      "loss": 0.0656,
      "num_tokens": 36938445.0,
      "reward": 1.2245346307754517,
      "reward_std": 0.27831125259399414,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.5309507846832275,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 166
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -4.390739093695917e-07,
      "aux_brier/mean_group_std": 0.058610101023658656,
      "aux_brier/mean_r": 0.9450262725855368,
      "aux_brier/n_active_tok": 281.75,
      "aux_brier/n_groups": 19.03125,
      "aux_brier/n_step_records": 70.4375,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.6231026785714285,
      "calib/avg_num_step_conf": 8.90234375,
      "calib/ece": 0.5113709677419356,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.01963095238095236,
      "calib/mean_conf": 0.1660483870967742,
      "calib/mu_c": 0.17238095238095238,
      "calib/mu_w": 0.15275000000000002,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0,
      "calib/std_conf": 0.061445209640143016,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2698.0,
      "completions/max_terminated_length": 2698.0,
      "completions/mean_length": 531.41015625,
      "completions/mean_terminated_length": 539.8452758789062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 196.0,
      "epoch": 0.17813333333333334,
      "grad_norm": 0.06966294348239899,
      "learning_rate": 9.166666666666666e-07,
      "loss": -0.018,
      "num_tokens": 37180094.0,
      "reward": 1.259904146194458,
      "reward_std": 0.23908957839012146,
      "rewards/accuracy_reward_step": 0.65625,
      "rewards/final_brier_reward_step": 0.5005543231964111,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 167
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.1016560679265606e-08,
      "aux_brier/mean_group_std": 0.07045650564183857,
      "aux_brier/mean_r": 0.9434491784728122,
      "aux_brier/n_active_tok": 259.75,
      "aux_brier/n_groups": 13.59375,
      "aux_brier/n_step_records": 64.9375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.42066115702479334,
      "calib/avg_num_step_conf": 8.31640625,
      "calib/ece": 0.4157312252964427,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.015405594405594386,
      "calib/mean_conf": 0.15501976284584978,
      "calib/mu_c": 0.14832167832167834,
      "calib/mu_w": 0.16372727272727272,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0027667984189723317,
      "calib/std_conf": 0.046127718433160075,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3038.0,
      "completions/max_terminated_length": 3038.0,
      "completions/mean_length": 534.515625,
      "completions/mean_terminated_length": 536.61181640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 180.0,
      "epoch": 0.1792,
      "grad_norm": 0.0411788746714592,
      "learning_rate": 8.88888888888889e-07,
      "loss": 0.0105,
      "num_tokens": 37421602.0,
      "reward": 1.1951189041137695,
      "reward_std": 0.2917238175868988,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.5695382952690125,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 168
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.2797146142773386e-07,
      "aux_brier/mean_group_std": 0.0398169424234487,
      "aux_brier/mean_r": 0.9645761018675307,
      "aux_brier/n_active_tok": 259.125,
      "aux_brier/n_groups": 14.90625,
      "aux_brier/n_step_records": 64.78125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5461334845906598,
      "calib/avg_num_step_conf": 8.22265625,
      "calib/ece": 0.3644444444444445,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.007585554925316679,
      "calib/mean_conf": 0.14746031746031749,
      "calib/mu_c": 0.1511627906976744,
      "calib/mu_w": 0.14357723577235773,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05206736338684675,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3066.0,
      "completions/max_terminated_length": 3066.0,
      "completions/mean_length": 527.26953125,
      "completions/mean_terminated_length": 529.3372802734375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 101.0,
      "epoch": 0.18026666666666666,
      "grad_norm": 0.024509698152542114,
      "learning_rate": 8.611111111111112e-07,
      "loss": 0.0458,
      "num_tokens": 37660767.0,
      "reward": 1.1482784748077393,
      "reward_std": 0.23623031377792358,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.6087390780448914,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 169
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -7.465195546729753e-07,
      "aux_brier/mean_group_std": 0.04840349667762759,
      "aux_brier/mean_r": 0.9621910303766671,
      "aux_brier/n_active_tok": 280.25,
      "aux_brier/n_groups": 14.96875,
      "aux_brier/n_step_records": 70.0625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5712414542201776,
      "calib/avg_num_step_conf": 8.7578125,
      "calib/ece": 0.39841269841269844,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.010841479777649987,
      "calib/mean_conf": 0.1611111111111111,
      "calib/mu_c": 0.16588652482269503,
      "calib/mu_w": 0.15504504504504504,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05118007769370506,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2377.0,
      "completions/max_terminated_length": 2377.0,
      "completions/mean_length": 535.265625,
      "completions/mean_terminated_length": 537.36474609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 191.0,
      "epoch": 0.18133333333333335,
      "grad_norm": 0.056401584297418594,
      "learning_rate": 8.333333333333333e-07,
      "loss": 0.0319,
      "num_tokens": 37901947.0,
      "reward": 1.1900184154510498,
      "reward_std": 0.23777510225772858,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.5881984233856201,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 170
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.491154891786195e-07,
      "aux_brier/mean_group_std": 0.06174421707246917,
      "aux_brier/mean_r": 0.9496703379452144,
      "aux_brier/n_active_tok": 265.0,
      "aux_brier/n_groups": 14.125,
      "aux_brier/n_step_records": 66.25,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5812636165577342,
      "calib/avg_num_step_conf": 8.38671875,
      "calib/ece": 0.31582677165354334,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0152007469654529,
      "calib/mean_conf": 0.1526771653543307,
      "calib/mu_c": 0.16075630252100845,
      "calib/mu_w": 0.14555555555555555,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05233069918321514,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1124.0,
      "completions/max_terminated_length": 1124.0,
      "completions/mean_length": 494.9375,
      "completions/mean_terminated_length": 496.8784484863281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 142.0,
      "epoch": 0.1824,
      "grad_norm": 0.04185402765870094,
      "learning_rate": 8.055555555555557e-07,
      "loss": -0.0002,
      "num_tokens": 38135547.0,
      "reward": 1.1236753463745117,
      "reward_std": 0.27767056226730347,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.6509515047073364,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 171
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -4.3899531909641354e-07,
      "aux_brier/mean_group_std": 0.05103582146149051,
      "aux_brier/mean_r": 0.9617057692948487,
      "aux_brier/n_active_tok": 267.125,
      "aux_brier/n_groups": 15.5625,
      "aux_brier/n_step_records": 66.78125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5143003064351379,
      "calib/avg_num_step_conf": 8.44921875,
      "calib/ece": 0.5061417322834646,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0026830098740211195,
      "calib/mean_conf": 0.14881889763779527,
      "calib/mu_c": 0.14787878787878786,
      "calib/mu_w": 0.15056179775280898,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0026771653543307085,
      "calib/std_conf": 0.05601763183097044,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1908.0,
      "completions/max_terminated_length": 1908.0,
      "completions/mean_length": 495.859375,
      "completions/mean_terminated_length": 497.803955078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 171.0,
      "epoch": 0.18346666666666667,
      "grad_norm": 0.0242488831281662,
      "learning_rate": 7.777777777777779e-07,
      "loss": -0.0376,
      "num_tokens": 38365839.0,
      "reward": 1.2689235210418701,
      "reward_std": 0.22546443343162537,
      "rewards/accuracy_reward_step": 0.64453125,
      "rewards/final_brier_reward_step": 0.5131937265396118,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 172
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 4.995026681164916e-07,
      "aux_brier/mean_group_std": 0.059560432252318406,
      "aux_brier/mean_r": 0.9545917724846392,
      "aux_brier/n_active_tok": 280.75,
      "aux_brier/n_groups": 18.46875,
      "aux_brier/n_step_records": 70.1875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6645595023078467,
      "calib/avg_num_step_conf": 8.7734375,
      "calib/ece": 0.44155999999999995,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.004,
      "calib/gap": 0.03430931834905343,
      "calib/mean_conf": 0.16244,
      "calib/mu_c": 0.17602649006622514,
      "calib/mu_w": 0.1417171717171717,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.07389753987786062,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2756.0,
      "completions/max_terminated_length": 2756.0,
      "completions/mean_length": 537.3359375,
      "completions/mean_terminated_length": 539.4431762695312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 147.0,
      "epoch": 0.18453333333333333,
      "grad_norm": 0.057891733944416046,
      "learning_rate": 7.5e-07,
      "loss": 0.0231,
      "num_tokens": 38606557.0,
      "reward": 1.222849726676941,
      "reward_std": 0.2555616497993469,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.5632737874984741,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 173
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 4.628371158554856e-07,
      "aux_brier/mean_group_std": 0.06359103303081172,
      "aux_brier/mean_r": 0.9525573415400864,
      "aux_brier/n_active_tok": 277.0,
      "aux_brier/n_groups": 15.96875,
      "aux_brier/n_step_records": 69.25,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5182238843813387,
      "calib/avg_num_step_conf": 8.65625,
      "calib/ece": 0.3131349206349206,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0010547667342798994,
      "calib/mean_conf": 0.1528968253968254,
      "calib/mu_c": 0.15232758620689657,
      "calib/mu_w": 0.15338235294117647,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0028571428571428567,
      "calib/std_conf": 0.048810669105158994,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2637.0,
      "completions/max_terminated_length": 2637.0,
      "completions/mean_length": 565.14453125,
      "completions/mean_terminated_length": 567.36083984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 221.0,
      "epoch": 0.1856,
      "grad_norm": 0.04420608654618263,
      "learning_rate": 7.222222222222222e-07,
      "loss": 0.0278,
      "num_tokens": 38855466.0,
      "reward": 1.102353811264038,
      "reward_std": 0.29684698581695557,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.6359777450561523,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 174
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -9.875406990489388e-07,
      "aux_brier/mean_group_std": 0.05294872455716927,
      "aux_brier/mean_r": 0.9574670859832486,
      "aux_brier/n_active_tok": 265.875,
      "aux_brier/n_groups": 16.09375,
      "aux_brier/n_step_records": 66.46875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5250064699792961,
      "calib/avg_num_step_conf": 8.51953125,
      "calib/ece": 0.31059999999999993,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004482401656314722,
      "calib/mean_conf": 0.14324,
      "calib/mu_c": 0.1457142857142857,
      "calib/mu_w": 0.141231884057971,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.00292,
      "calib/std_conf": 0.06464288359904748,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2578.0,
      "completions/max_terminated_length": 2578.0,
      "completions/mean_length": 533.72265625,
      "completions/mean_terminated_length": 540.0513916015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.18666666666666668,
      "grad_norm": 0.07568599283695221,
      "learning_rate": 6.944444444444446e-07,
      "loss": -0.0386,
      "num_tokens": 39097923.0,
      "reward": 1.0815097093582153,
      "reward_std": 0.30057209730148315,
      "rewards/accuracy_reward_step": 0.4375,
      "rewards/final_brier_reward_step": 0.6385387182235718,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 175
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.1575440644895352e-07,
      "aux_brier/mean_group_std": 0.08753972842421873,
      "aux_brier/mean_r": 0.9381399282236724,
      "aux_brier/n_active_tok": 283.25,
      "aux_brier/n_groups": 17.0625,
      "aux_brier/n_step_records": 70.8125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5977443609022557,
      "calib/avg_num_step_conf": 8.8515625,
      "calib/ece": 0.32285714285714284,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.013471915081822244,
      "calib/mean_conf": 0.15373015873015872,
      "calib/mu_c": 0.16084033613445384,
      "calib/mu_w": 0.1473684210526316,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0021825396825396826,
      "calib/std_conf": 0.05098605671494791,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2717.0,
      "completions/max_terminated_length": 2717.0,
      "completions/mean_length": 521.390625,
      "completions/mean_terminated_length": 521.390625,
      "completions/min_length": 171.0,
      "completions/min_terminated_length": 171.0,
      "epoch": 0.18773333333333334,
      "grad_norm": 0.054171666502952576,
      "learning_rate": 6.666666666666667e-07,
      "loss": 0.0236,
      "num_tokens": 39335463.0,
      "reward": 1.1178412437438965,
      "reward_std": 0.2490749955177307,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.6432398557662964,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 176
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.689186719504136e-07,
      "aux_brier/mean_group_std": 0.04862063620140382,
      "aux_brier/mean_r": 0.9607684011459631,
      "aux_brier/n_active_tok": 269.75,
      "aux_brier/n_groups": 16.375,
      "aux_brier/n_step_records": 67.4375,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.4508516905737705,
      "calib/avg_num_step_conf": 8.51171875,
      "calib/ece": 0.38048,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.004,
      "calib/gap": -0.0166265368852459,
      "calib/mean_conf": 0.14952,
      "calib/mu_c": 0.14140625,
      "calib/mu_w": 0.1580327868852459,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.009000000000000001,
      "calib/std_conf": 0.07035744168174395,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2750.0,
      "completions/max_terminated_length": 2750.0,
      "completions/mean_length": 538.55859375,
      "completions/mean_terminated_length": 542.7991943359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 223.0,
      "epoch": 0.1888,
      "grad_norm": 0.045605529099702835,
      "learning_rate": 6.388888888888889e-07,
      "loss": 0.0226,
      "num_tokens": 39577166.0,
      "reward": 1.1320149898529053,
      "reward_std": 0.2338375747203827,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.5905601382255554,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 177
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -2.3104294180620322e-07,
      "aux_brier/mean_group_std": 0.06723367407909693,
      "aux_brier/mean_r": 0.9492157188413886,
      "aux_brier/n_active_tok": 267.5,
      "aux_brier/n_groups": 14.9375,
      "aux_brier/n_step_records": 66.875,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5969959946595461,
      "calib/avg_num_step_conf": 9.0078125,
      "calib/ece": 0.4315384615384616,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009666889185580785,
      "calib/mean_conf": 0.14174089068825912,
      "calib/mu_c": 0.14592857142857144,
      "calib/mu_w": 0.13626168224299065,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0032388663967611335,
      "calib/std_conf": 0.0500182726531682,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2538.0,
      "completions/max_terminated_length": 2538.0,
      "completions/mean_length": 497.9921875,
      "completions/mean_terminated_length": 507.912353515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.18986666666666666,
      "grad_norm": 0.06169963628053665,
      "learning_rate": 6.111111111111112e-07,
      "loss": 0.0261,
      "num_tokens": 39810724.0,
      "reward": 1.1655563116073608,
      "reward_std": 0.2794933617115021,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.5528503656387329,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 178
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.433350124147026e-07,
      "aux_brier/mean_group_std": 0.05176466906621195,
      "aux_brier/mean_r": 0.9615216922126134,
      "aux_brier/n_active_tok": 264.625,
      "aux_brier/n_groups": 13.90625,
      "aux_brier/n_step_records": 66.15625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5617105263157894,
      "calib/avg_num_step_conf": 8.41796875,
      "calib/ece": 0.46158730158730155,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00498157894736842,
      "calib/mean_conf": 0.1469047619047619,
      "calib/mu_c": 0.14888157894736842,
      "calib/mu_w": 0.1439,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.002658730158730158,
      "calib/std_conf": 0.04943272301126441,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1943.0,
      "completions/max_terminated_length": 1943.0,
      "completions/mean_length": 500.34765625,
      "completions/mean_terminated_length": 502.3098449707031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 198.0,
      "epoch": 0.19093333333333334,
      "grad_norm": 0.04521411284804344,
      "learning_rate": 5.833333333333334e-07,
      "loss": -0.0184,
      "num_tokens": 40045077.0,
      "reward": 1.2150800228118896,
      "reward_std": 0.2715832591056824,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.540007472038269,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 179
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.557433655350996e-07,
      "aux_brier/mean_group_std": 0.07278899981776597,
      "aux_brier/mean_r": 0.9373554568692914,
      "aux_brier/n_active_tok": 300.25,
      "aux_brier/n_groups": 17.125,
      "aux_brier/n_step_records": 75.0625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5727829543233206,
      "calib/avg_num_step_conf": 9.61328125,
      "calib/ece": 0.43198412698412697,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.011193067048934607,
      "calib/mean_conf": 0.16166666666666665,
      "calib/mu_c": 0.166241610738255,
      "calib/mu_w": 0.1550485436893204,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.0011904761904761904,
      "calib/std_conf": 0.04701823354523526,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2816.0,
      "completions/max_terminated_length": 2816.0,
      "completions/mean_length": 596.3359375,
      "completions/mean_terminated_length": 598.674560546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 209.0,
      "epoch": 0.192,
      "grad_norm": 0.01824566349387169,
      "learning_rate": 5.555555555555555e-07,
      "loss": -0.0015,
      "num_tokens": 40301595.0,
      "reward": 1.204345941543579,
      "reward_std": 0.22382229566574097,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.5517585873603821,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 180
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -3.7227140847750917e-07,
      "aux_brier/mean_group_std": 0.03710208709554556,
      "aux_brier/mean_r": 0.9678823743283032,
      "aux_brier/n_active_tok": 273.125,
      "aux_brier/n_groups": 17.25,
      "aux_brier/n_step_records": 68.28125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5787562125653116,
      "calib/avg_num_step_conf": 8.73828125,
      "calib/ece": 0.3343426294820717,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.012076589779533603,
      "calib/mean_conf": 0.14334661354581674,
      "calib/mu_c": 0.14974576271186443,
      "calib/mu_w": 0.13766917293233083,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.003784860557768924,
      "calib/std_conf": 0.056680724768170314,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2814.0,
      "completions/max_terminated_length": 2814.0,
      "completions/mean_length": 496.55078125,
      "completions/mean_terminated_length": 500.46063232421875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 194.0,
      "epoch": 0.19306666666666666,
      "grad_norm": 0.08360396325588226,
      "learning_rate": 5.277777777777779e-07,
      "loss": 0.0483,
      "num_tokens": 40534976.0,
      "reward": 1.104553461074829,
      "reward_std": 0.31083574891090393,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.6291515827178955,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 181
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.922425588816613e-07,
      "aux_brier/mean_group_std": 0.06789074830965955,
      "aux_brier/mean_r": 0.954944220972813,
      "aux_brier/n_active_tok": 277.875,
      "aux_brier/n_groups": 16.5625,
      "aux_brier/n_step_records": 69.46875,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5816562721018453,
      "calib/avg_num_step_conf": 9.1640625,
      "calib/ece": 0.4632283464566928,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0035594419083135043,
      "calib/mean_conf": 0.144251968503937,
      "calib/mu_c": 0.1456953642384106,
      "calib/mu_w": 0.1421359223300971,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.006496062992125986,
      "calib/std_conf": 0.0589076373674881,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1588.0,
      "completions/max_terminated_length": 1588.0,
      "completions/mean_length": 532.20703125,
      "completions/mean_terminated_length": 534.2941284179688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 203.0,
      "epoch": 0.19413333333333332,
      "grad_norm": 0.04704170301556587,
      "learning_rate": 5.000000000000001e-07,
      "loss": -0.0073,
      "num_tokens": 40777381.0,
      "reward": 1.220548152923584,
      "reward_std": 0.2288506031036377,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.5462551116943359,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 182
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 5.345288603963994e-07,
      "aux_brier/mean_group_std": 0.05264424985501321,
      "aux_brier/mean_r": 0.9604438890020913,
      "aux_brier/n_active_tok": 277.25,
      "aux_brier/n_groups": 16.46875,
      "aux_brier/n_step_records": 69.3125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5346075692151384,
      "calib/avg_num_step_conf": 8.87890625,
      "calib/ece": 0.35828685258964144,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00028956057912110644,
      "calib/mean_conf": 0.14211155378486057,
      "calib/mu_c": 0.14225806451612905,
      "calib/mu_w": 0.14196850393700794,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0031872509960159364,
      "calib/std_conf": 0.050652352981139896,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2832.0,
      "completions/max_terminated_length": 2832.0,
      "completions/mean_length": 558.50390625,
      "completions/mean_terminated_length": 560.6941528320312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.1952,
      "grad_norm": 0.02439642697572708,
      "learning_rate": 4.7222222222222226e-07,
      "loss": 0.0363,
      "num_tokens": 41027038.0,
      "reward": 1.1275067329406738,
      "reward_std": 0.28274720907211304,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.6115894317626953,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 183
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 3.103694012906999e-07,
      "aux_brier/mean_group_std": 0.06436204785975919,
      "aux_brier/mean_r": 0.9484624329261578,
      "aux_brier/n_active_tok": 295.25,
      "aux_brier/n_groups": 19.1875,
      "aux_brier/n_step_records": 73.8125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.4677537681739362,
      "calib/avg_num_step_conf": 9.53515625,
      "calib/ece": 0.45848605577689244,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.006673336001067065,
      "calib/mean_conf": 0.15776892430278885,
      "calib/mu_c": 0.15516339869281048,
      "calib/mu_w": 0.16183673469387755,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0033466135458167334,
      "calib/std_conf": 0.05640457857745456,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2917.0,
      "completions/max_terminated_length": 2917.0,
      "completions/mean_length": 547.421875,
      "completions/mean_terminated_length": 551.7322998046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 178.0,
      "epoch": 0.19626666666666667,
      "grad_norm": 0.03321665897965431,
      "learning_rate": 4.444444444444445e-07,
      "loss": 0.02,
      "num_tokens": 41272458.0,
      "reward": 1.2182791233062744,
      "reward_std": 0.31848832964897156,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.5371793508529663,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 184
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.4893604744224973e-07,
      "aux_brier/mean_group_std": 0.06732063692971484,
      "aux_brier/mean_r": 0.9515054704357953,
      "aux_brier/n_active_tok": 312.0,
      "aux_brier/n_groups": 21.53125,
      "aux_brier/n_step_records": 78.0,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5551517598761694,
      "calib/avg_num_step_conf": 10.41796875,
      "calib/ece": 0.36415573770491805,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.008297866612827276,
      "calib/mean_conf": 0.1563360655737705,
      "calib/mu_c": 0.1603149606299213,
      "calib/mu_w": 0.152017094017094,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.0536060478995191,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2945.0,
      "completions/max_terminated_length": 2945.0,
      "completions/mean_length": 563.51953125,
      "completions/mean_terminated_length": 574.7450561523438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.19733333333333333,
      "grad_norm": 0.08410869538784027,
      "learning_rate": 4.1666666666666667e-07,
      "loss": 0.0669,
      "num_tokens": 41523639.0,
      "reward": 1.1133431196212769,
      "reward_std": 0.2527396082878113,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5861849784851074,
      "rewards/format_reward_step_strict": 0.94140625,
      "step": 185
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.1460473081648104e-07,
      "aux_brier/mean_group_std": 0.052871889902070625,
      "aux_brier/mean_r": 0.9604263236935935,
      "aux_brier/n_active_tok": 290.875,
      "aux_brier/n_groups": 16.5,
      "aux_brier/n_step_records": 72.71875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6115372220431841,
      "calib/avg_num_step_conf": 9.45703125,
      "calib/ece": 0.4247222222222222,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.01968868836609733,
      "calib/mean_conf": 0.15067460317460318,
      "calib/mu_c": 0.1590344827586207,
      "calib/mu_w": 0.13934579439252337,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05303432990669396,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2657.0,
      "completions/max_terminated_length": 2657.0,
      "completions/mean_length": 542.7109375,
      "completions/mean_terminated_length": 546.9842529296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 193.0,
      "epoch": 0.1984,
      "grad_norm": 0.03420405834913254,
      "learning_rate": 3.8888888888888895e-07,
      "loss": 0.0136,
      "num_tokens": 41767613.0,
      "reward": 1.196563959121704,
      "reward_std": 0.2683338522911072,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.5675058364868164,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 186
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.9278789737130495e-07,
      "aux_brier/mean_group_std": 0.05675664947652703,
      "aux_brier/mean_r": 0.9555893828683435,
      "aux_brier/n_active_tok": 323.625,
      "aux_brier/n_groups": 24.78125,
      "aux_brier/n_step_records": 80.90625,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5007753505933118,
      "calib/avg_num_step_conf": 10.59375,
      "calib/ece": 0.4300404858299595,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0038592233009708543,
      "calib/mean_conf": 0.1591093117408907,
      "calib/mu_c": 0.1575,
      "calib/mu_w": 0.16135922330097086,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.003076923076923075,
      "calib/std_conf": 0.0782915771133442,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2369.0,
      "completions/max_terminated_length": 2369.0,
      "completions/mean_length": 568.6328125,
      "completions/mean_terminated_length": 577.6587524414062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 173.0,
      "epoch": 0.19946666666666665,
      "grad_norm": 0.08125673234462738,
      "learning_rate": 3.611111111111111e-07,
      "loss": -0.0061,
      "num_tokens": 42014727.0,
      "reward": 1.179487705230713,
      "reward_std": 0.32495832443237305,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.5460761785507202,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 187
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.0506600884307993e-06,
      "aux_brier/mean_group_std": 0.06685216259607211,
      "aux_brier/mean_r": 0.948845332300965,
      "aux_brier/n_active_tok": 284.875,
      "aux_brier/n_groups": 16.03125,
      "aux_brier/n_step_records": 71.21875,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5872108843537416,
      "calib/avg_num_step_conf": 9.54296875,
      "calib/ece": 0.4564516129032259,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.013360544217687065,
      "calib/mean_conf": 0.14838709677419357,
      "calib/mu_c": 0.15366666666666665,
      "calib/mu_w": 0.14030612244897958,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.0503197373059421,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2482.0,
      "completions/max_terminated_length": 2482.0,
      "completions/mean_length": 535.859375,
      "completions/mean_terminated_length": 546.5338745117188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 190.0,
      "epoch": 0.20053333333333334,
      "grad_norm": 0.11312317103147507,
      "learning_rate": 3.3333333333333335e-07,
      "loss": -0.0073,
      "num_tokens": 42255979.0,
      "reward": 1.2021760940551758,
      "reward_std": 0.27750808000564575,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.5352667570114136,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 188
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -7.452008510544506e-07,
      "aux_brier/mean_group_std": 0.06444877414433915,
      "aux_brier/mean_r": 0.9399538708497756,
      "aux_brier/n_active_tok": 261.875,
      "aux_brier/n_groups": 14.21875,
      "aux_brier/n_step_records": 65.46875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6512432795698925,
      "calib/avg_num_step_conf": 8.33984375,
      "calib/ece": 0.4774501992031872,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.019141801075268833,
      "calib/mean_conf": 0.14254980079681276,
      "calib/mu_c": 0.14987096774193548,
      "calib/mu_w": 0.13072916666666665,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0012350597609561752,
      "calib/std_conf": 0.042910456544259845,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1524.0,
      "completions/max_terminated_length": 1524.0,
      "completions/mean_length": 498.23046875,
      "completions/mean_terminated_length": 500.1843566894531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.2016,
      "grad_norm": 0.03259199112653732,
      "learning_rate": 3.055555555555556e-07,
      "loss": -0.0372,
      "num_tokens": 42491294.0,
      "reward": 1.2293919324874878,
      "reward_std": 0.23676735162734985,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.5347554683685303,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 189
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.5741437181526763e-08,
      "aux_brier/mean_group_std": 0.07738672575917957,
      "aux_brier/mean_r": 0.9435339355484508,
      "aux_brier/n_active_tok": 299.0,
      "aux_brier/n_groups": 18.1875,
      "aux_brier/n_step_records": 74.75,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5185710515899196,
      "calib/avg_num_step_conf": 9.6875,
      "calib/ece": 0.4170682730923695,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.010780445969125252,
      "calib/mean_conf": 0.15722891566265063,
      "calib/mu_c": 0.16181818181818183,
      "calib/mu_w": 0.15103773584905658,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.07050557390045874,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2984.0,
      "completions/max_terminated_length": 2984.0,
      "completions/mean_length": 585.86328125,
      "completions/mean_terminated_length": 590.4763793945312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 210.0,
      "epoch": 0.20266666666666666,
      "grad_norm": 0.02406657487154007,
      "learning_rate": 2.7777777777777776e-07,
      "loss": 0.0149,
      "num_tokens": 42746883.0,
      "reward": 1.1835261583328247,
      "reward_std": 0.2722218930721283,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.5622296929359436,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 190
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.2340974625834065e-07,
      "aux_brier/mean_group_std": 0.037090143465623206,
      "aux_brier/mean_r": 0.96761951313658,
      "aux_brier/n_active_tok": 301.875,
      "aux_brier/n_groups": 15.8125,
      "aux_brier/n_step_records": 75.46875,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6338861788617887,
      "calib/avg_num_step_conf": 9.60546875,
      "calib/ece": 0.3501612903225806,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.01989853658536586,
      "calib/mean_conf": 0.15637096774193548,
      "calib/mu_c": 0.16624,
      "calib/mu_w": 0.14634146341463414,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00125,
      "calib/std_conf": 0.04924158152086891,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2871.0,
      "completions/max_terminated_length": 2871.0,
      "completions/mean_length": 555.515625,
      "completions/mean_terminated_length": 559.8897705078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 166.0,
      "epoch": 0.20373333333333332,
      "grad_norm": 0.03271899372339249,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0392,
      "num_tokens": 42993263.0,
      "reward": 1.1268501281738281,
      "reward_std": 0.2318863868713379,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.6167757511138916,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 191
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -1.861136611491787e-07,
      "aux_brier/mean_group_std": 0.055214087686892996,
      "aux_brier/mean_r": 0.9545805860948601,
      "aux_brier/n_active_tok": 285.125,
      "aux_brier/n_groups": 16.9375,
      "aux_brier/n_step_records": 71.28125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6114485075591161,
      "calib/avg_num_step_conf": 9.3046875,
      "calib/ece": 0.4221513944223108,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.016469182064866222,
      "calib/mean_conf": 0.14390438247011955,
      "calib/mu_c": 0.151056338028169,
      "calib/mu_w": 0.13458715596330278,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.00015936254980079717,
      "calib/std_conf": 0.0577387199510932,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2050.0,
      "completions/max_terminated_length": 2050.0,
      "completions/mean_length": 556.12109375,
      "completions/mean_terminated_length": 560.5,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 169.0,
      "epoch": 0.2048,
      "grad_norm": 0.05804295092821121,
      "learning_rate": 2.2222222222222224e-07,
      "loss": 0.0509,
      "num_tokens": 43240606.0,
      "reward": 1.181755781173706,
      "reward_std": 0.28981447219848633,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.5629605054855347,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 192
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -6.803905278074396e-07,
      "aux_brier/mean_group_std": 0.0431717765249665,
      "aux_brier/mean_r": 0.9606889633754716,
      "aux_brier/n_active_tok": 309.75,
      "aux_brier/n_groups": 20.21875,
      "aux_brier/n_step_records": 77.4375,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.43971282454760036,
      "calib/avg_num_step_conf": 10.140625,
      "calib/ece": 0.36020242914979755,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.017089562024652505,
      "calib/mean_conf": 0.15947368421052632,
      "calib/mu_c": 0.15089430894308944,
      "calib/mu_w": 0.16798387096774195,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.010850202429149799,
      "calib/std_conf": 0.06391516324710692,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2719.0,
      "completions/max_terminated_length": 2719.0,
      "completions/mean_length": 590.11328125,
      "completions/mean_terminated_length": 597.1107177734375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 205.0,
      "epoch": 0.20586666666666667,
      "grad_norm": 0.03458983823657036,
      "learning_rate": 1.9444444444444447e-07,
      "loss": 0.0748,
      "num_tokens": 43497387.0,
      "reward": 1.1143351793289185,
      "reward_std": 0.34169265627861023,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.5979660749435425,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 193
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 1.0541671985331291e-06,
      "aux_brier/mean_group_std": 0.08417545681814177,
      "aux_brier/mean_r": 0.9387734846250257,
      "aux_brier/n_active_tok": 257.0,
      "aux_brier/n_groups": 13.90625,
      "aux_brier/n_step_records": 64.25,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6547619047619048,
      "calib/avg_num_step_conf": 8.24609375,
      "calib/ece": 0.4482868525896414,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.028838958660387257,
      "calib/mean_conf": 0.13737051792828686,
      "calib/mu_c": 0.1493197278911565,
      "calib/mu_w": 0.12048076923076924,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05068709344166836,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1582.0,
      "completions/max_terminated_length": 1582.0,
      "completions/mean_length": 493.59375,
      "completions/mean_terminated_length": 497.4803161621094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 195.0,
      "epoch": 0.20693333333333333,
      "grad_norm": 0.0344914086163044,
      "learning_rate": 1.6666666666666668e-07,
      "loss": -0.0077,
      "num_tokens": 43729691.0,
      "reward": 1.2007018327713013,
      "reward_std": 0.27053576707839966,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.5528074502944946,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 194
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.2498041210572417e-07,
      "aux_brier/mean_group_std": 0.068059353922213,
      "aux_brier/mean_r": 0.9491068867397771,
      "aux_brier/n_active_tok": 294.75,
      "aux_brier/n_groups": 18.625,
      "aux_brier/n_step_records": 73.6875,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5792780215396889,
      "calib/avg_num_step_conf": 9.890625,
      "calib/ece": 0.4136842105263157,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009962106102911844,
      "calib/mean_conf": 0.1471255060728745,
      "calib/mu_c": 0.1515217391304348,
      "calib/mu_w": 0.14155963302752295,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0010526315789473684,
      "calib/std_conf": 0.052602670354152316,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2718.0,
      "completions/max_terminated_length": 2718.0,
      "completions/mean_length": 544.3046875,
      "completions/mean_terminated_length": 557.3680419921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 178.0,
      "epoch": 0.208,
      "grad_norm": 0.145261749625206,
      "learning_rate": 1.3888888888888888e-07,
      "loss": -0.0407,
      "num_tokens": 43975017.0,
      "reward": 1.1599828004837036,
      "reward_std": 0.259069561958313,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5618062615394592,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 195
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.4591270117169106e-06,
      "aux_brier/mean_group_std": 0.061708016616617144,
      "aux_brier/mean_r": 0.953872105484891,
      "aux_brier/n_active_tok": 255.75,
      "aux_brier/n_groups": 13.1875,
      "aux_brier/n_step_records": 63.9375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.534438612213545,
      "calib/avg_num_step_conf": 8.17578125,
      "calib/ece": 0.42758893280632415,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005587632825502509,
      "calib/mean_conf": 0.14948616600790515,
      "calib/mu_c": 0.15184931506849317,
      "calib/mu_w": 0.14626168224299066,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04645679711671892,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1719.0,
      "completions/max_terminated_length": 1719.0,
      "completions/mean_length": 459.8203125,
      "completions/mean_terminated_length": 461.6235656738281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.20906666666666668,
      "grad_norm": 0.018504036590456963,
      "learning_rate": 1.1111111111111112e-07,
      "loss": -0.0032,
      "num_tokens": 44195275.0,
      "reward": 1.2061917781829834,
      "reward_std": 0.2512151896953583,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.5669547319412231,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 196
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -4.589672547850032e-07,
      "aux_brier/mean_group_std": 0.09339156506784498,
      "aux_brier/mean_r": 0.9286565395622205,
      "aux_brier/n_active_tok": 307.25,
      "aux_brier/n_groups": 21.875,
      "aux_brier/n_step_records": 76.8125,
      "calib/answer_extract_rate": 0.9453125,
      "calib/auroc": 0.5016392322928762,
      "calib/avg_num_step_conf": 10.1640625,
      "calib/ece": 0.34995867768595046,
      "calib/final_conf_rate": 0.9453125,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005371900826446313,
      "calib/mean_conf": 0.15260330578512393,
      "calib/mu_c": 0.15528925619834713,
      "calib/mu_w": 0.14991735537190082,
      "calib/nonempty_final_conf_rate": 0.9453125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0012809917355371902,
      "calib/std_conf": 0.06047538661113296,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2845.0,
      "completions/max_terminated_length": 2845.0,
      "completions/mean_length": 572.05078125,
      "completions/mean_terminated_length": 583.4462280273438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 153.0,
      "epoch": 0.21013333333333334,
      "grad_norm": 0.0341520756483078,
      "learning_rate": 8.333333333333334e-08,
      "loss": 0.0792,
      "num_tokens": 44446776.0,
      "reward": 1.0938079357147217,
      "reward_std": 0.2998431921005249,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.5939816236495972,
      "rewards/format_reward_step_strict": 0.9453125,
      "step": 197
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": -5.442882395068072e-07,
      "aux_brier/mean_group_std": 0.0549987736354695,
      "aux_brier/mean_r": 0.9597876312209791,
      "aux_brier/n_active_tok": 291.625,
      "aux_brier/n_groups": 17.75,
      "aux_brier/n_step_records": 72.90625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6205846936044956,
      "calib/avg_num_step_conf": 9.453125,
      "calib/ece": 0.44405622489959834,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.004016064257028112,
      "calib/gap": 0.02602689322986354,
      "calib/mean_conf": 0.15032128514056228,
      "calib/mu_c": 0.1608783783783784,
      "calib/mu_w": 0.13485148514851486,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.07592466931371257,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2248.0,
      "completions/max_terminated_length": 2248.0,
      "completions/mean_length": 511.1484375,
      "completions/mean_terminated_length": 519.261962890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 138.0,
      "epoch": 0.2112,
      "grad_norm": 0.059009380638599396,
      "learning_rate": 5.555555555555556e-08,
      "loss": -0.0015,
      "num_tokens": 44683014.0,
      "reward": 1.1997989416122437,
      "reward_std": 0.24536868929862976,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.5491960644721985,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 198
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 2.0662572798244017e-07,
      "aux_brier/mean_group_std": 0.0600427922814054,
      "aux_brier/mean_r": 0.9515209018543054,
      "aux_brier/n_active_tok": 301.0,
      "aux_brier/n_groups": 16.8125,
      "aux_brier/n_step_records": 75.25,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5038126361655774,
      "calib/avg_num_step_conf": 9.54296875,
      "calib/ece": 0.4215040650406504,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0012745098039215752,
      "calib/mean_conf": 0.1638617886178862,
      "calib/mu_c": 0.16333333333333333,
      "calib/mu_w": 0.1646078431372549,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05597732194746908,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2327.0,
      "completions/max_terminated_length": 2327.0,
      "completions/mean_length": 585.2265625,
      "completions/mean_terminated_length": 592.166015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 107.0,
      "epoch": 0.21226666666666666,
      "grad_norm": 0.04053623229265213,
      "learning_rate": 2.777777777777778e-08,
      "loss": 0.0188,
      "num_tokens": 44937032.0,
      "reward": 1.1813123226165771,
      "reward_std": 0.3064155578613281,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.5533746480941772,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 199
    },
    {
      "aux_brier/lambda": 0.10000000000000005,
      "aux_brier/loss": 7.429067104963716e-07,
      "aux_brier/mean_group_std": 0.05776427918373577,
      "aux_brier/mean_r": 0.9539067025547618,
      "aux_brier/n_active_tok": 261.375,
      "aux_brier/n_groups": 15.53125,
      "aux_brier/n_step_records": 65.34375,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6913060897435896,
      "calib/avg_num_step_conf": 8.19921875,
      "calib/ece": 0.43596774193548393,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.026597222222222217,
      "calib/mean_conf": 0.15169354838709678,
      "calib/mu_c": 0.16284722222222223,
      "calib/mu_w": 0.13625,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0035080645161290323,
      "calib/std_conf": 0.06901638157798934,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2382.0,
      "completions/max_terminated_length": 2382.0,
      "completions/mean_length": 528.5390625,
      "completions/mean_terminated_length": 532.7008056640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 78.0,
      "epoch": 0.21333333333333335,
      "grad_norm": 0.04022820293903351,
      "learning_rate": 0.0,
      "loss": -0.0159,
      "num_tokens": 45180386.0,
      "reward": 1.187511682510376,
      "reward_std": 0.22799977660179138,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.5625468492507935,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 200
    },
    {
      "epoch": 0.21333333333333335,
      "step": 200,
      "total_flos": 0.0,
      "train_loss": 0.035836528669169636,
      "train_runtime": 16174.3808,
      "train_samples_per_second": 3.165,
      "train_steps_per_second": 0.012
    }
  ],
  "logging_steps": 1,
  "max_steps": 200,
  "num_input_tokens_seen": 45180386,
  "num_train_epochs": 1,
  "save_steps": 40,
  "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": 8,
  "trial_name": null,
  "trial_params": null
}