{
  "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": 0.25,
      "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": 0.8019090294837952,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0318,
      "num_tokens": 264685.0,
      "reward": 0.04124843701720238,
      "reward_std": 0.0838509351015091,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.01655624993145466,
      "rewards/format_reward_step": 0.04296875,
      "step": 1
    },
    {
      "aux_brier/lambda": 0.25,
      "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": 0.007053256966173649,
      "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": 0.08984375,
      "step": 2
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.4675386239559884e-08,
      "aux_brier/mean_group_std": 0.010562199023242428,
      "aux_brier/mean_r": 0.4475614425895778,
      "aux_brier/n_active_tok": 22.666666666666668,
      "aux_brier/n_groups": 4.833333333333333,
      "aux_brier/n_step_records": 5.666666666666667,
      "calib/answer_extract_rate": 0.04296875,
      "calib/auroc": 0.625,
      "calib/avg_num_step_conf": 0.26953125,
      "calib/ece": 0.6136363636363635,
      "calib/final_conf_rate": 0.04296875,
      "calib/format_rate": 0.03125,
      "calib/frac_conf_gt_0.9": 0.6363636363636364,
      "calib/gap": 0.10124999999999995,
      "calib/mean_conf": 0.8863636363636364,
      "calib/mu_c": 0.96,
      "calib/mu_w": 0.85875,
      "calib/nonempty_final_conf_rate": 0.04296875,
      "calib/nonempty_reasoning_rate": 0.06640625,
      "calib/nonempty_step_conf_rate": 0.05859375,
      "calib/pce": 0.6136363636363635,
      "calib/std_conf": 0.14169575134919613,
      "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": 0.05859375,
      "completions/max_length": 3062.0,
      "completions/max_terminated_length": 3062.0,
      "completions/mean_length": 726.01171875,
      "completions/mean_terminated_length": 771.19921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0032,
      "grad_norm": 0.26548826694488525,
      "learning_rate": 7.5e-07,
      "loss": 0.0159,
      "num_tokens": 824582.0,
      "reward": 0.03019765578210354,
      "reward_std": 0.05533502995967865,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.011415624991059303,
      "rewards/format_reward_step": 0.03125,
      "step": 3
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.966125440233668e-08,
      "aux_brier/mean_group_std": 0.026093932111437953,
      "aux_brier/mean_r": 0.5140029739995847,
      "aux_brier/n_active_tok": 22.666666666666668,
      "aux_brier/n_groups": 4.833333333333333,
      "aux_brier/n_step_records": 5.666666666666667,
      "calib/answer_extract_rate": 0.06640625,
      "calib/auroc": 0.7727272727272727,
      "calib/avg_num_step_conf": 0.27734375,
      "calib/ece": 0.7430769230769232,
      "calib/final_conf_rate": 0.05078125,
      "calib/format_rate": 0.0390625,
      "calib/frac_conf_gt_0.9": 0.7692307692307693,
      "calib/gap": 0.09818181818181826,
      "calib/mean_conf": 0.8969230769230769,
      "calib/mu_c": 0.98,
      "calib/mu_w": 0.8818181818181817,
      "calib/nonempty_final_conf_rate": 0.05078125,
      "calib/nonempty_reasoning_rate": 0.07421875,
      "calib/nonempty_step_conf_rate": 0.05859375,
      "calib/pce": 0.7430769230769232,
      "calib/std_conf": 0.20521600684999117,
      "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": 0.05859375,
      "completions/max_length": 3048.0,
      "completions/max_terminated_length": 3048.0,
      "completions/mean_length": 709.09375,
      "completions/mean_terminated_length": 753.228271484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 5.0,
      "epoch": 0.004266666666666667,
      "grad_norm": 0.003563561709597707,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0307,
      "num_tokens": 1112278.0,
      "reward": 0.030818065628409386,
      "reward_std": 0.07425656169652939,
      "rewards/accuracy_reward_step": 0.0078125,
      "rewards/final_brier_reward_step": 0.013897265307605267,
      "rewards/format_reward_step": 0.0390625,
      "step": 4
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.3686993069781137e-08,
      "aux_brier/mean_group_std": 0.04379681054330711,
      "aux_brier/mean_r": 0.42881758133179143,
      "aux_brier/n_active_tok": 32.888888888888886,
      "aux_brier/n_groups": 5.888888888888889,
      "aux_brier/n_step_records": 8.222222222222221,
      "calib/answer_extract_rate": 0.046875,
      "calib/auroc": 0.6190476190476191,
      "calib/avg_num_step_conf": 0.29296875,
      "calib/ece": 0.565,
      "calib/final_conf_rate": 0.0390625,
      "calib/format_rate": 0.03515625,
      "calib/frac_conf_gt_0.9": 0.7,
      "calib/gap": 0.1461904761904762,
      "calib/mean_conf": 0.841,
      "calib/mu_c": 0.9433333333333334,
      "calib/mu_w": 0.7971428571428572,
      "calib/nonempty_final_conf_rate": 0.0390625,
      "calib/nonempty_reasoning_rate": 0.0625,
      "calib/nonempty_step_conf_rate": 0.0546875,
      "calib/pce": 0.5529999999999999,
      "calib/std_conf": 0.2564546743578678,
      "calib/step_conf_rate": 0.0546875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.10546875,
      "completions/max_length": 2843.0,
      "completions/max_terminated_length": 2843.0,
      "completions/mean_length": 610.21875,
      "completions/mean_terminated_length": 682.1659545898438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.005333333333333333,
      "grad_norm": 0.00464936438947916,
      "learning_rate": 1.25e-06,
      "loss": 0.0103,
      "num_tokens": 1375182.0,
      "reward": 0.03314834088087082,
      "reward_std": 0.06054586544632912,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.015405858866870403,
      "rewards/format_reward_step": 0.03515625,
      "step": 5
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.232474389212257e-08,
      "aux_brier/mean_group_std": 0.013661028843392754,
      "aux_brier/mean_r": 0.4007950204987466,
      "aux_brier/n_active_tok": 28.533333333333335,
      "aux_brier/n_groups": 6.2,
      "aux_brier/n_step_records": 7.133333333333334,
      "calib/answer_extract_rate": 0.11328125,
      "calib/auroc": 0.0625,
      "calib/avg_num_step_conf": 0.49609375,
      "calib/ece": 0.9123529411764707,
      "calib/final_conf_rate": 0.06640625,
      "calib/format_rate": 0.0546875,
      "calib/frac_conf_gt_0.9": 1.0,
      "calib/gap": -0.022499999999999964,
      "calib/mean_conf": 0.9711764705882354,
      "calib/mu_c": 0.95,
      "calib/mu_w": 0.9724999999999999,
      "calib/nonempty_final_conf_rate": 0.06640625,
      "calib/nonempty_reasoning_rate": 0.125,
      "calib/nonempty_step_conf_rate": 0.08203125,
      "calib/pce": 0.9123529411764707,
      "calib/std_conf": 0.019668533526235873,
      "calib/step_conf_rate": 0.08203125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05859375,
      "completions/max_length": 2834.0,
      "completions/max_terminated_length": 2834.0,
      "completions/mean_length": 598.41015625,
      "completions/mean_terminated_length": 635.6556396484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0064,
      "grad_norm": 0.05702491104602814,
      "learning_rate": 1.5e-06,
      "loss": -0.0057,
      "num_tokens": 1634327.0,
      "reward": 0.03291259706020355,
      "reward_std": 0.07377687096595764,
      "rewards/accuracy_reward_step": 0.00390625,
      "rewards/final_brier_reward_step": 0.0066503905691206455,
      "rewards/format_reward_step": 0.0546875,
      "step": 6
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.07054222048464e-09,
      "aux_brier/mean_group_std": 0.03345712440611508,
      "aux_brier/mean_r": 0.45309954942507086,
      "aux_brier/n_active_tok": 22.133333333333333,
      "aux_brier/n_groups": 4.8,
      "aux_brier/n_step_records": 5.533333333333333,
      "calib/answer_extract_rate": 0.08203125,
      "calib/auroc": 0.33333333333333337,
      "calib/avg_num_step_conf": 0.34765625,
      "calib/ece": 0.7453333333333334,
      "calib/final_conf_rate": 0.05859375,
      "calib/format_rate": 0.04296875,
      "calib/frac_conf_gt_0.9": 0.8666666666666667,
      "calib/gap": -0.006666666666666599,
      "calib/mean_conf": 0.9453333333333334,
      "calib/mu_c": 0.9400000000000001,
      "calib/mu_w": 0.9466666666666667,
      "calib/nonempty_final_conf_rate": 0.05859375,
      "calib/nonempty_reasoning_rate": 0.109375,
      "calib/nonempty_step_conf_rate": 0.078125,
      "calib/pce": 0.7453333333333334,
      "calib/std_conf": 0.05667058809960908,
      "calib/step_conf_rate": 0.078125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.13671875,
      "completions/max_length": 3054.0,
      "completions/max_terminated_length": 3054.0,
      "completions/mean_length": 657.5234375,
      "completions/mean_terminated_length": 761.6561279296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.007466666666666667,
      "grad_norm": 0.9773487448692322,
      "learning_rate": 1.75e-06,
      "loss": 0.0245,
      "num_tokens": 1910077.0,
      "reward": 0.03702617064118385,
      "reward_std": 0.09901833534240723,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.015292187221348286,
      "rewards/format_reward_step": 0.04296875,
      "step": 7
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -8.059045120134556e-08,
      "aux_brier/mean_group_std": 0.04683999810883638,
      "aux_brier/mean_r": 0.5308816023273762,
      "aux_brier/n_active_tok": 26.0,
      "aux_brier/n_groups": 5.083333333333333,
      "aux_brier/n_step_records": 6.5,
      "calib/answer_extract_rate": 0.0703125,
      "calib/auroc": 0.796875,
      "calib/avg_num_step_conf": 0.3046875,
      "calib/ece": 0.4618750000000001,
      "calib/final_conf_rate": 0.0625,
      "calib/format_rate": 0.05078125,
      "calib/frac_conf_gt_0.9": 0.75,
      "calib/gap": 0.03874999999999995,
      "calib/mean_conf": 0.836875,
      "calib/mu_c": 0.85625,
      "calib/mu_w": 0.8175,
      "calib/nonempty_final_conf_rate": 0.0625,
      "calib/nonempty_reasoning_rate": 0.08203125,
      "calib/nonempty_step_conf_rate": 0.06640625,
      "calib/pce": 0.3993750000000001,
      "calib/std_conf": 0.2816185085803133,
      "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": 0.08203125,
      "completions/max_length": 2851.0,
      "completions/max_terminated_length": 2851.0,
      "completions/mean_length": 576.5703125,
      "completions/mean_terminated_length": 628.0935668945312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.008533333333333334,
      "grad_norm": 0.23815672099590302,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0261,
      "num_tokens": 2164191.0,
      "reward": 0.06468446552753448,
      "reward_std": 0.1265413910150528,
      "rewards/accuracy_reward_step": 0.03125,
      "rewards/final_brier_reward_step": 0.03217539191246033,
      "rewards/format_reward_step": 0.05078125,
      "step": 8
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 9.461055149194584e-10,
      "aux_brier/mean_group_std": 0.015730202045537118,
      "aux_brier/mean_r": 0.43023866184339343,
      "aux_brier/n_active_tok": 28.444444444444443,
      "aux_brier/n_groups": 6.777777777777778,
      "aux_brier/n_step_records": 7.111111111111111,
      "calib/answer_extract_rate": 0.0625,
      "calib/auroc": 0.9444444444444444,
      "calib/avg_num_step_conf": 0.2578125,
      "calib/ece": 0.7490000000000001,
      "calib/final_conf_rate": 0.0390625,
      "calib/format_rate": 0.02734375,
      "calib/frac_conf_gt_0.9": 0.8,
      "calib/gap": 0.14555555555555555,
      "calib/mean_conf": 0.849,
      "calib/mu_c": 0.98,
      "calib/mu_w": 0.8344444444444444,
      "calib/nonempty_final_conf_rate": 0.0390625,
      "calib/nonempty_reasoning_rate": 0.06640625,
      "calib/nonempty_step_conf_rate": 0.04296875,
      "calib/pce": 0.7490000000000001,
      "calib/std_conf": 0.28727861041156544,
      "calib/step_conf_rate": 0.04296875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.109375,
      "completions/max_length": 2978.0,
      "completions/max_terminated_length": 2978.0,
      "completions/mean_length": 613.953125,
      "completions/mean_terminated_length": 689.3508911132812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0096,
      "grad_norm": 0.005154241342097521,
      "learning_rate": 2.25e-06,
      "loss": -0.0141,
      "num_tokens": 2428899.0,
      "reward": 0.018406933173537254,
      "reward_std": 0.04806169122457504,
      "rewards/accuracy_reward_step": 0.00390625,
      "rewards/final_brier_reward_step": 0.0033152345567941666,
      "rewards/format_reward_step": 0.02734375,
      "step": 9
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.5254254793226636e-08,
      "aux_brier/mean_group_std": 0.04791594090406018,
      "aux_brier/mean_r": 0.4576816674277268,
      "aux_brier/n_active_tok": 24.75,
      "aux_brier/n_groups": 4.875,
      "aux_brier/n_step_records": 6.1875,
      "calib/answer_extract_rate": 0.11328125,
      "calib/auroc": 0.6086956521739131,
      "calib/avg_num_step_conf": 0.421875,
      "calib/ece": 0.8261875000000001,
      "calib/final_conf_rate": 0.09375,
      "calib/format_rate": 0.0546875,
      "calib/frac_conf_gt_0.9": 0.6666666666666666,
      "calib/gap": 0.09615217391304343,
      "calib/mean_conf": 0.8678541666666666,
      "calib/mu_c": 0.96,
      "calib/mu_w": 0.8638478260869565,
      "calib/nonempty_final_conf_rate": 0.09375,
      "calib/nonempty_reasoning_rate": 0.1328125,
      "calib/nonempty_step_conf_rate": 0.09375,
      "calib/pce": 0.8261875000000001,
      "calib/std_conf": 0.2272346199033917,
      "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": 0.05859375,
      "completions/max_length": 3017.0,
      "completions/max_terminated_length": 3017.0,
      "completions/mean_length": 668.12109375,
      "completions/mean_terminated_length": 709.7054443359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.010666666666666666,
      "grad_norm": 1.2165741920471191,
      "learning_rate": 2.5e-06,
      "loss": 0.0376,
      "num_tokens": 2706738.0,
      "reward": 0.03873037174344063,
      "reward_std": 0.08526559174060822,
      "rewards/accuracy_reward_step": 0.0078125,
      "rewards/final_brier_reward_step": 0.014296485111117363,
      "rewards/format_reward_step": 0.0546875,
      "step": 10
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.191568579338384e-06,
      "aux_brier/mean_group_std": 0.08926198206335412,
      "aux_brier/mean_r": 0.382425037350804,
      "aux_brier/n_active_tok": 30.105263157894736,
      "aux_brier/n_groups": 5.315789473684211,
      "aux_brier/n_step_records": 7.526315789473684,
      "calib/answer_extract_rate": 0.15625,
      "calib/auroc": 0.5607142857142857,
      "calib/avg_num_step_conf": 0.5703125,
      "calib/ece": 0.7930303030303029,
      "calib/final_conf_rate": 0.12890625,
      "calib/format_rate": 0.1015625,
      "calib/frac_conf_gt_0.9": 0.8484848484848485,
      "calib/gap": 0.011142857142857121,
      "calib/mean_conf": 0.9445454545454545,
      "calib/mu_c": 0.954,
      "calib/mu_w": 0.9428571428571428,
      "calib/nonempty_final_conf_rate": 0.12890625,
      "calib/nonempty_reasoning_rate": 0.171875,
      "calib/nonempty_step_conf_rate": 0.12109375,
      "calib/pce": 0.7930303030303029,
      "calib/std_conf": 0.04710342643979464,
      "calib/step_conf_rate": 0.12109375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2967.0,
      "completions/max_terminated_length": 2967.0,
      "completions/mean_length": 628.97265625,
      "completions/mean_terminated_length": 694.038818359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.011733333333333333,
      "grad_norm": 0.3624080419540405,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.0328,
      "num_tokens": 2972235.0,
      "reward": 0.07586494088172913,
      "reward_std": 0.13519909977912903,
      "rewards/accuracy_reward_step": 0.01953125,
      "rewards/final_brier_reward_step": 0.022209765389561653,
      "rewards/format_reward_step": 0.1015625,
      "step": 11
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.0138040675961317e-08,
      "aux_brier/mean_group_std": 0.08816123498156225,
      "aux_brier/mean_r": 0.5617093218650877,
      "aux_brier/n_active_tok": 36.0,
      "aux_brier/n_groups": 5.368421052631579,
      "aux_brier/n_step_records": 9.0,
      "calib/answer_extract_rate": 0.16796875,
      "calib/auroc": 0.20238095238095238,
      "calib/avg_num_step_conf": 0.703125,
      "calib/ece": 0.7838235294117647,
      "calib/final_conf_rate": 0.1328125,
      "calib/format_rate": 0.10546875,
      "calib/frac_conf_gt_0.9": 0.8529411764705882,
      "calib/gap": -0.07119047619047625,
      "calib/mean_conf": 0.9302941176470588,
      "calib/mu_c": 0.8716666666666666,
      "calib/mu_w": 0.9428571428571428,
      "calib/nonempty_final_conf_rate": 0.1328125,
      "calib/nonempty_reasoning_rate": 0.19921875,
      "calib/nonempty_step_conf_rate": 0.1484375,
      "calib/pce": 0.7688235294117648,
      "calib/std_conf": 0.07946457938706168,
      "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": 0.0703125,
      "completions/max_length": 3066.0,
      "completions/max_terminated_length": 3066.0,
      "completions/mean_length": 601.875,
      "completions/mean_terminated_length": 647.39501953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0128,
      "grad_norm": 1.0927518606185913,
      "learning_rate": 3e-06,
      "loss": 0.0794,
      "num_tokens": 3230491.0,
      "reward": 0.08419764786958694,
      "reward_std": 0.1351093053817749,
      "rewards/accuracy_reward_step": 0.0234375,
      "rewards/final_brier_reward_step": 0.032103125005960464,
      "rewards/format_reward_step": 0.10546875,
      "step": 12
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 6.745897776186085e-08,
      "aux_brier/mean_group_std": 0.030241205883298414,
      "aux_brier/mean_r": 0.3736955558950782,
      "aux_brier/n_active_tok": 24.444444444444443,
      "aux_brier/n_groups": 4.722222222222222,
      "aux_brier/n_step_records": 6.111111111111111,
      "calib/answer_extract_rate": 0.1171875,
      "calib/auroc": 0.4789473684210526,
      "calib/avg_num_step_conf": 0.45703125,
      "calib/ece": 0.7216666666666666,
      "calib/final_conf_rate": 0.09375,
      "calib/format_rate": 0.07421875,
      "calib/frac_conf_gt_0.9": 0.875,
      "calib/gap": 0.030315789473683963,
      "calib/mean_conf": 0.93,
      "calib/mu_c": 0.954,
      "calib/mu_w": 0.923684210526316,
      "calib/nonempty_final_conf_rate": 0.09375,
      "calib/nonempty_reasoning_rate": 0.1328125,
      "calib/nonempty_step_conf_rate": 0.09765625,
      "calib/pce": 0.7216666666666666,
      "calib/std_conf": 0.08883505314157619,
      "calib/step_conf_rate": 0.09765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2781.0,
      "completions/max_terminated_length": 2781.0,
      "completions/mean_length": 578.16796875,
      "completions/mean_terminated_length": 627.165283203125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.013866666666666666,
      "grad_norm": 0.3745361864566803,
      "learning_rate": 3.2500000000000002e-06,
      "loss": -0.0048,
      "num_tokens": 3483094.0,
      "reward": 0.062227047979831696,
      "reward_std": 0.13563629984855652,
      "rewards/accuracy_reward_step": 0.01953125,
      "rewards/final_brier_reward_step": 0.022345703095197678,
      "rewards/format_reward_step": 0.07421875,
      "step": 13
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -7.824561436553523e-07,
      "aux_brier/mean_group_std": 0.0509677167700097,
      "aux_brier/mean_r": 0.3953121614686363,
      "aux_brier/n_active_tok": 32.285714285714285,
      "aux_brier/n_groups": 5.571428571428571,
      "aux_brier/n_step_records": 8.071428571428571,
      "calib/answer_extract_rate": 0.1875,
      "calib/auroc": 0.6645161290322581,
      "calib/avg_num_step_conf": 0.9140625,
      "calib/ece": 0.782222222222222,
      "calib/final_conf_rate": 0.140625,
      "calib/format_rate": 0.12109375,
      "calib/frac_conf_gt_0.9": 0.7777777777777778,
      "calib/gap": 0.05909677419354864,
      "calib/mean_conf": 0.9211111111111112,
      "calib/mu_c": 0.9720000000000001,
      "calib/mu_w": 0.9129032258064514,
      "calib/nonempty_final_conf_rate": 0.140625,
      "calib/nonempty_reasoning_rate": 0.2265625,
      "calib/nonempty_step_conf_rate": 0.1796875,
      "calib/pce": 0.782222222222222,
      "calib/std_conf": 0.11413388283010675,
      "calib/step_conf_rate": 0.1796875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2817.0,
      "completions/max_terminated_length": 2817.0,
      "completions/mean_length": 643.81640625,
      "completions/mean_terminated_length": 686.737548828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.014933333333333333,
      "grad_norm": 0.3816582262516022,
      "learning_rate": 3.5e-06,
      "loss": 0.022,
      "num_tokens": 3753311.0,
      "reward": 0.0880625993013382,
      "reward_std": 0.16737763583660126,
      "rewards/accuracy_reward_step": 0.01953125,
      "rewards/final_brier_reward_step": 0.031937889754772186,
      "rewards/format_reward_step": 0.12109375,
      "step": 14
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.788997709183158e-08,
      "aux_brier/mean_group_std": 0.06431303721101182,
      "aux_brier/mean_r": 0.494478371672257,
      "aux_brier/n_active_tok": 34.22222222222222,
      "aux_brier/n_groups": 6.0,
      "aux_brier/n_step_records": 8.555555555555555,
      "calib/answer_extract_rate": 0.1640625,
      "calib/auroc": 0.41538461538461546,
      "calib/avg_num_step_conf": 0.90234375,
      "calib/ece": 0.7741935483870968,
      "calib/final_conf_rate": 0.12109375,
      "calib/format_rate": 0.10546875,
      "calib/frac_conf_gt_0.9": 0.6451612903225806,
      "calib/gap": -0.06784615384615378,
      "calib/mean_conf": 0.9129032258064514,
      "calib/mu_c": 0.8560000000000001,
      "calib/mu_w": 0.9238461538461539,
      "calib/nonempty_final_conf_rate": 0.12109375,
      "calib/nonempty_reasoning_rate": 0.22265625,
      "calib/nonempty_step_conf_rate": 0.17578125,
      "calib/pce": 0.7629032258064516,
      "calib/std_conf": 0.07788302015397096,
      "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": 0.0625,
      "completions/max_length": 2918.0,
      "completions/max_terminated_length": 2918.0,
      "completions/mean_length": 580.35546875,
      "completions/mean_terminated_length": 619.0458374023438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.016,
      "grad_norm": 0.12358175963163376,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.0893,
      "num_tokens": 4009762.0,
      "reward": 0.08004707098007202,
      "reward_std": 0.1672627329826355,
      "rewards/accuracy_reward_step": 0.01953125,
      "rewards/final_brier_reward_step": 0.031125782057642937,
      "rewards/format_reward_step": 0.10546875,
      "step": 15
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 6.605632350200417e-08,
      "aux_brier/mean_group_std": 0.10973796878610398,
      "aux_brier/mean_r": 0.5119668618019643,
      "aux_brier/n_active_tok": 38.0,
      "aux_brier/n_groups": 5.7727272727272725,
      "aux_brier/n_step_records": 9.5,
      "calib/answer_extract_rate": 0.2109375,
      "calib/auroc": 0.5340501792114696,
      "calib/avg_num_step_conf": 0.81640625,
      "calib/ece": 0.6762499999999998,
      "calib/final_conf_rate": 0.15625,
      "calib/format_rate": 0.11328125,
      "calib/frac_conf_gt_0.9": 0.825,
      "calib/gap": 0.05516129032258066,
      "calib/mean_conf": 0.8772500000000001,
      "calib/mu_c": 0.9199999999999999,
      "calib/mu_w": 0.8648387096774193,
      "calib/nonempty_final_conf_rate": 0.15625,
      "calib/nonempty_reasoning_rate": 0.234375,
      "calib/nonempty_step_conf_rate": 0.15625,
      "calib/pce": 0.6642499999999998,
      "calib/std_conf": 0.24191927889277448,
      "calib/step_conf_rate": 0.15625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2984.0,
      "completions/max_terminated_length": 2984.0,
      "completions/mean_length": 693.84375,
      "completions/mean_terminated_length": 768.93505859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.017066666666666667,
      "grad_norm": 0.10435499250888824,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0697,
      "num_tokens": 4296234.0,
      "reward": 0.10126493871212006,
      "reward_std": 0.21034234762191772,
      "rewards/accuracy_reward_step": 0.03515625,
      "rewards/final_brier_reward_step": 0.03787226602435112,
      "rewards/format_reward_step": 0.11328125,
      "step": 16
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.1079165019320953e-08,
      "aux_brier/mean_group_std": 0.0960157782281253,
      "aux_brier/mean_r": 0.49889396337404934,
      "aux_brier/n_active_tok": 41.54838709677419,
      "aux_brier/n_groups": 6.709677419354839,
      "aux_brier/n_step_records": 10.387096774193548,
      "calib/answer_extract_rate": 0.25390625,
      "calib/auroc": 0.43181818181818177,
      "calib/avg_num_step_conf": 1.33984375,
      "calib/ece": 0.7013529411764705,
      "calib/final_conf_rate": 0.19921875,
      "calib/format_rate": 0.16015625,
      "calib/frac_conf_gt_0.9": 0.8235294117647058,
      "calib/gap": 0.003252272727272687,
      "calib/mean_conf": 0.9001764705882351,
      "calib/mu_c": 0.9027272727272727,
      "calib/mu_w": 0.899475,
      "calib/nonempty_final_conf_rate": 0.19921875,
      "calib/nonempty_reasoning_rate": 0.3203125,
      "calib/nonempty_step_conf_rate": 0.2421875,
      "calib/pce": 0.6929215686274509,
      "calib/std_conf": 0.1827727299651559,
      "calib/step_conf_rate": 0.2421875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2843.0,
      "completions/max_terminated_length": 2843.0,
      "completions/mean_length": 585.78515625,
      "completions/mean_terminated_length": 635.427978515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.018133333333333335,
      "grad_norm": 0.7509145140647888,
      "learning_rate": 4.25e-06,
      "loss": 0.0676,
      "num_tokens": 4549723.0,
      "reward": 0.13649648427963257,
      "reward_std": 0.23179839551448822,
      "rewards/accuracy_reward_step": 0.04296875,
      "rewards/final_brier_reward_step": 0.053798433393239975,
      "rewards/format_reward_step": 0.16015625,
      "step": 17
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 8.246359257748002e-08,
      "aux_brier/mean_group_std": 0.06599999581363429,
      "aux_brier/mean_r": 0.4889344608586582,
      "aux_brier/n_active_tok": 37.857142857142854,
      "aux_brier/n_groups": 7.285714285714286,
      "aux_brier/n_step_records": 9.464285714285714,
      "calib/answer_extract_rate": 0.18359375,
      "calib/auroc": 0.6096774193548388,
      "calib/avg_num_step_conf": 1.08203125,
      "calib/ece": 0.5914634146341464,
      "calib/final_conf_rate": 0.16015625,
      "calib/format_rate": 0.140625,
      "calib/frac_conf_gt_0.9": 0.7317073170731707,
      "calib/gap": 0.13574193548387092,
      "calib/mean_conf": 0.8353658536585364,
      "calib/mu_c": 0.938,
      "calib/mu_w": 0.802258064516129,
      "calib/nonempty_final_conf_rate": 0.16015625,
      "calib/nonempty_reasoning_rate": 0.23828125,
      "calib/nonempty_step_conf_rate": 0.1953125,
      "calib/pce": 0.5914634146341464,
      "calib/std_conf": 0.2663859071290262,
      "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": 0.07421875,
      "completions/max_length": 2874.0,
      "completions/max_terminated_length": 2874.0,
      "completions/mean_length": 635.3828125,
      "completions/mean_terminated_length": 686.3206176757812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0192,
      "grad_norm": 0.23033437132835388,
      "learning_rate": 4.5e-06,
      "loss": 0.0745,
      "num_tokens": 4823101.0,
      "reward": 0.12414237856864929,
      "reward_std": 0.1830832064151764,
      "rewards/accuracy_reward_step": 0.0390625,
      "rewards/final_brier_reward_step": 0.05906952917575836,
      "rewards/format_reward_step": 0.140625,
      "step": 18
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.1799148291497127e-07,
      "aux_brier/mean_group_std": 0.12539111681370793,
      "aux_brier/mean_r": 0.4819700571199433,
      "aux_brier/n_active_tok": 49.29032258064516,
      "aux_brier/n_groups": 6.096774193548387,
      "aux_brier/n_step_records": 12.32258064516129,
      "calib/answer_extract_rate": 0.30078125,
      "calib/auroc": 0.48460591133004927,
      "calib/avg_num_step_conf": 1.55078125,
      "calib/ece": 0.7018055555555555,
      "calib/final_conf_rate": 0.28125,
      "calib/format_rate": 0.24609375,
      "calib/frac_conf_gt_0.9": 0.75,
      "calib/gap": 0.041896551724137865,
      "calib/mean_conf": 0.8812500000000001,
      "calib/mu_c": 0.9149999999999999,
      "calib/mu_w": 0.8731034482758621,
      "calib/nonempty_final_conf_rate": 0.28125,
      "calib/nonempty_reasoning_rate": 0.390625,
      "calib/nonempty_step_conf_rate": 0.34765625,
      "calib/pce": 0.6943055555555555,
      "calib/std_conf": 0.21670472422169296,
      "calib/step_conf_rate": 0.34765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.046875,
      "completions/max_length": 3002.0,
      "completions/max_terminated_length": 3002.0,
      "completions/mean_length": 526.578125,
      "completions/mean_terminated_length": 552.4754028320312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.020266666666666665,
      "grad_norm": 0.6649148464202881,
      "learning_rate": 4.75e-06,
      "loss": 0.0971,
      "num_tokens": 5062665.0,
      "reward": 0.1994456946849823,
      "reward_std": 0.3087695837020874,
      "rewards/accuracy_reward_step": 0.0546875,
      "rewards/final_brier_reward_step": 0.08684530854225159,
      "rewards/format_reward_step": 0.24609375,
      "step": 19
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.661198010376012e-09,
      "aux_brier/mean_group_std": 0.18190807637293757,
      "aux_brier/mean_r": 0.49438430379502507,
      "aux_brier/n_active_tok": 79.125,
      "aux_brier/n_groups": 7.40625,
      "aux_brier/n_step_records": 19.78125,
      "calib/answer_extract_rate": 0.51953125,
      "calib/auroc": 0.5978576615831518,
      "calib/avg_num_step_conf": 2.5234375,
      "calib/ece": 0.6113913043478262,
      "calib/final_conf_rate": 0.44921875,
      "calib/format_rate": 0.40625,
      "calib/frac_conf_gt_0.9": 0.7565217391304347,
      "calib/gap": 0.05180101670297743,
      "calib/mean_conf": 0.9070434782608695,
      "calib/mu_c": 0.9435294117647058,
      "calib/mu_w": 0.8917283950617284,
      "calib/nonempty_final_conf_rate": 0.44921875,
      "calib/nonempty_reasoning_rate": 0.58984375,
      "calib/nonempty_step_conf_rate": 0.50390625,
      "calib/pce": 0.6113913043478262,
      "calib/std_conf": 0.14471545642063302,
      "calib/step_conf_rate": 0.50390625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 2971.0,
      "completions/max_terminated_length": 2971.0,
      "completions/mean_length": 437.01953125,
      "completions/mean_terminated_length": 449.3052062988281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 7.0,
      "epoch": 0.021333333333333333,
      "grad_norm": 0.36984461545944214,
      "learning_rate": 5e-06,
      "loss": 0.0571,
      "num_tokens": 5279414.0,
      "reward": 0.3783619999885559,
      "reward_std": 0.48967522382736206,
      "rewards/accuracy_reward_step": 0.1328125,
      "rewards/final_brier_reward_step": 0.16969804465770721,
      "rewards/format_reward_step": 0.40625,
      "step": 20
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.3740270443939337e-09,
      "aux_brier/mean_group_std": 0.21181267982006952,
      "aux_brier/mean_r": 0.5149232417855898,
      "aux_brier/n_active_tok": 101.25,
      "aux_brier/n_groups": 8.78125,
      "aux_brier/n_step_records": 25.3125,
      "calib/answer_extract_rate": 0.640625,
      "calib/auroc": 0.5442108294930876,
      "calib/avg_num_step_conf": 3.29296875,
      "calib/ece": 0.7090269736842105,
      "calib/final_conf_rate": 0.59375,
      "calib/format_rate": 0.5234375,
      "calib/frac_conf_gt_0.9": 0.7236842105263158,
      "calib/gap": 0.03186025345622123,
      "calib/mean_conf": 0.8822230263157893,
      "calib/mu_c": 0.9082142857142859,
      "calib/mu_w": 0.8763540322580646,
      "calib/nonempty_final_conf_rate": 0.59375,
      "calib/nonempty_reasoning_rate": 0.734375,
      "calib/nonempty_step_conf_rate": 0.6328125,
      "calib/pce": 0.7035197368421053,
      "calib/std_conf": 0.19938412492578328,
      "calib/step_conf_rate": 0.6328125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2691.0,
      "completions/max_terminated_length": 2691.0,
      "completions/mean_length": 443.609375,
      "completions/mean_terminated_length": 447.10235595703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 5.0,
      "epoch": 0.0224,
      "grad_norm": 0.6879671216011047,
      "learning_rate": 4.9722222222222224e-06,
      "loss": 0.1353,
      "num_tokens": 5495938.0,
      "reward": 0.41851744055747986,
      "reward_std": 0.4342387318611145,
      "rewards/accuracy_reward_step": 0.11328125,
      "rewards/final_brier_reward_step": 0.17406976222991943,
      "rewards/format_reward_step": 0.5234375,
      "step": 21
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 9.913477306416896e-08,
      "aux_brier/mean_group_std": 0.19722453778596868,
      "aux_brier/mean_r": 0.5062000081744715,
      "aux_brier/n_active_tok": 104.0,
      "aux_brier/n_groups": 7.78125,
      "aux_brier/n_step_records": 26.0,
      "calib/answer_extract_rate": 0.6953125,
      "calib/auroc": 0.49372206852521816,
      "calib/avg_num_step_conf": 3.26953125,
      "calib/ece": 0.6656463414634147,
      "calib/final_conf_rate": 0.640625,
      "calib/format_rate": 0.5625,
      "calib/frac_conf_gt_0.9": 0.725609756097561,
      "calib/gap": 0.007945945945945776,
      "calib/mean_conf": 0.8897926829268293,
      "calib/mu_c": 0.8959459459459459,
      "calib/mu_w": 0.8880000000000001,
      "calib/nonempty_final_conf_rate": 0.640625,
      "calib/nonempty_reasoning_rate": 0.765625,
      "calib/nonempty_step_conf_rate": 0.6640625,
      "calib/pce": 0.6649146341463416,
      "calib/std_conf": 0.16682525089657244,
      "calib/step_conf_rate": 0.6640625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2846.0,
      "completions/max_terminated_length": 2846.0,
      "completions/mean_length": 386.79296875,
      "completions/mean_terminated_length": 392.93255615234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.023466666666666667,
      "grad_norm": 0.3435189723968506,
      "learning_rate": 4.944444444444445e-06,
      "loss": 0.2041,
      "num_tokens": 5696773.0,
      "reward": 0.48116904497146606,
      "reward_std": 0.45967790484428406,
      "rewards/accuracy_reward_step": 0.1484375,
      "rewards/final_brier_reward_step": 0.20592617988586426,
      "rewards/format_reward_step": 0.5625,
      "step": 22
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.6693964287328722e-08,
      "aux_brier/mean_group_std": 0.19466317894734803,
      "aux_brier/mean_r": 0.46551549790218716,
      "aux_brier/n_active_tok": 111.125,
      "aux_brier/n_groups": 7.8125,
      "aux_brier/n_step_records": 27.78125,
      "calib/answer_extract_rate": 0.74609375,
      "calib/auroc": 0.488888888888889,
      "calib/avg_num_step_conf": 3.51171875,
      "calib/ece": 0.6667318435754189,
      "calib/final_conf_rate": 0.69921875,
      "calib/format_rate": 0.62890625,
      "calib/frac_conf_gt_0.9": 0.7486033519553073,
      "calib/gap": 0.01290235690235686,
      "calib/mean_conf": 0.9125418994413406,
      "calib/mu_c": 0.9222727272727272,
      "calib/mu_w": 0.9093703703703704,
      "calib/nonempty_final_conf_rate": 0.69921875,
      "calib/nonempty_reasoning_rate": 0.80078125,
      "calib/nonempty_step_conf_rate": 0.69921875,
      "calib/pce": 0.6667318435754189,
      "calib/std_conf": 0.12324224307871748,
      "calib/step_conf_rate": 0.69921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2747.0,
      "completions/max_terminated_length": 2747.0,
      "completions/mean_length": 365.08203125,
      "completions/mean_terminated_length": 370.87701416015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.024533333333333334,
      "grad_norm": 0.3371759355068207,
      "learning_rate": 4.9166666666666665e-06,
      "loss": 0.1058,
      "num_tokens": 5894170.0,
      "reward": 0.5453325510025024,
      "reward_std": 0.5091565251350403,
      "rewards/accuracy_reward_step": 0.171875,
      "rewards/final_brier_reward_step": 0.23601758480072021,
      "rewards/format_reward_step": 0.62890625,
      "step": 23
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -6.6316448651959625e-09,
      "aux_brier/mean_group_std": 0.20653811194818572,
      "aux_brier/mean_r": 0.4973872652994401,
      "aux_brier/n_active_tok": 134.5,
      "aux_brier/n_groups": 9.03125,
      "aux_brier/n_step_records": 33.625,
      "calib/answer_extract_rate": 0.84765625,
      "calib/auroc": 0.5679466230936819,
      "calib/avg_num_step_conf": 4.30078125,
      "calib/ece": 0.6650044776119403,
      "calib/final_conf_rate": 0.78515625,
      "calib/format_rate": 0.734375,
      "calib/frac_conf_gt_0.9": 0.7412935323383084,
      "calib/gap": 0.04283562091503279,
      "calib/mean_conf": 0.9038104477611941,
      "calib/mu_c": 0.9364166666666667,
      "calib/mu_w": 0.8935810457516339,
      "calib/nonempty_final_conf_rate": 0.78515625,
      "calib/nonempty_reasoning_rate": 0.91015625,
      "calib/nonempty_step_conf_rate": 0.8203125,
      "calib/pce": 0.6650044776119403,
      "calib/std_conf": 0.16970596737734678,
      "calib/step_conf_rate": 0.8203125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1742.0,
      "completions/max_terminated_length": 1742.0,
      "completions/mean_length": 334.46875,
      "completions/mean_terminated_length": 341.1314697265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 11.0,
      "epoch": 0.0256,
      "grad_norm": 1.9992388486862183,
      "learning_rate": 4.888888888888889e-06,
      "loss": 0.0465,
      "num_tokens": 6084306.0,
      "reward": 0.6237559914588928,
      "reward_std": 0.47158321738243103,
      "rewards/accuracy_reward_step": 0.1875,
      "rewards/final_brier_reward_step": 0.2762739658355713,
      "rewards/format_reward_step": 0.734375,
      "step": 24
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.9139240043514194e-08,
      "aux_brier/mean_group_std": 0.20935374056120784,
      "aux_brier/mean_r": 0.505291674313617,
      "aux_brier/n_active_tok": 132.875,
      "aux_brier/n_groups": 8.1875,
      "aux_brier/n_step_records": 33.21875,
      "calib/answer_extract_rate": 0.8984375,
      "calib/auroc": 0.5071524577572964,
      "calib/avg_num_step_conf": 4.203125,
      "calib/ece": 0.6408339130434781,
      "calib/final_conf_rate": 0.8984375,
      "calib/format_rate": 0.8203125,
      "calib/frac_conf_gt_0.9": 0.7086956521739131,
      "calib/gap": -0.012659600614439381,
      "calib/mean_conf": 0.8769921739130435,
      "calib/mu_c": 0.8677451612903226,
      "calib/mu_w": 0.880404761904762,
      "calib/nonempty_final_conf_rate": 0.8984375,
      "calib/nonempty_reasoning_rate": 0.953125,
      "calib/nonempty_step_conf_rate": 0.88671875,
      "calib/pce": 0.6241304347826085,
      "calib/std_conf": 0.20299148864051295,
      "calib/step_conf_rate": 0.88671875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2546.0,
      "completions/max_terminated_length": 2546.0,
      "completions/mean_length": 291.7265625,
      "completions/mean_terminated_length": 295.185791015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 26.0,
      "epoch": 0.02666666666666667,
      "grad_norm": 0.3488129675388336,
      "learning_rate": 4.861111111111111e-06,
      "loss": 0.0202,
      "num_tokens": 6262212.0,
      "reward": 0.7352092862129211,
      "reward_std": 0.4682455062866211,
      "rewards/accuracy_reward_step": 0.24609375,
      "rewards/final_brier_reward_step": 0.31583714485168457,
      "rewards/format_reward_step": 0.8203125,
      "step": 25
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 3.5249401730830243e-09,
      "aux_brier/mean_group_std": 0.2004840616017761,
      "aux_brier/mean_r": 0.5115577585789893,
      "aux_brier/n_active_tok": 152.25,
      "aux_brier/n_groups": 8.8125,
      "aux_brier/n_step_records": 38.0625,
      "calib/answer_extract_rate": 0.8671875,
      "calib/auroc": 0.483560545308741,
      "calib/avg_num_step_conf": 4.81640625,
      "calib/ece": 0.700069585253456,
      "calib/final_conf_rate": 0.84765625,
      "calib/format_rate": 0.7890625,
      "calib/frac_conf_gt_0.9": 0.7603686635944701,
      "calib/gap": 0.019425380914194323,
      "calib/mean_conf": 0.8869820276497697,
      "calib/mu_c": 0.902558139534884,
      "calib/mu_w": 0.8831327586206896,
      "calib/nonempty_final_conf_rate": 0.84765625,
      "calib/nonempty_reasoning_rate": 0.9609375,
      "calib/nonempty_step_conf_rate": 0.9140625,
      "calib/pce": 0.6944474654377878,
      "calib/std_conf": 0.20601633575678802,
      "calib/step_conf_rate": 0.9140625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1818.0,
      "completions/max_terminated_length": 1818.0,
      "completions/mean_length": 303.15234375,
      "completions/mean_terminated_length": 304.3411865234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.027733333333333332,
      "grad_norm": 1.5694950819015503,
      "learning_rate": 4.833333333333333e-06,
      "loss": 0.0266,
      "num_tokens": 6445059.0,
      "reward": 0.6304481029510498,
      "reward_std": 0.4700429439544678,
      "rewards/accuracy_reward_step": 0.171875,
      "rewards/final_brier_reward_step": 0.256167471408844,
      "rewards/format_reward_step": 0.7890625,
      "step": 26
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 6.1927870401157215e-09,
      "aux_brier/mean_group_std": 0.19561626236250582,
      "aux_brier/mean_r": 0.5135547247653034,
      "aux_brier/n_active_tok": 148.625,
      "aux_brier/n_groups": 8.625,
      "aux_brier/n_step_records": 37.15625,
      "calib/answer_extract_rate": 0.90625,
      "calib/auroc": 0.5291819291819292,
      "calib/avg_num_step_conf": 4.734375,
      "calib/ece": 0.677325697503671,
      "calib/final_conf_rate": 0.88671875,
      "calib/format_rate": 0.83203125,
      "calib/frac_conf_gt_0.9": 0.6475770925110133,
      "calib/gap": 0.023012454212454014,
      "calib/mean_conf": 0.8675494860499267,
      "calib/mu_c": 0.8859999999999999,
      "calib/mu_w": 0.8629875457875459,
      "calib/nonempty_final_conf_rate": 0.88671875,
      "calib/nonempty_reasoning_rate": 0.96484375,
      "calib/nonempty_step_conf_rate": 0.9140625,
      "calib/pce": 0.6733186490455212,
      "calib/std_conf": 0.2010847250794651,
      "calib/step_conf_rate": 0.9140625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2712.0,
      "completions/max_terminated_length": 2712.0,
      "completions/mean_length": 308.921875,
      "completions/mean_terminated_length": 312.5849914550781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 34.0,
      "epoch": 0.0288,
      "grad_norm": 4.943506717681885,
      "learning_rate": 4.805555555555556e-06,
      "loss": 0.0659,
      "num_tokens": 6629359.0,
      "reward": 0.6669065952301025,
      "reward_std": 0.42902442812919617,
      "rewards/accuracy_reward_step": 0.17578125,
      "rewards/final_brier_reward_step": 0.3004390001296997,
      "rewards/format_reward_step": 0.83203125,
      "step": 27
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.2140759217404984e-08,
      "aux_brier/mean_group_std": 0.21330967080190893,
      "aux_brier/mean_r": 0.5095590760513937,
      "aux_brier/n_active_tok": 161.25,
      "aux_brier/n_groups": 9.59375,
      "aux_brier/n_step_records": 40.3125,
      "calib/answer_extract_rate": 0.89453125,
      "calib/auroc": 0.4992000752870318,
      "calib/avg_num_step_conf": 5.07421875,
      "calib/ece": 0.6084577092511015,
      "calib/final_conf_rate": 0.88671875,
      "calib/format_rate": 0.84375,
      "calib/frac_conf_gt_0.9": 0.7180616740088106,
      "calib/gap": 0.012993920572181183,
      "calib/mean_conf": 0.8877537444933922,
      "calib/mu_c": 0.8969696969696969,
      "calib/mu_w": 0.8839757763975157,
      "calib/nonempty_final_conf_rate": 0.88671875,
      "calib/nonempty_reasoning_rate": 0.96484375,
      "calib/nonempty_step_conf_rate": 0.9453125,
      "calib/pce": 0.6027312775330398,
      "calib/std_conf": 0.18186719563245113,
      "calib/step_conf_rate": 0.9453125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 318.5625,
      "completions/mean_terminated_length": 322.3399353027344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 60.0,
      "epoch": 0.029866666666666666,
      "grad_norm": 3.2336666584014893,
      "learning_rate": 4.777777777777778e-06,
      "loss": 0.049,
      "num_tokens": 6817855.0,
      "reward": 0.770412802696228,
      "reward_std": 0.4679546654224396,
      "rewards/accuracy_reward_step": 0.26171875,
      "rewards/final_brier_reward_step": 0.3472761809825897,
      "rewards/format_reward_step": 0.84375,
      "step": 28
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.2023973334238676e-09,
      "aux_brier/mean_group_std": 0.22325800543385368,
      "aux_brier/mean_r": 0.5106411767447425,
      "aux_brier/n_active_tok": 180.25,
      "aux_brier/n_groups": 10.875,
      "aux_brier/n_step_records": 45.0625,
      "calib/answer_extract_rate": 0.921875,
      "calib/auroc": 0.5476190476190476,
      "calib/avg_num_step_conf": 5.75390625,
      "calib/ece": 0.7278418803418804,
      "calib/final_conf_rate": 0.9140625,
      "calib/format_rate": 0.8671875,
      "calib/frac_conf_gt_0.9": 0.7777777777777778,
      "calib/gap": 0.04591145833333332,
      "calib/mean_conf": 0.9073290598290599,
      "calib/mu_c": 0.945,
      "calib/mu_w": 0.8990885416666666,
      "calib/nonempty_final_conf_rate": 0.9140625,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.96875,
      "calib/pce": 0.7278418803418804,
      "calib/std_conf": 0.16810984990225533,
      "calib/step_conf_rate": 0.96875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2094.0,
      "completions/max_terminated_length": 2094.0,
      "completions/mean_length": 341.28515625,
      "completions/mean_terminated_length": 341.28515625,
      "completions/min_length": 35.0,
      "completions/min_terminated_length": 35.0,
      "epoch": 0.030933333333333334,
      "grad_norm": 7.195005893707275,
      "learning_rate": 4.75e-06,
      "loss": 0.0357,
      "num_tokens": 7012352.0,
      "reward": 0.6658509373664856,
      "reward_std": 0.4067962169647217,
      "rewards/accuracy_reward_step": 0.1640625,
      "rewards/final_brier_reward_step": 0.2727789282798767,
      "rewards/format_reward_step": 0.8671875,
      "step": 29
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.267420791210675e-09,
      "aux_brier/mean_group_std": 0.21119422611148528,
      "aux_brier/mean_r": 0.5397748448564954,
      "aux_brier/n_active_tok": 170.25,
      "aux_brier/n_groups": 9.6875,
      "aux_brier/n_step_records": 42.5625,
      "calib/answer_extract_rate": 0.9375,
      "calib/auroc": 0.5150462962962963,
      "calib/avg_num_step_conf": 5.3984375,
      "calib/ece": 0.6678237693389593,
      "calib/final_conf_rate": 0.92578125,
      "calib/format_rate": 0.890625,
      "calib/frac_conf_gt_0.9": 0.6286919831223629,
      "calib/gap": 0.05370184082892415,
      "calib/mean_conf": 0.865714064697609,
      "calib/mu_c": 0.9085395833333333,
      "calib/mu_w": 0.8548377425044091,
      "calib/nonempty_final_conf_rate": 0.92578125,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.9609375,
      "calib/pce": 0.665503094233474,
      "calib/std_conf": 0.21664601599961172,
      "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": 3028.0,
      "completions/max_terminated_length": 3028.0,
      "completions/mean_length": 345.0859375,
      "completions/mean_terminated_length": 346.4392395019531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 47.0,
      "epoch": 0.032,
      "grad_norm": 0.8084984421730042,
      "learning_rate": 4.722222222222222e-06,
      "loss": 0.0394,
      "num_tokens": 7207678.0,
      "reward": 0.7224695682525635,
      "reward_std": 0.42729613184928894,
      "rewards/accuracy_reward_step": 0.1953125,
      "rewards/final_brier_reward_step": 0.3273780047893524,
      "rewards/format_reward_step": 0.890625,
      "step": 30
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.4573048948141043e-08,
      "aux_brier/mean_group_std": 0.19195187264691851,
      "aux_brier/mean_r": 0.5383719147935322,
      "aux_brier/n_active_tok": 179.0,
      "aux_brier/n_groups": 10.125,
      "aux_brier/n_step_records": 44.75,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.4319251456745854,
      "calib/avg_num_step_conf": 5.69921875,
      "calib/ece": 0.68948625,
      "calib/final_conf_rate": 0.9375,
      "calib/format_rate": 0.921875,
      "calib/frac_conf_gt_0.9": 0.6666666666666666,
      "calib/gap": 0.005395764231286182,
      "calib/mean_conf": 0.87448625,
      "calib/mu_c": 0.8788478260869563,
      "calib/mu_w": 0.8734520618556701,
      "calib/nonempty_final_conf_rate": 0.9375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.6861529166666667,
      "calib/std_conf": 0.195830454141512,
      "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": 2112.0,
      "completions/max_terminated_length": 2112.0,
      "completions/mean_length": 319.078125,
      "completions/mean_terminated_length": 320.3294372558594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 54.0,
      "epoch": 0.03306666666666667,
      "grad_norm": 0.4686136841773987,
      "learning_rate": 4.694444444444445e-06,
      "loss": 0.0378,
      "num_tokens": 7395274.0,
      "reward": 0.7197814583778381,
      "reward_std": 0.38697105646133423,
      "rewards/accuracy_reward_step": 0.1796875,
      "rewards/final_brier_reward_step": 0.31662583351135254,
      "rewards/format_reward_step": 0.921875,
      "step": 31
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.3076910377349833e-09,
      "aux_brier/mean_group_std": 0.23219807412544136,
      "aux_brier/mean_r": 0.6105224431929146,
      "aux_brier/n_active_tok": 180.875,
      "aux_brier/n_groups": 9.5625,
      "aux_brier/n_step_records": 45.21875,
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.4489905498281787,
      "calib/avg_num_step_conf": 5.6796875,
      "calib/ece": 0.6518677685950413,
      "calib/final_conf_rate": 0.9453125,
      "calib/format_rate": 0.8984375,
      "calib/frac_conf_gt_0.9": 0.5909090909090909,
      "calib/gap": -0.021713487972508716,
      "calib/mean_conf": 0.842198347107438,
      "calib/mu_c": 0.8247916666666666,
      "calib/mu_w": 0.8465051546391753,
      "calib/nonempty_final_conf_rate": 0.9453125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.6478595041322314,
      "calib/std_conf": 0.20719590361927231,
      "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": 0.00390625,
      "completions/max_length": 1583.0,
      "completions/max_terminated_length": 1583.0,
      "completions/mean_length": 307.84765625,
      "completions/mean_terminated_length": 309.054931640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 29.0,
      "epoch": 0.034133333333333335,
      "grad_norm": 0.2770783305168152,
      "learning_rate": 4.666666666666667e-06,
      "loss": 0.0358,
      "num_tokens": 7580787.0,
      "reward": 0.7233457565307617,
      "reward_std": 0.33742696046829224,
      "rewards/accuracy_reward_step": 0.1875,
      "rewards/final_brier_reward_step": 0.34650811553001404,
      "rewards/format_reward_step": 0.8984375,
      "step": 32
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -8.45548617234293e-09,
      "aux_brier/mean_group_std": 0.2160932896811416,
      "aux_brier/mean_r": 0.5734913379692018,
      "aux_brier/n_active_tok": 188.5,
      "aux_brier/n_groups": 10.15625,
      "aux_brier/n_step_records": 47.125,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5422324387896931,
      "calib/avg_num_step_conf": 5.953125,
      "calib/ece": 0.6710995934959348,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.921875,
      "calib/frac_conf_gt_0.9": 0.6382113821138211,
      "calib/gap": 0.01798636801026432,
      "calib/mean_conf": 0.8490670731707317,
      "calib/mu_c": 0.8636170212765958,
      "calib/mu_w": 0.8456306532663315,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.6645548780487804,
      "calib/std_conf": 0.2214576263245123,
      "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": 0.00390625,
      "completions/max_length": 1345.0,
      "completions/max_terminated_length": 1345.0,
      "completions/mean_length": 339.8515625,
      "completions/mean_terminated_length": 341.184326171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 100.0,
      "epoch": 0.0352,
      "grad_norm": 2.2709271907806396,
      "learning_rate": 4.638888888888889e-06,
      "loss": 0.0367,
      "num_tokens": 7774661.0,
      "reward": 0.735588788986206,
      "reward_std": 0.37076908349990845,
      "rewards/accuracy_reward_step": 0.1875,
      "rewards/final_brier_reward_step": 0.3486052453517914,
      "rewards/format_reward_step": 0.921875,
      "step": 33
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.514862030260744e-08,
      "aux_brier/mean_group_std": 0.22333645302988725,
      "aux_brier/mean_r": 0.6118464752477848,
      "aux_brier/n_active_tok": 177.875,
      "aux_brier/n_groups": 9.40625,
      "aux_brier/n_step_records": 44.46875,
      "calib/answer_extract_rate": 0.93359375,
      "calib/auroc": 0.49775705347656707,
      "calib/avg_num_step_conf": 5.62109375,
      "calib/ece": 0.6464166666666666,
      "calib/final_conf_rate": 0.9375,
      "calib/format_rate": 0.90234375,
      "calib/frac_conf_gt_0.9": 0.5541666666666667,
      "calib/gap": 0.0071561799079213895,
      "calib/mean_conf": 0.8169166666666666,
      "calib/mu_c": 0.8227906976744187,
      "calib/mu_w": 0.8156345177664973,
      "calib/nonempty_final_conf_rate": 0.9375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.6420833333333333,
      "calib/std_conf": 0.23541203535267452,
      "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": 1946.0,
      "completions/max_terminated_length": 1946.0,
      "completions/mean_length": 308.40234375,
      "completions/mean_terminated_length": 308.40234375,
      "completions/min_length": 34.0,
      "completions/min_terminated_length": 34.0,
      "epoch": 0.03626666666666667,
      "grad_norm": 6.610110759735107,
      "learning_rate": 4.611111111111112e-06,
      "loss": 0.0728,
      "num_tokens": 7958724.0,
      "reward": 0.7139061689376831,
      "reward_std": 0.3879733085632324,
      "rewards/accuracy_reward_step": 0.171875,
      "rewards/final_brier_reward_step": 0.36343711614608765,
      "rewards/format_reward_step": 0.90234375,
      "step": 34
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.088691583579383e-08,
      "aux_brier/mean_group_std": 0.2220033179775186,
      "aux_brier/mean_r": 0.6694873177641538,
      "aux_brier/n_active_tok": 208.25,
      "aux_brier/n_groups": 11.75,
      "aux_brier/n_step_records": 52.0625,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.4580443828016643,
      "calib/avg_num_step_conf": 6.63671875,
      "calib/ece": 0.6261298387096774,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.93359375,
      "calib/frac_conf_gt_0.9": 0.5,
      "calib/gap": 0.005149283402681504,
      "calib/mean_conf": 0.7954846774193549,
      "calib/mu_c": 0.7997619047619047,
      "calib/mu_w": 0.7946126213592232,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.6261298387096774,
      "calib/std_conf": 0.24226275210098042,
      "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": 1544.0,
      "completions/max_terminated_length": 1544.0,
      "completions/mean_length": 342.21484375,
      "completions/mean_terminated_length": 343.556884765625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 96.0,
      "epoch": 0.037333333333333336,
      "grad_norm": 2.5186874866485596,
      "learning_rate": 4.583333333333333e-06,
      "loss": 0.0205,
      "num_tokens": 8155587.0,
      "reward": 0.7271343469619751,
      "reward_std": 0.38158851861953735,
      "rewards/accuracy_reward_step": 0.1640625,
      "rewards/final_brier_reward_step": 0.38509997725486755,
      "rewards/format_reward_step": 0.93359375,
      "step": 35
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 6.169562194457967e-09,
      "aux_brier/mean_group_std": 0.2203165829964861,
      "aux_brier/mean_r": 0.7090527466461379,
      "aux_brier/n_active_tok": 189.625,
      "aux_brier/n_groups": 10.28125,
      "aux_brier/n_step_records": 47.40625,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4370542635658915,
      "calib/avg_num_step_conf": 5.97265625,
      "calib/ece": 0.482085020242915,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.4089068825910931,
      "calib/gap": -0.02508565891472858,
      "calib/mean_conf": 0.7729352226720648,
      "calib/mu_c": 0.7554666666666667,
      "calib/mu_w": 0.7805523255813953,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.4756882591093118,
      "calib/std_conf": 0.24132308180527398,
      "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": 1496.0,
      "completions/max_terminated_length": 1496.0,
      "completions/mean_length": 313.98046875,
      "completions/mean_terminated_length": 313.98046875,
      "completions/min_length": 58.0,
      "completions/min_terminated_length": 58.0,
      "epoch": 0.0384,
      "grad_norm": 1.0801153182983398,
      "learning_rate": 4.555555555555556e-06,
      "loss": 0.0461,
      "num_tokens": 8338678.0,
      "reward": 0.8887618184089661,
      "reward_std": 0.43464434146881104,
      "rewards/accuracy_reward_step": 0.296875,
      "rewards/final_brier_reward_step": 0.4769221842288971,
      "rewards/format_reward_step": 0.9453125,
      "step": 36
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -8.800347457704749e-09,
      "aux_brier/mean_group_std": 0.1925654434566419,
      "aux_brier/mean_r": 0.7558997547162454,
      "aux_brier/n_active_tok": 197.125,
      "aux_brier/n_groups": 11.15625,
      "aux_brier/n_step_records": 49.28125,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.4782876501385894,
      "calib/avg_num_step_conf": 6.234375,
      "calib/ece": 0.508456611570248,
      "calib/final_conf_rate": 0.9453125,
      "calib/format_rate": 0.9140625,
      "calib/frac_conf_gt_0.9": 0.33884297520661155,
      "calib/gap": 0.009708294836259235,
      "calib/mean_conf": 0.7192004132231404,
      "calib/mu_c": 0.7268627450980393,
      "calib/mu_w": 0.71715445026178,
      "calib/nonempty_final_conf_rate": 0.9453125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.508456611570248,
      "calib/std_conf": 0.25263505945195347,
      "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": 2816.0,
      "completions/max_terminated_length": 2816.0,
      "completions/mean_length": 330.3359375,
      "completions/mean_terminated_length": 331.6313781738281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 17.0,
      "epoch": 0.039466666666666664,
      "grad_norm": 0.6208831071853638,
      "learning_rate": 4.527777777777778e-06,
      "loss": 0.0345,
      "num_tokens": 8530340.0,
      "reward": 0.782370924949646,
      "reward_std": 0.41898417472839355,
      "rewards/accuracy_reward_step": 0.20703125,
      "rewards/final_brier_reward_step": 0.4732334613800049,
      "rewards/format_reward_step": 0.9140625,
      "step": 37
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.332314524318036e-08,
      "aux_brier/mean_group_std": 0.17843746533874077,
      "aux_brier/mean_r": 0.7999370721329889,
      "aux_brier/n_active_tok": 207.75,
      "aux_brier/n_groups": 12.4375,
      "aux_brier/n_step_records": 51.9375,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5330945208825206,
      "calib/avg_num_step_conf": 6.6015625,
      "calib/ece": 0.40371311475409843,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.2827868852459016,
      "calib/gap": 0.028847512822600452,
      "calib/mean_conf": 0.6736311475409835,
      "calib/mu_c": 0.6940845070422537,
      "calib/mu_w": 0.6652369942196532,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.3931803278688525,
      "calib/std_conf": 0.277502464937204,
      "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": 2271.0,
      "completions/max_terminated_length": 2271.0,
      "completions/mean_length": 350.91796875,
      "completions/mean_terminated_length": 352.29412841796875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 116.0,
      "epoch": 0.04053333333333333,
      "grad_norm": 0.953909158706665,
      "learning_rate": 4.5e-06,
      "loss": 0.0602,
      "num_tokens": 8727063.0,
      "reward": 0.8888434171676636,
      "reward_std": 0.40952152013778687,
      "rewards/accuracy_reward_step": 0.28125,
      "rewards/final_brier_reward_step": 0.5475612878799438,
      "rewards/format_reward_step": 0.94140625,
      "step": 38
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.5018593230164967e-08,
      "aux_brier/mean_group_std": 0.15156664724203292,
      "aux_brier/mean_r": 0.810969457909356,
      "aux_brier/n_active_tok": 192.75,
      "aux_brier/n_groups": 9.90625,
      "aux_brier/n_step_records": 48.1875,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4874641148325359,
      "calib/avg_num_step_conf": 6.1484375,
      "calib/ece": 0.42951183673469384,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.24081632653061225,
      "calib/gap": -0.0030888038277511187,
      "calib/mean_conf": 0.6313044897959184,
      "calib/mu_c": 0.628909090909091,
      "calib/mu_w": 0.6319978947368421,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.4181632653061224,
      "calib/std_conf": 0.2752871157578769,
      "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": 1205.0,
      "completions/max_terminated_length": 1205.0,
      "completions/mean_length": 329.890625,
      "completions/mean_terminated_length": 331.184326171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.0416,
      "grad_norm": 0.23614369332790375,
      "learning_rate": 4.472222222222223e-06,
      "loss": 0.0236,
      "num_tokens": 8917603.0,
      "reward": 0.8228167295455933,
      "reward_std": 0.32237982749938965,
      "rewards/accuracy_reward_step": 0.21484375,
      "rewards/final_brier_reward_step": 0.5490795373916626,
      "rewards/format_reward_step": 0.94140625,
      "step": 39
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 4.045960014242844e-08,
      "aux_brier/mean_group_std": 0.13373408419202218,
      "aux_brier/mean_r": 0.8406878158306899,
      "aux_brier/n_active_tok": 215.625,
      "aux_brier/n_groups": 13.15625,
      "aux_brier/n_step_records": 53.90625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5196078431372548,
      "calib/avg_num_step_conf": 6.77734375,
      "calib/ece": 0.4654435483870968,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.27419354838709675,
      "calib/gap": 0.015691251119737126,
      "calib/mean_conf": 0.6616532258064516,
      "calib/mu_c": 0.6741176470588235,
      "calib/mu_w": 0.6584263959390864,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.4607258064516129,
      "calib/std_conf": 0.2739056096056909,
      "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": 2065.0,
      "completions/max_terminated_length": 2065.0,
      "completions/mean_length": 357.421875,
      "completions/mean_terminated_length": 357.421875,
      "completions/min_length": 11.0,
      "completions/min_terminated_length": 11.0,
      "epoch": 0.042666666666666665,
      "grad_norm": 0.7531865835189819,
      "learning_rate": 4.444444444444444e-06,
      "loss": 0.0167,
      "num_tokens": 9115863.0,
      "reward": 0.8074332475662231,
      "reward_std": 0.3995877206325531,
      "rewards/accuracy_reward_step": 0.203125,
      "rewards/final_brier_reward_step": 0.5266082286834717,
      "rewards/format_reward_step": 0.9453125,
      "step": 40
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 7.874983145098047e-08,
      "aux_brier/mean_group_std": 0.1312267855764217,
      "aux_brier/mean_r": 0.8464493463466005,
      "aux_brier/n_active_tok": 202.125,
      "aux_brier/n_groups": 11.8125,
      "aux_brier/n_step_records": 50.53125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4947436245252306,
      "calib/avg_num_step_conf": 6.36328125,
      "calib/ece": 0.30325301204819277,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.1646586345381526,
      "calib/gap": -0.005603635377102734,
      "calib/mean_conf": 0.5608433734939758,
      "calib/mu_c": 0.5574226804123711,
      "calib/mu_w": 0.5630263157894738,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.2372690763052209,
      "calib/std_conf": 0.29027146200216886,
      "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": 1700.0,
      "completions/max_terminated_length": 1700.0,
      "completions/mean_length": 329.859375,
      "completions/mean_terminated_length": 331.1529541015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 64.0,
      "epoch": 0.04373333333333333,
      "grad_norm": 0.32473674416542053,
      "learning_rate": 4.416666666666667e-06,
      "loss": 0.0376,
      "num_tokens": 9307555.0,
      "reward": 1.0193958282470703,
      "reward_std": 0.39183253049850464,
      "rewards/accuracy_reward_step": 0.3828125,
      "rewards/final_brier_reward_step": 0.6244581937789917,
      "rewards/format_reward_step": 0.9609375,
      "step": 41
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.636152155316429e-07,
      "aux_brier/mean_group_std": 0.11594292246522589,
      "aux_brier/mean_r": 0.8885825685236638,
      "aux_brier/n_active_tok": 182.75,
      "aux_brier/n_groups": 9.9375,
      "aux_brier/n_step_records": 45.6875,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.4811442989634479,
      "calib/avg_num_step_conf": 5.7734375,
      "calib/ece": 0.279268,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.172,
      "calib/gap": -0.024231178396071917,
      "calib/mean_conf": 0.5247159999999998,
      "calib/mu_c": 0.5095957446808511,
      "calib/mu_w": 0.533826923076923,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.21399200000000002,
      "calib/std_conf": 0.30292278115717874,
      "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": 1987.0,
      "completions/max_terminated_length": 1987.0,
      "completions/mean_length": 292.33203125,
      "completions/mean_terminated_length": 292.33203125,
      "completions/min_length": 73.0,
      "completions/min_terminated_length": 73.0,
      "epoch": 0.0448,
      "grad_norm": 0.9810383915901184,
      "learning_rate": 4.388888888888889e-06,
      "loss": 0.0275,
      "num_tokens": 9486760.0,
      "reward": 0.9959656000137329,
      "reward_std": 0.3806919455528259,
      "rewards/accuracy_reward_step": 0.3671875,
      "rewards/final_brier_reward_step": 0.6166751384735107,
      "rewards/format_reward_step": 0.94921875,
      "step": 42
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.5842941802207688e-07,
      "aux_brier/mean_group_std": 0.09925830655635783,
      "aux_brier/mean_r": 0.895804819442369,
      "aux_brier/n_active_tok": 201.0,
      "aux_brier/n_groups": 13.15625,
      "aux_brier/n_step_records": 50.25,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.48248781368478066,
      "calib/avg_num_step_conf": 6.37890625,
      "calib/ece": 0.32383132530120484,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.1686746987951807,
      "calib/gap": -0.024683336342300022,
      "calib/mean_conf": 0.517726907630522,
      "calib/mu_c": 0.4987931034482759,
      "calib/mu_w": 0.5234764397905759,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.30431325301204826,
      "calib/std_conf": 0.2982098492263088,
      "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": 2567.0,
      "completions/max_terminated_length": 2567.0,
      "completions/mean_length": 347.87890625,
      "completions/mean_terminated_length": 347.87890625,
      "completions/min_length": 79.0,
      "completions/min_terminated_length": 79.0,
      "epoch": 0.04586666666666667,
      "grad_norm": 0.764629602432251,
      "learning_rate": 4.361111111111112e-06,
      "loss": 0.0761,
      "num_tokens": 9681041.0,
      "reward": 0.8540087938308716,
      "reward_std": 0.35527944564819336,
      "rewards/accuracy_reward_step": 0.2265625,
      "rewards/final_brier_reward_step": 0.6113476753234863,
      "rewards/format_reward_step": 0.94921875,
      "step": 43
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 7.65690757575932e-07,
      "aux_brier/mean_group_std": 0.0811326167114847,
      "aux_brier/mean_r": 0.920624391641562,
      "aux_brier/n_active_tok": 215.25,
      "aux_brier/n_groups": 13.03125,
      "aux_brier/n_step_records": 53.8125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.4535250631085468,
      "calib/avg_num_step_conf": 6.765625,
      "calib/ece": 0.2585951417004048,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.08502024291497975,
      "calib/gap": -0.06511206274792636,
      "calib/mean_conf": 0.4246437246963562,
      "calib/mu_c": 0.3750847457627119,
      "calib/mu_w": 0.44019680851063825,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.22218623481781374,
      "calib/std_conf": 0.2849744831802655,
      "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": 2513.0,
      "completions/max_terminated_length": 2513.0,
      "completions/mean_length": 345.7109375,
      "completions/mean_terminated_length": 345.7109375,
      "completions/min_length": 76.0,
      "completions/min_terminated_length": 76.0,
      "epoch": 0.046933333333333334,
      "grad_norm": 0.33147311210632324,
      "learning_rate": 4.333333333333334e-06,
      "loss": 0.0725,
      "num_tokens": 9875863.0,
      "reward": 0.8748396635055542,
      "reward_std": 0.3271452784538269,
      "rewards/accuracy_reward_step": 0.234375,
      "rewards/final_brier_reward_step": 0.647796094417572,
      "rewards/format_reward_step": 0.95703125,
      "step": 44
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.765974979086373e-07,
      "aux_brier/mean_group_std": 0.08300107892415177,
      "aux_brier/mean_r": 0.9306819157320256,
      "aux_brier/n_active_tok": 203.25,
      "aux_brier/n_groups": 11.65625,
      "aux_brier/n_step_records": 50.8125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.44570881226053644,
      "calib/avg_num_step_conf": 6.41796875,
      "calib/ece": 0.27049397590361446,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0642570281124498,
      "calib/gap": -0.046901149425287336,
      "calib/mean_conf": 0.37124096385542166,
      "calib/mu_c": 0.3384666666666667,
      "calib/mu_w": 0.38536781609195403,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.17026506024096388,
      "calib/std_conf": 0.28194382321033484,
      "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": 2254.0,
      "completions/max_terminated_length": 2254.0,
      "completions/mean_length": 330.65234375,
      "completions/mean_terminated_length": 331.94903564453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 99.0,
      "epoch": 0.048,
      "grad_norm": 0.18330220878124237,
      "learning_rate": 4.305555555555556e-06,
      "loss": 0.0437,
      "num_tokens": 10065558.0,
      "reward": 0.9469988346099854,
      "reward_std": 0.3695826530456543,
      "rewards/accuracy_reward_step": 0.296875,
      "rewards/final_brier_reward_step": 0.6629955768585205,
      "rewards/format_reward_step": 0.96875,
      "step": 45
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.4409363044361476e-07,
      "aux_brier/mean_group_std": 0.08869492678138623,
      "aux_brier/mean_r": 0.9119198388583334,
      "aux_brier/n_active_tok": 225.75,
      "aux_brier/n_groups": 15.34375,
      "aux_brier/n_step_records": 56.4375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4519160951996772,
      "calib/avg_num_step_conf": 7.08203125,
      "calib/ece": 0.2646785714285714,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.10317460317460317,
      "calib/gap": -0.0675923356192013,
      "calib/mean_conf": 0.3448452380952381,
      "calib/mu_c": 0.295223880597015,
      "calib/mu_w": 0.3628162162162163,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.17182539682539683,
      "calib/std_conf": 0.29150665654605146,
      "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": 2304.0,
      "completions/max_terminated_length": 2304.0,
      "completions/mean_length": 347.5,
      "completions/mean_terminated_length": 347.5,
      "completions/min_length": 105.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.04906666666666667,
      "grad_norm": 0.10056949406862259,
      "learning_rate": 4.277777777777778e-06,
      "loss": 0.0901,
      "num_tokens": 10259286.0,
      "reward": 0.9220494627952576,
      "reward_std": 0.24605713784694672,
      "rewards/accuracy_reward_step": 0.265625,
      "rewards/final_brier_reward_step": 0.6725727319717407,
      "rewards/format_reward_step": 0.9765625,
      "step": 46
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 9.879659923495154e-08,
      "aux_brier/mean_group_std": 0.07213828201589893,
      "aux_brier/mean_r": 0.9364409053049382,
      "aux_brier/n_active_tok": 231.0,
      "aux_brier/n_groups": 14.78125,
      "aux_brier/n_step_records": 57.75,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.45559593023255807,
      "calib/avg_num_step_conf": 7.39453125,
      "calib/ece": 0.2782520325203252,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.06504065040650407,
      "calib/gap": -0.04196366279069774,
      "calib/mean_conf": 0.34264227642276424,
      "calib/mu_c": 0.3153488372093023,
      "calib/mu_w": 0.35731250000000003,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.13565040650406504,
      "calib/std_conf": 0.2758289540117945,
      "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": 2426.0,
      "completions/max_terminated_length": 2426.0,
      "completions/mean_length": 372.76953125,
      "completions/mean_terminated_length": 374.23138427734375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 135.0,
      "epoch": 0.050133333333333335,
      "grad_norm": 0.22952257096767426,
      "learning_rate": 4.25e-06,
      "loss": 0.0817,
      "num_tokens": 10460691.0,
      "reward": 0.9732900261878967,
      "reward_std": 0.3508404493331909,
      "rewards/accuracy_reward_step": 0.3359375,
      "rewards/final_brier_reward_step": 0.6431602239608765,
      "rewards/format_reward_step": 0.953125,
      "step": 47
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.0490021446792923e-07,
      "aux_brier/mean_group_std": 0.06687772576759139,
      "aux_brier/mean_r": 0.9495411315696214,
      "aux_brier/n_active_tok": 200.25,
      "aux_brier/n_groups": 10.40625,
      "aux_brier/n_step_records": 50.0625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.39001855287569576,
      "calib/avg_num_step_conf": 6.3046875,
      "calib/ece": 0.27710317460317463,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.027777777777777776,
      "calib/gap": -0.07735584415584418,
      "calib/mean_conf": 0.2681349206349206,
      "calib/mu_c": 0.2144155844155844,
      "calib/mu_w": 0.29177142857142857,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.11984126984126983,
      "calib/std_conf": 0.23697015595280196,
      "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": 1802.0,
      "completions/max_terminated_length": 1802.0,
      "completions/mean_length": 327.828125,
      "completions/mean_terminated_length": 329.1137390136719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 97.0,
      "epoch": 0.0512,
      "grad_norm": 0.18228919804096222,
      "learning_rate": 4.222222222222223e-06,
      "loss": 0.0133,
      "num_tokens": 10648303.0,
      "reward": 0.9566383361816406,
      "reward_std": 0.31275951862335205,
      "rewards/accuracy_reward_step": 0.30078125,
      "rewards/final_brier_reward_step": 0.6781160235404968,
      "rewards/format_reward_step": 0.97265625,
      "step": 48
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.0355335841594204e-07,
      "aux_brier/mean_group_std": 0.07027406013126304,
      "aux_brier/mean_r": 0.9404911111550787,
      "aux_brier/n_active_tok": 221.125,
      "aux_brier/n_groups": 13.5,
      "aux_brier/n_step_records": 55.28125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5274584929757344,
      "calib/avg_num_step_conf": 6.94921875,
      "calib/ece": 0.25136546184738956,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0321285140562249,
      "calib/gap": -0.004244359301830591,
      "calib/mean_conf": 0.24839357429718875,
      "calib/mu_c": 0.24563218390804598,
      "calib/mu_w": 0.24987654320987657,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.07518072289156627,
      "calib/std_conf": 0.23972002938601275,
      "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": 2439.0,
      "completions/max_terminated_length": 2439.0,
      "completions/mean_length": 354.859375,
      "completions/mean_terminated_length": 354.859375,
      "completions/min_length": 37.0,
      "completions/min_terminated_length": 37.0,
      "epoch": 0.05226666666666667,
      "grad_norm": 0.07151960581541061,
      "learning_rate": 4.194444444444445e-06,
      "loss": 0.0643,
      "num_tokens": 10843683.0,
      "reward": 0.9951850771903992,
      "reward_std": 0.3210059404373169,
      "rewards/accuracy_reward_step": 0.34375,
      "rewards/final_brier_reward_step": 0.6760527491569519,
      "rewards/format_reward_step": 0.96484375,
      "step": 49
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 9.784847908589356e-07,
      "aux_brier/mean_group_std": 0.055116995024101514,
      "aux_brier/mean_r": 0.9598924751937429,
      "aux_brier/n_active_tok": 224.75,
      "aux_brier/n_groups": 13.90625,
      "aux_brier/n_step_records": 56.1875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.4138856280395774,
      "calib/avg_num_step_conf": 7.078125,
      "calib/ece": 0.24460040816326534,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.05714285714285714,
      "calib/gap": -0.038986609089384505,
      "calib/mean_conf": 0.2233995918367347,
      "calib/mu_c": 0.19507462686567167,
      "calib/mu_w": 0.23406123595505618,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.09726530612244898,
      "calib/std_conf": 0.2539838697269298,
      "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": 2101.0,
      "completions/max_terminated_length": 2101.0,
      "completions/mean_length": 381.96484375,
      "completions/mean_terminated_length": 381.96484375,
      "completions/min_length": 134.0,
      "completions/min_terminated_length": 134.0,
      "epoch": 0.05333333333333334,
      "grad_norm": 0.08293453603982925,
      "learning_rate": 4.166666666666667e-06,
      "loss": 0.0643,
      "num_tokens": 11046826.0,
      "reward": 0.9112589955329895,
      "reward_std": 0.33396610617637634,
      "rewards/accuracy_reward_step": 0.26953125,
      "rewards/final_brier_reward_step": 0.6762858629226685,
      "rewards/format_reward_step": 0.9453125,
      "step": 50
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 7.984037968533642e-07,
      "aux_brier/mean_group_std": 0.05989052847027103,
      "aux_brier/mean_r": 0.9579583998545194,
      "aux_brier/n_active_tok": 228.625,
      "aux_brier/n_groups": 13.46875,
      "aux_brier/n_step_records": 57.15625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5154973231896309,
      "calib/avg_num_step_conf": 7.265625,
      "calib/ece": 0.24108142292490115,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.043478260869565216,
      "calib/gap": 0.007696703296703272,
      "calib/mean_conf": 0.22200158102766795,
      "calib/mu_c": 0.2271428571428571,
      "calib/mu_w": 0.21944615384615382,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.06553359683794466,
      "calib/std_conf": 0.2472112133968055,
      "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": 1084.0,
      "completions/max_terminated_length": 1084.0,
      "completions/mean_length": 360.72265625,
      "completions/mean_terminated_length": 362.13726806640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 87.0,
      "epoch": 0.0544,
      "grad_norm": 0.16797003149986267,
      "learning_rate": 4.138888888888889e-06,
      "loss": 0.007,
      "num_tokens": 11248467.0,
      "reward": 0.9927654266357422,
      "reward_std": 0.2776499092578888,
      "rewards/accuracy_reward_step": 0.328125,
      "rewards/final_brier_reward_step": 0.697624146938324,
      "rewards/format_reward_step": 0.98046875,
      "step": 51
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.3240418723636793e-06,
      "aux_brier/mean_group_std": 0.05680214748533047,
      "aux_brier/mean_r": 0.9560514305138509,
      "aux_brier/n_active_tok": 206.375,
      "aux_brier/n_groups": 11.3125,
      "aux_brier/n_step_records": 51.59375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4833976833976834,
      "calib/avg_num_step_conf": 6.5234375,
      "calib/ece": 0.3255098814229249,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.03162055335968379,
      "calib/gap": -0.03864414414414416,
      "calib/mean_conf": 0.16927272727272727,
      "calib/mu_c": 0.14666666666666667,
      "calib/mu_w": 0.18531081081081083,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.03988142292490118,
      "calib/std_conf": 0.2041630188551689,
      "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": 1201.0,
      "completions/max_terminated_length": 1201.0,
      "completions/mean_length": 344.078125,
      "completions/mean_terminated_length": 345.4274597167969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 121.0,
      "epoch": 0.055466666666666664,
      "grad_norm": 0.2156234234571457,
      "learning_rate": 4.111111111111111e-06,
      "loss": 0.0292,
      "num_tokens": 11444503.0,
      "reward": 1.0537123680114746,
      "reward_std": 0.3458700180053711,
      "rewards/accuracy_reward_step": 0.41015625,
      "rewards/final_brier_reward_step": 0.6210998296737671,
      "rewards/format_reward_step": 0.9765625,
      "step": 52
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.8243179855659264e-07,
      "aux_brier/mean_group_std": 0.0626040370378366,
      "aux_brier/mean_r": 0.9475969372039725,
      "aux_brier/n_active_tok": 245.5,
      "aux_brier/n_groups": 15.84375,
      "aux_brier/n_step_records": 61.375,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.49686520376175547,
      "calib/avg_num_step_conf": 7.75,
      "calib/ece": 0.306639344262295,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.012295081967213115,
      "calib/gap": -0.0031083246255660113,
      "calib/mean_conf": 0.16204918032786889,
      "calib/mu_c": 0.1602020202020202,
      "calib/mu_w": 0.16331034482758622,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.031475409836065574,
      "calib/std_conf": 0.1878891961777253,
      "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": 2851.0,
      "completions/max_terminated_length": 2851.0,
      "completions/mean_length": 425.61328125,
      "completions/mean_terminated_length": 425.61328125,
      "completions/min_length": 106.0,
      "completions/min_terminated_length": 106.0,
      "epoch": 0.05653333333333333,
      "grad_norm": 0.20086102187633514,
      "learning_rate": 4.083333333333334e-06,
      "loss": 0.064,
      "num_tokens": 11659284.0,
      "reward": 1.0182605981826782,
      "reward_std": 0.38351237773895264,
      "rewards/accuracy_reward_step": 0.390625,
      "rewards/final_brier_reward_step": 0.6277297139167786,
      "rewards/format_reward_step": 0.94140625,
      "step": 53
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.044353998762518e-06,
      "aux_brier/mean_group_std": 0.07300635731622351,
      "aux_brier/mean_r": 0.9530345758029888,
      "aux_brier/n_active_tok": 220.875,
      "aux_brier/n_groups": 13.0625,
      "aux_brier/n_step_records": 55.21875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.49387254901960786,
      "calib/avg_num_step_conf": 7.00390625,
      "calib/ece": 0.3587075098814229,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.007905138339920948,
      "calib/gap": -0.005667546505781773,
      "calib/mean_conf": 0.12295256916996049,
      "calib/mu_c": 0.11990598290598292,
      "calib/mu_w": 0.1255735294117647,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00960474308300395,
      "calib/std_conf": 0.14050042452540257,
      "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": 1032.0,
      "completions/max_terminated_length": 1032.0,
      "completions/mean_length": 349.11328125,
      "completions/mean_terminated_length": 349.11328125,
      "completions/min_length": 85.0,
      "completions/min_terminated_length": 85.0,
      "epoch": 0.0576,
      "grad_norm": 0.32629361748695374,
      "learning_rate": 4.055555555555556e-06,
      "loss": -0.0086,
      "num_tokens": 11854889.0,
      "reward": 1.1027724742889404,
      "reward_std": 0.30096855759620667,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.6064023375511169,
      "rewards/format_reward_step": 0.98828125,
      "step": 54
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.1379246602216675e-06,
      "aux_brier/mean_group_std": 0.05071299854557984,
      "aux_brier/mean_r": 0.9610989423086578,
      "aux_brier/n_active_tok": 240.0,
      "aux_brier/n_groups": 14.75,
      "aux_brier/n_step_records": 60.0,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.44242015294646875,
      "calib/avg_num_step_conf": 7.58203125,
      "calib/ece": 0.2533530120481928,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.01606425702811245,
      "calib/gap": -0.007733040935672536,
      "calib/mean_conf": 0.13339397590361446,
      "calib/mu_c": 0.12808333333333333,
      "calib/mu_w": 0.13581637426900586,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.036746987951807225,
      "calib/std_conf": 0.17100283894592416,
      "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": 2919.0,
      "completions/max_terminated_length": 2919.0,
      "completions/mean_length": 394.859375,
      "completions/mean_terminated_length": 394.859375,
      "completions/min_length": 93.0,
      "completions/min_terminated_length": 93.0,
      "epoch": 0.058666666666666666,
      "grad_norm": 0.1013939380645752,
      "learning_rate": 4.027777777777779e-06,
      "loss": 0.0937,
      "num_tokens": 12063797.0,
      "reward": 0.9612206816673279,
      "reward_std": 0.2767617404460907,
      "rewards/accuracy_reward_step": 0.30859375,
      "rewards/final_brier_reward_step": 0.6886326670646667,
      "rewards/format_reward_step": 0.9609375,
      "step": 55
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 4.685081207433939e-07,
      "aux_brier/mean_group_std": 0.05100227364623314,
      "aux_brier/mean_r": 0.9606296541145303,
      "aux_brier/n_active_tok": 242.75,
      "aux_brier/n_groups": 14.75,
      "aux_brier/n_step_records": 60.6875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4962231489495633,
      "calib/avg_num_step_conf": 7.68359375,
      "calib/ece": 0.229168,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.012,
      "calib/gap": -0.017303328349988184,
      "calib/mean_conf": 0.100192,
      "calib/mu_c": 0.08780281690140847,
      "calib/mu_w": 0.10510614525139665,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.02268,
      "calib/std_conf": 0.1443027759123157,
      "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": 2414.0,
      "completions/max_terminated_length": 2414.0,
      "completions/mean_length": 398.92578125,
      "completions/mean_terminated_length": 398.92578125,
      "completions/min_length": 113.0,
      "completions/min_terminated_length": 113.0,
      "epoch": 0.05973333333333333,
      "grad_norm": 0.03072967194020748,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0337,
      "num_tokens": 12272762.0,
      "reward": 0.9499537944793701,
      "reward_std": 0.25985878705978394,
      "rewards/accuracy_reward_step": 0.28515625,
      "rewards/final_brier_reward_step": 0.7138776183128357,
      "rewards/format_reward_step": 0.97265625,
      "step": 56
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -6.040157607412056e-07,
      "aux_brier/mean_group_std": 0.04829101798079878,
      "aux_brier/mean_r": 0.9699569409987028,
      "aux_brier/n_active_tok": 239.375,
      "aux_brier/n_groups": 16.5625,
      "aux_brier/n_step_records": 59.84375,
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5027433448486079,
      "calib/avg_num_step_conf": 7.5546875,
      "calib/ece": 0.39521311475409826,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.020491803278688523,
      "calib/gap": -0.0017990923254080915,
      "calib/mean_conf": 0.08937704918032786,
      "calib/mu_c": 0.0883963963963964,
      "calib/mu_w": 0.0901954887218045,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.014836065573770494,
      "calib/std_conf": 0.14308247809859057,
      "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": 3072.0,
      "completions/max_terminated_length": 3072.0,
      "completions/mean_length": 414.01953125,
      "completions/mean_terminated_length": 415.6431579589844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 19.0,
      "epoch": 0.0608,
      "grad_norm": 0.33087193965911865,
      "learning_rate": 3.972222222222223e-06,
      "loss": 0.0824,
      "num_tokens": 12485543.0,
      "reward": 1.0536420345306396,
      "reward_std": 0.32177644968032837,
      "rewards/accuracy_reward_step": 0.4375,
      "rewards/final_brier_reward_step": 0.566131055355072,
      "rewards/format_reward_step": 0.94921875,
      "step": 57
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.3840898555605996e-06,
      "aux_brier/mean_group_std": 0.042049013825433725,
      "aux_brier/mean_r": 0.9660714632088365,
      "aux_brier/n_active_tok": 296.875,
      "aux_brier/n_groups": 23.40625,
      "aux_brier/n_step_records": 74.21875,
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.4646460503650234,
      "calib/avg_num_step_conf": 9.4765625,
      "calib/ece": 0.26274166666666665,
      "calib/final_conf_rate": 0.9375,
      "calib/format_rate": 0.921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.025771962923468117,
      "calib/mean_conf": 0.08450833333333334,
      "calib/mu_c": 0.06657534246575343,
      "calib/mu_w": 0.09234730538922155,
      "calib/nonempty_final_conf_rate": 0.9375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.021541666666666664,
      "calib/std_conf": 0.10103291508491456,
      "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": 2606.0,
      "completions/max_terminated_length": 2606.0,
      "completions/mean_length": 494.7265625,
      "completions/mean_terminated_length": 496.66668701171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 144.0,
      "epoch": 0.06186666666666667,
      "grad_norm": 0.24724014103412628,
      "learning_rate": 3.944444444444445e-06,
      "loss": 0.1501,
      "num_tokens": 12718513.0,
      "reward": 0.9154220819473267,
      "reward_std": 0.31696373224258423,
      "rewards/accuracy_reward_step": 0.2890625,
      "rewards/final_brier_reward_step": 0.6616881489753723,
      "rewards/format_reward_step": 0.921875,
      "step": 58
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.0737893708923085e-06,
      "aux_brier/mean_group_std": 0.04600677352950576,
      "aux_brier/mean_r": 0.9679426185910952,
      "aux_brier/n_active_tok": 244.0,
      "aux_brier/n_groups": 16.0625,
      "aux_brier/n_step_records": 61.0,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5842910551093881,
      "calib/avg_num_step_conf": 7.66796875,
      "calib/ece": 0.31970281124498,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.004016064257028112,
      "calib/gap": -0.0007900166158958943,
      "calib/mean_conf": 0.07194377510040162,
      "calib/mu_c": 0.07144565217391304,
      "calib/mu_w": 0.07223566878980893,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.011084337349397589,
      "calib/std_conf": 0.0969266110162633,
      "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": 2307.0,
      "completions/max_terminated_length": 2307.0,
      "completions/mean_length": 425.6796875,
      "completions/mean_terminated_length": 425.6796875,
      "completions/min_length": 63.0,
      "completions/min_terminated_length": 63.0,
      "epoch": 0.06293333333333333,
      "grad_norm": 0.08634313941001892,
      "learning_rate": 3.916666666666667e-06,
      "loss": 0.0597,
      "num_tokens": 12933735.0,
      "reward": 1.0083181858062744,
      "reward_std": 0.2918752133846283,
      "rewards/accuracy_reward_step": 0.359375,
      "rewards/final_brier_reward_step": 0.6504605412483215,
      "rewards/format_reward_step": 0.97265625,
      "step": 59
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.6949528880180225e-06,
      "aux_brier/mean_group_std": 0.03616843391882674,
      "aux_brier/mean_r": 0.9745536707640898,
      "aux_brier/n_active_tok": 260.625,
      "aux_brier/n_groups": 18.9375,
      "aux_brier/n_step_records": 65.15625,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.46282598749428094,
      "calib/avg_num_step_conf": 8.30078125,
      "calib/ece": 0.26178775510204083,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.004081632653061225,
      "calib/gap": -0.008513268262925117,
      "calib/mean_conf": 0.07274285714285715,
      "calib/mu_c": 0.0669746835443038,
      "calib/mu_w": 0.07548795180722892,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0060408163265306125,
      "calib/std_conf": 0.08467365443337166,
      "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": 2996.0,
      "completions/max_terminated_length": 2996.0,
      "completions/mean_length": 440.3203125,
      "completions/mean_terminated_length": 440.3203125,
      "completions/min_length": 129.0,
      "completions/min_terminated_length": 129.0,
      "epoch": 0.064,
      "grad_norm": 0.017184965312480927,
      "learning_rate": 3.88888888888889e-06,
      "loss": 0.1188,
      "num_tokens": 13155313.0,
      "reward": 0.9585400819778442,
      "reward_std": 0.31010979413986206,
      "rewards/accuracy_reward_step": 0.31640625,
      "rewards/final_brier_reward_step": 0.6700977087020874,
      "rewards/format_reward_step": 0.94921875,
      "step": 60
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 7.387941521885821e-07,
      "aux_brier/mean_group_std": 0.052413984962238584,
      "aux_brier/mean_r": 0.9652360239111484,
      "aux_brier/n_active_tok": 221.75,
      "aux_brier/n_groups": 15.0,
      "aux_brier/n_step_records": 55.4375,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.4169480519480519,
      "calib/avg_num_step_conf": 6.984375,
      "calib/ece": 0.39087599999999995,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.014322077922077923,
      "calib/mean_conf": 0.052084000000000005,
      "calib/mu_c": 0.04406363636363636,
      "calib/mu_w": 0.058385714285714284,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00148,
      "calib/std_conf": 0.04510180643832351,
      "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": 2342.0,
      "completions/max_terminated_length": 2342.0,
      "completions/mean_length": 371.046875,
      "completions/mean_terminated_length": 371.046875,
      "completions/min_length": 101.0,
      "completions/min_terminated_length": 101.0,
      "epoch": 0.06506666666666666,
      "grad_norm": 0.10837318748235703,
      "learning_rate": 3.861111111111112e-06,
      "loss": 0.0767,
      "num_tokens": 13354365.0,
      "reward": 1.0542142391204834,
      "reward_std": 0.2764996886253357,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.5684198141098022,
      "rewards/format_reward_step": 0.96484375,
      "step": 61
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.868187364002406e-07,
      "aux_brier/mean_group_std": 0.04381141609160259,
      "aux_brier/mean_r": 0.9716618393481739,
      "aux_brier/n_active_tok": 263.25,
      "aux_brier/n_groups": 21.09375,
      "aux_brier/n_step_records": 65.8125,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4961106309420916,
      "calib/avg_num_step_conf": 8.4921875,
      "calib/ece": 0.3216856326530612,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.004081632653061225,
      "calib/gap": -0.012113928262748486,
      "calib/mean_conf": 0.0677021224489796,
      "calib/mu_c": 0.059988764044943826,
      "calib/mu_w": 0.07210269230769231,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.013061224489795919,
      "calib/std_conf": 0.10893542482040143,
      "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": 2354.0,
      "completions/max_terminated_length": 2354.0,
      "completions/mean_length": 471.04296875,
      "completions/mean_terminated_length": 471.04296875,
      "completions/min_length": 105.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.06613333333333334,
      "grad_norm": 0.0680285394191742,
      "learning_rate": 3.833333333333334e-06,
      "loss": 0.1161,
      "num_tokens": 13582032.0,
      "reward": 0.9762334227561951,
      "reward_std": 0.34210383892059326,
      "rewards/accuracy_reward_step": 0.34765625,
      "rewards/final_brier_reward_step": 0.6236836910247803,
      "rewards/format_reward_step": 0.9453125,
      "step": 62
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 6.40636866733324e-06,
      "aux_brier/mean_group_std": 0.05836626167407911,
      "aux_brier/mean_r": 0.960481865817939,
      "aux_brier/n_active_tok": 242.0,
      "aux_brier/n_groups": 14.84375,
      "aux_brier/n_step_records": 60.5,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.46469387755102043,
      "calib/avg_num_step_conf": 7.89453125,
      "calib/ece": 0.3517943548387097,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.004032258064516129,
      "calib/gap": -0.01288421768707481,
      "calib/mean_conf": 0.05941532258064517,
      "calib/mu_c": 0.051622448979591846,
      "calib/mu_w": 0.06450666666666666,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.008024193548387095,
      "calib/std_conf": 0.0885411837329467,
      "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": 2944.0,
      "completions/max_terminated_length": 2944.0,
      "completions/mean_length": 445.30078125,
      "completions/mean_terminated_length": 448.8070983886719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.0672,
      "grad_norm": 0.5678281188011169,
      "learning_rate": 3.8055555555555556e-06,
      "loss": 0.0613,
      "num_tokens": 13804669.0,
      "reward": 1.012493371963501,
      "reward_std": 0.34109848737716675,
      "rewards/accuracy_reward_step": 0.3828125,
      "rewards/final_brier_reward_step": 0.6046613454818726,
      "rewards/format_reward_step": 0.95703125,
      "step": 63
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -6.058569844147321e-06,
      "aux_brier/mean_group_std": 0.033108966515736575,
      "aux_brier/mean_r": 0.9747501982796533,
      "aux_brier/n_active_tok": 253.0,
      "aux_brier/n_groups": 17.96875,
      "aux_brier/n_step_records": 63.25,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.46250336292709177,
      "calib/avg_num_step_conf": 8.15625,
      "calib/ece": 0.44395368852459016,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.004098360655737705,
      "calib/gap": -0.011361864406779644,
      "calib/mean_conf": 0.05350532786885247,
      "calib/mu_c": 0.04763813559322035,
      "calib/mu_w": 0.059,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.006926229508196722,
      "calib/std_conf": 0.07581863560102291,
      "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": 2956.0,
      "completions/max_terminated_length": 2956.0,
      "completions/mean_length": 444.5078125,
      "completions/mean_terminated_length": 446.2510070800781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 132.0,
      "epoch": 0.06826666666666667,
      "grad_norm": 0.09891669452190399,
      "learning_rate": 3.777777777777778e-06,
      "loss": 0.0667,
      "num_tokens": 14022239.0,
      "reward": 1.0685046911239624,
      "reward_std": 0.32946476340293884,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5162064433097839,
      "rewards/format_reward_step": 0.94140625,
      "step": 64
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.775143360556065e-06,
      "aux_brier/mean_group_std": 0.03827895670426919,
      "aux_brier/mean_r": 0.9771603448523676,
      "aux_brier/n_active_tok": 206.0,
      "aux_brier/n_groups": 13.0,
      "aux_brier/n_step_records": 51.5,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.569016290726817,
      "calib/avg_num_step_conf": 6.640625,
      "calib/ece": 0.41270861767322836,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.010049043442731838,
      "calib/mean_conf": 0.03611027996456693,
      "calib/mu_c": 0.04164912280701755,
      "calib/mu_w": 0.03160007936428571,
      "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": 0.034171112056635054,
      "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": 2107.0,
      "completions/max_terminated_length": 2107.0,
      "completions/mean_length": 355.828125,
      "completions/mean_terminated_length": 355.828125,
      "completions/min_length": 118.0,
      "completions/min_terminated_length": 118.0,
      "epoch": 0.06933333333333333,
      "grad_norm": 0.033202022314071655,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.0187,
      "num_tokens": 14218355.0,
      "reward": 1.0857701301574707,
      "reward_std": 0.20154376327991486,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.5774555206298828,
      "rewards/format_reward_step": 0.984375,
      "step": 65
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.3306250042574574e-06,
      "aux_brier/mean_group_std": 0.04320104418073151,
      "aux_brier/mean_r": 0.9743550899357948,
      "aux_brier/n_active_tok": 265.875,
      "aux_brier/n_groups": 17.1875,
      "aux_brier/n_step_records": 66.46875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5000348383500557,
      "calib/avg_num_step_conf": 8.51953125,
      "calib/ece": 0.325970564516129,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.003624275362318838,
      "calib/mean_conf": 0.04499717741935485,
      "calib/mu_c": 0.042717391304347825,
      "calib/mu_w": 0.04634166666666666,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04443168496838503,
      "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": 2883.0,
      "completions/max_terminated_length": 2883.0,
      "completions/mean_length": 462.66796875,
      "completions/mean_terminated_length": 468.1541748046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 96.0,
      "epoch": 0.0704,
      "grad_norm": 0.3320234417915344,
      "learning_rate": 3.7222222222222225e-06,
      "loss": 0.0577,
      "num_tokens": 14443150.0,
      "reward": 0.9998792409896851,
      "reward_std": 0.28078022599220276,
      "rewards/accuracy_reward_step": 0.359375,
      "rewards/final_brier_reward_step": 0.6323295831680298,
      "rewards/format_reward_step": 0.96484375,
      "step": 66
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.064743734445096e-06,
      "aux_brier/mean_group_std": 0.04949975251528722,
      "aux_brier/mean_r": 0.9677108378406305,
      "aux_brier/n_active_tok": 236.375,
      "aux_brier/n_groups": 15.5625,
      "aux_brier/n_step_records": 59.09375,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.502158273381295,
      "calib/avg_num_step_conf": 7.41796875,
      "calib/ece": 0.398421686746988,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0009283845650752201,
      "calib/mean_conf": 0.043345381526104426,
      "calib/mu_c": 0.04386363636363637,
      "calib/mu_w": 0.04293525179856115,
      "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": 0.04063345295135549,
      "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": 3007.0,
      "completions/max_terminated_length": 3007.0,
      "completions/mean_length": 435.51953125,
      "completions/mean_terminated_length": 435.51953125,
      "completions/min_length": 118.0,
      "completions/min_terminated_length": 118.0,
      "epoch": 0.07146666666666666,
      "grad_norm": 0.05349089205265045,
      "learning_rate": 3.694444444444445e-06,
      "loss": 0.0561,
      "num_tokens": 14659651.0,
      "reward": 1.0583751201629639,
      "reward_std": 0.26450416445732117,
      "rewards/accuracy_reward_step": 0.43359375,
      "rewards/final_brier_reward_step": 0.5694377422332764,
      "rewards/format_reward_step": 0.96484375,
      "step": 67
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 8.650902338147581e-06,
      "aux_brier/mean_group_std": 0.03427101187295996,
      "aux_brier/mean_r": 0.9768078425432928,
      "aux_brier/n_active_tok": 204.5,
      "aux_brier/n_groups": 12.875,
      "aux_brier/n_step_records": 51.125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.46068627450980393,
      "calib/avg_num_step_conf": 6.5703125,
      "calib/ece": 0.36509486166007904,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.007049019607843134,
      "calib/mean_conf": 0.04676284584980237,
      "calib/mu_c": 0.0425,
      "calib/mu_w": 0.04954901960784314,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.008300395256916995,
      "calib/std_conf": 0.08306335385689934,
      "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": 2545.0,
      "completions/max_terminated_length": 2545.0,
      "completions/mean_length": 379.18359375,
      "completions/mean_terminated_length": 379.18359375,
      "completions/min_length": 57.0,
      "completions/min_terminated_length": 57.0,
      "epoch": 0.07253333333333334,
      "grad_norm": 0.48980000615119934,
      "learning_rate": 3.6666666666666666e-06,
      "loss": 0.0623,
      "num_tokens": 14860810.0,
      "reward": 1.0373156070709229,
      "reward_std": 0.25164324045181274,
      "rewards/accuracy_reward_step": 0.390625,
      "rewards/final_brier_reward_step": 0.6180124282836914,
      "rewards/format_reward_step": 0.984375,
      "step": 68
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.078681342526645e-06,
      "aux_brier/mean_group_std": 0.05264741637255274,
      "aux_brier/mean_r": 0.9663651153189144,
      "aux_brier/n_active_tok": 215.125,
      "aux_brier/n_groups": 13.15625,
      "aux_brier/n_step_records": 53.78125,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5235158862876255,
      "calib/avg_num_step_conf": 6.765625,
      "calib/ece": 0.3296016129032258,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0007544035674470484,
      "calib/mean_conf": 0.04459193548387098,
      "calib/mu_c": 0.044117391304347824,
      "calib/mu_w": 0.04487179487179487,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0016129032258064516,
      "calib/std_conf": 0.0434310235422943,
      "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": 2472.0,
      "completions/max_terminated_length": 2472.0,
      "completions/mean_length": 436.8203125,
      "completions/mean_terminated_length": 436.8203125,
      "completions/min_length": 7.0,
      "completions/min_terminated_length": 7.0,
      "epoch": 0.0736,
      "grad_norm": 0.05785338953137398,
      "learning_rate": 3.638888888888889e-06,
      "loss": 0.0771,
      "num_tokens": 15077132.0,
      "reward": 1.0069888830184937,
      "reward_std": 0.3025099039077759,
      "rewards/accuracy_reward_step": 0.36328125,
      "rewards/final_brier_reward_step": 0.6373307704925537,
      "rewards/format_reward_step": 0.96875,
      "step": 69
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.0532646641545647e-06,
      "aux_brier/mean_group_std": 0.039275487480210904,
      "aux_brier/mean_r": 0.9749847672449794,
      "aux_brier/n_active_tok": 231.0,
      "aux_brier/n_groups": 15.3125,
      "aux_brier/n_step_records": 57.75,
      "calib/answer_extract_rate": 0.94140625,
      "calib/auroc": 0.5602575231481481,
      "calib/avg_num_step_conf": 7.3359375,
      "calib/ece": 0.36384333333333335,
      "calib/final_conf_rate": 0.9375,
      "calib/format_rate": 0.93359375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.011329166666666675,
      "calib/mean_conf": 0.03615666666666668,
      "calib/mu_c": 0.042954166666666675,
      "calib/mu_w": 0.031625,
      "calib/nonempty_final_conf_rate": 0.9375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04059276769190404,
      "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": 2475.0,
      "completions/max_terminated_length": 2475.0,
      "completions/mean_length": 463.0703125,
      "completions/mean_terminated_length": 466.7165222167969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 67.0,
      "epoch": 0.07466666666666667,
      "grad_norm": 0.02348187379539013,
      "learning_rate": 3.6111111111111115e-06,
      "loss": 0.0838,
      "num_tokens": 15302670.0,
      "reward": 0.9888066053390503,
      "reward_std": 0.2804641127586365,
      "rewards/accuracy_reward_step": 0.375,
      "rewards/final_brier_reward_step": 0.5880390405654907,
      "rewards/format_reward_step": 0.93359375,
      "step": 70
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.4394772646374463e-08,
      "aux_brier/mean_group_std": 0.04867247399872675,
      "aux_brier/mean_r": 0.9634075252441048,
      "aux_brier/n_active_tok": 231.625,
      "aux_brier/n_groups": 14.71875,
      "aux_brier/n_step_records": 57.90625,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.4202279202279202,
      "calib/avg_num_step_conf": 7.4765625,
      "calib/ece": 0.33001343873517786,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.003952569169960474,
      "calib/gap": -0.011526427893094562,
      "calib/mean_conf": 0.039314624505928854,
      "calib/mu_c": 0.03193406593406593,
      "calib/mu_w": 0.043460493827160496,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.004822134387351778,
      "calib/std_conf": 0.06989159069698907,
      "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": 1896.0,
      "completions/max_terminated_length": 1896.0,
      "completions/mean_length": 422.97265625,
      "completions/mean_terminated_length": 426.30316162109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 111.0,
      "epoch": 0.07573333333333333,
      "grad_norm": 0.8687294125556946,
      "learning_rate": 3.5833333333333335e-06,
      "loss": 0.0305,
      "num_tokens": 15515359.0,
      "reward": 0.9972602725028992,
      "reward_std": 0.2537091076374054,
      "rewards/accuracy_reward_step": 0.35546875,
      "rewards/final_brier_reward_step": 0.6296659708023071,
      "rewards/format_reward_step": 0.96875,
      "step": 71
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.560168510720942e-06,
      "aux_brier/mean_group_std": 0.02920832451230697,
      "aux_brier/mean_r": 0.9761008591631589,
      "aux_brier/n_active_tok": 246.25,
      "aux_brier/n_groups": 16.15625,
      "aux_brier/n_step_records": 61.5625,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5556539420175783,
      "calib/avg_num_step_conf": 7.73046875,
      "calib/ece": 0.3620662450592885,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.003952569169960474,
      "calib/gap": 0.009279480519480507,
      "calib/mean_conf": 0.038566166007905134,
      "calib/mu_c": 0.044214545454545444,
      "calib/mu_w": 0.03493506493506494,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0046640316205533595,
      "calib/std_conf": 0.08305488922652862,
      "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": 3043.0,
      "completions/max_terminated_length": 3043.0,
      "completions/mean_length": 421.51171875,
      "completions/mean_terminated_length": 421.51171875,
      "completions/min_length": 130.0,
      "completions/min_terminated_length": 130.0,
      "epoch": 0.0768,
      "grad_norm": 0.30694445967674255,
      "learning_rate": 3.555555555555556e-06,
      "loss": 0.0423,
      "num_tokens": 15727674.0,
      "reward": 1.0391556024551392,
      "reward_std": 0.2421858310699463,
      "rewards/accuracy_reward_step": 0.390625,
      "rewards/final_brier_reward_step": 0.6253725290298462,
      "rewards/format_reward_step": 0.984375,
      "step": 72
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 8.052404001479552e-06,
      "aux_brier/mean_group_std": 0.04169540844016346,
      "aux_brier/mean_r": 0.9758179795916286,
      "aux_brier/n_active_tok": 227.875,
      "aux_brier/n_groups": 15.03125,
      "aux_brier/n_step_records": 56.96875,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5363523187052599,
      "calib/avg_num_step_conf": 7.26953125,
      "calib/ece": 0.5017786968503938,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004129070339246813,
      "calib/mean_conf": 0.029717366141732288,
      "calib/mu_c": 0.03165185185185185,
      "calib/mu_w": 0.027522781512605034,
      "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": 0.023934978877244266,
      "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": 2992.0,
      "completions/max_terminated_length": 2992.0,
      "completions/mean_length": 409.98828125,
      "completions/mean_terminated_length": 409.98828125,
      "completions/min_length": 108.0,
      "completions/min_terminated_length": 108.0,
      "epoch": 0.07786666666666667,
      "grad_norm": 0.05047299340367317,
      "learning_rate": 3.5277777777777784e-06,
      "loss": 0.043,
      "num_tokens": 15939663.0,
      "reward": 1.1476329565048218,
      "reward_std": 0.26514506340026855,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.4967818856239319,
      "rewards/format_reward_step": 0.9921875,
      "step": 73
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 6.184344703008549e-08,
      "aux_brier/mean_group_std": 0.023980905783726972,
      "aux_brier/mean_r": 0.9863200664382628,
      "aux_brier/n_active_tok": 241.375,
      "aux_brier/n_groups": 15.21875,
      "aux_brier/n_step_records": 60.34375,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.49262187088274045,
      "calib/avg_num_step_conf": 7.55859375,
      "calib/ece": 0.4146693548387097,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00381159420289855,
      "calib/mean_conf": 0.028879032258064517,
      "calib/mu_c": 0.031,
      "calib/mu_w": 0.02718840579710145,
      "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": 0.03330530635381605,
      "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": 2533.0,
      "completions/max_terminated_length": 2533.0,
      "completions/mean_length": 430.07421875,
      "completions/mean_terminated_length": 430.07421875,
      "completions/min_length": 115.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.07893333333333333,
      "grad_norm": 0.2849559485912323,
      "learning_rate": 3.5e-06,
      "loss": 0.0786,
      "num_tokens": 16153690.0,
      "reward": 1.0550175905227661,
      "reward_std": 0.25433430075645447,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.5638206005096436,
      "rewards/format_reward_step": 0.96875,
      "step": 74
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -6.103603219287379e-06,
      "aux_brier/mean_group_std": 0.05549154691148288,
      "aux_brier/mean_r": 0.9642726666882724,
      "aux_brier/n_active_tok": 216.375,
      "aux_brier/n_groups": 11.59375,
      "aux_brier/n_step_records": 54.09375,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6524572445800751,
      "calib/avg_num_step_conf": 6.82421875,
      "calib/ece": 0.5501291338582676,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.011855909466590379,
      "calib/mean_conf": 0.028611023622047243,
      "calib/mu_c": 0.03360544217687075,
      "calib/mu_w": 0.021749532710280375,
      "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": 0.02749721597459955,
      "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": 1224.0,
      "completions/max_terminated_length": 1224.0,
      "completions/mean_length": 369.51953125,
      "completions/mean_terminated_length": 370.9686584472656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 103.0,
      "epoch": 0.08,
      "grad_norm": 0.9460354447364807,
      "learning_rate": 3.4722222222222224e-06,
      "loss": -0.0031,
      "num_tokens": 16353039.0,
      "reward": 1.1918749809265137,
      "reward_std": 0.24501259624958038,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.455000102519989,
      "rewards/format_reward_step": 0.9921875,
      "step": 75
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.4513692827854427e-06,
      "aux_brier/mean_group_std": 0.051347752775991504,
      "aux_brier/mean_r": 0.9641691588949147,
      "aux_brier/n_active_tok": 246.375,
      "aux_brier/n_groups": 16.5,
      "aux_brier/n_step_records": 61.59375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.480296343001261,
      "calib/avg_num_step_conf": 7.69921875,
      "calib/ece": 0.45480158730158726,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.007307692307692317,
      "calib/mean_conf": 0.03376984126984127,
      "calib/mu_c": 0.029999999999999995,
      "calib/mu_w": 0.03730769230769231,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0022222222222222222,
      "calib/std_conf": 0.042403889964268485,
      "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": 2782.0,
      "completions/max_terminated_length": 2782.0,
      "completions/mean_length": 436.00390625,
      "completions/mean_terminated_length": 436.00390625,
      "completions/min_length": 132.0,
      "completions/min_terminated_length": 132.0,
      "epoch": 0.08106666666666666,
      "grad_norm": 0.18404695391654968,
      "learning_rate": 3.444444444444445e-06,
      "loss": 0.0844,
      "num_tokens": 16567712.0,
      "reward": 1.099198818206787,
      "reward_std": 0.2967877984046936,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.5296074151992798,
      "rewards/format_reward_step": 0.98046875,
      "step": 76
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.9130282626355086e-06,
      "aux_brier/mean_group_std": 0.05058887783171625,
      "aux_brier/mean_r": 0.9618940982608526,
      "aux_brier/n_active_tok": 218.125,
      "aux_brier/n_groups": 13.71875,
      "aux_brier/n_step_records": 54.53125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4209269752339721,
      "calib/avg_num_step_conf": 6.84375,
      "calib/ece": 0.4193650793650794,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004851340166804614,
      "calib/mean_conf": 0.029047619047619048,
      "calib/mu_c": 0.0263716814159292,
      "calib/mu_w": 0.031223021582733813,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.033044668456177394,
      "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": 2614.0,
      "completions/max_terminated_length": 2614.0,
      "completions/mean_length": 388.171875,
      "completions/mean_terminated_length": 389.69415283203125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 108.0,
      "epoch": 0.08213333333333334,
      "grad_norm": 0.07342797517776489,
      "learning_rate": 3.416666666666667e-06,
      "loss": 0.0574,
      "num_tokens": 16771748.0,
      "reward": 1.0746798515319824,
      "reward_std": 0.28718793392181396,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.5643445253372192,
      "rewards/format_reward_step": 0.984375,
      "step": 77
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.696898912166404e-06,
      "aux_brier/mean_group_std": 0.046484346640904464,
      "aux_brier/mean_r": 0.970809476392736,
      "aux_brier/n_active_tok": 233.5,
      "aux_brier/n_groups": 14.03125,
      "aux_brier/n_step_records": 58.375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5840215095064336,
      "calib/avg_num_step_conf": 7.375,
      "calib/ece": 0.479112,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0068533384546443955,
      "calib/mean_conf": 0.028887999999999997,
      "calib/mu_c": 0.03225984251968504,
      "calib/mu_w": 0.025406504065040646,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.0,
      "calib/std_conf": 0.024696952362589196,
      "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": 2616.0,
      "completions/max_terminated_length": 2616.0,
      "completions/mean_length": 468.4765625,
      "completions/mean_terminated_length": 468.4765625,
      "completions/min_length": 114.0,
      "completions/min_terminated_length": 114.0,
      "epoch": 0.0832,
      "grad_norm": 0.2748223543167114,
      "learning_rate": 3.3888888888888893e-06,
      "loss": 0.0472,
      "num_tokens": 16999702.0,
      "reward": 1.1009531021118164,
      "reward_std": 0.3101786971092224,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.497562050819397,
      "rewards/format_reward_step": 0.953125,
      "step": 78
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.572590723781687e-07,
      "aux_brier/mean_group_std": 0.04789018717446725,
      "aux_brier/mean_r": 0.9645207600269317,
      "aux_brier/n_active_tok": 249.0,
      "aux_brier/n_groups": 15.5,
      "aux_brier/n_step_records": 62.25,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5566919191919192,
      "calib/avg_num_step_conf": 7.85546875,
      "calib/ece": 0.49617857142857147,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00297045454545454,
      "calib/mean_conf": 0.027630952380952385,
      "calib/mu_c": 0.029045454545454544,
      "calib/mu_w": 0.026075000000000004,
      "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": 0.020211449545833277,
      "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": 2454.0,
      "completions/max_terminated_length": 2454.0,
      "completions/mean_length": 482.28125,
      "completions/mean_terminated_length": 482.28125,
      "completions/min_length": 104.0,
      "completions/min_terminated_length": 104.0,
      "epoch": 0.08426666666666667,
      "grad_norm": 0.2128184586763382,
      "learning_rate": 3.3611111111111117e-06,
      "loss": 0.0467,
      "num_tokens": 17229542.0,
      "reward": 1.132199764251709,
      "reward_std": 0.28818950057029724,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.49754947423934937,
      "rewards/format_reward_step": 0.984375,
      "step": 79
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.174135792287757e-06,
      "aux_brier/mean_group_std": 0.06495822288625715,
      "aux_brier/mean_r": 0.9570226080739604,
      "aux_brier/n_active_tok": 239.0,
      "aux_brier/n_groups": 15.15625,
      "aux_brier/n_step_records": 59.75,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5057889822595705,
      "calib/avg_num_step_conf": 7.47265625,
      "calib/ece": 0.4432283464566929,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00035107376283846636,
      "calib/mean_conf": 0.025275590551181105,
      "calib/mu_c": 0.02546218487394958,
      "calib/mu_w": 0.025111111111111115,
      "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": 0.019715566880187718,
      "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": 2224.0,
      "completions/max_terminated_length": 2224.0,
      "completions/mean_length": 393.703125,
      "completions/mean_terminated_length": 393.703125,
      "completions/min_length": 131.0,
      "completions/min_terminated_length": 131.0,
      "epoch": 0.08533333333333333,
      "grad_norm": 0.45626622438430786,
      "learning_rate": 3.3333333333333333e-06,
      "loss": 0.0332,
      "num_tokens": 17432490.0,
      "reward": 1.0984364748001099,
      "reward_std": 0.2826603353023529,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5499960780143738,
      "rewards/format_reward_step": 0.9921875,
      "step": 80
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 3.332123914540075e-06,
      "aux_brier/mean_group_std": 0.05346998538001507,
      "aux_brier/mean_r": 0.9647629380884053,
      "aux_brier/n_active_tok": 250.5,
      "aux_brier/n_groups": 15.625,
      "aux_brier/n_step_records": 62.625,
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5023799946366317,
      "calib/avg_num_step_conf": 8.0625,
      "calib/ece": 0.44011428571428574,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.004081632653061225,
      "calib/gap": -0.007569455618128187,
      "calib/mean_conf": 0.031069387755102044,
      "calib/mu_c": 0.026991150442477876,
      "calib/mu_w": 0.03456060606060606,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.004979591836734694,
      "calib/std_conf": 0.06577028112466142,
      "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": 2342.0,
      "completions/max_terminated_length": 2342.0,
      "completions/mean_length": 459.61328125,
      "completions/mean_terminated_length": 463.2322692871094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 147.0,
      "epoch": 0.0864,
      "grad_norm": 0.2887077331542969,
      "learning_rate": 3.3055555555555558e-06,
      "loss": 0.077,
      "num_tokens": 17656399.0,
      "reward": 1.0515084266662598,
      "reward_std": 0.2824578285217285,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.5341585874557495,
      "rewards/format_reward_step": 0.953125,
      "step": 81
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.6703866285469076e-06,
      "aux_brier/mean_group_std": 0.03935703469611457,
      "aux_brier/mean_r": 0.9688550529060806,
      "aux_brier/n_active_tok": 224.0,
      "aux_brier/n_groups": 14.5625,
      "aux_brier/n_step_records": 56.0,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5902911376170322,
      "calib/avg_num_step_conf": 7.01953125,
      "calib/ece": 0.5311908366533865,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004536885981787867,
      "calib/mean_conf": 0.019406772908366534,
      "calib/mu_c": 0.021449275362318842,
      "calib/mu_w": 0.016912389380530975,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.00039840637450199205,
      "calib/std_conf": 0.015356439464022069,
      "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": 2759.0,
      "completions/max_terminated_length": 2759.0,
      "completions/mean_length": 417.1796875,
      "completions/mean_terminated_length": 417.1796875,
      "completions/min_length": 80.0,
      "completions/min_terminated_length": 80.0,
      "epoch": 0.08746666666666666,
      "grad_norm": 0.08206015080213547,
      "learning_rate": 3.277777777777778e-06,
      "loss": 0.0486,
      "num_tokens": 17868749.0,
      "reward": 1.1433264017105103,
      "reward_std": 0.2854713201522827,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.46393078565597534,
      "rewards/format_reward_step": 0.9765625,
      "step": 82
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.3666570329196048e-05,
      "aux_brier/mean_group_std": 0.05717790892446513,
      "aux_brier/mean_r": 0.9562057854616453,
      "aux_brier/n_active_tok": 251.75,
      "aux_brier/n_groups": 14.78125,
      "aux_brier/n_step_records": 62.9375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.46120606875360093,
      "calib/avg_num_step_conf": 7.87109375,
      "calib/ece": 0.48732000000000003,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.002982523526022667,
      "calib/mean_conf": 0.02068,
      "calib/mu_c": 0.019212598425196844,
      "calib/mu_w": 0.02219512195121951,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.01408323826397892,
      "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": 2036.0,
      "completions/max_terminated_length": 2036.0,
      "completions/mean_length": 467.68359375,
      "completions/mean_terminated_length": 471.36614990234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 126.0,
      "epoch": 0.08853333333333334,
      "grad_norm": 0.28322744369506836,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.0183,
      "num_tokens": 18095740.0,
      "reward": 1.1100877523422241,
      "reward_std": 0.2294122874736786,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.4950386583805084,
      "rewards/format_reward_step": 0.97265625,
      "step": 83
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.3964810212652168e-06,
      "aux_brier/mean_group_std": 0.05856871356778768,
      "aux_brier/mean_r": 0.9582298212054214,
      "aux_brier/n_active_tok": 225.625,
      "aux_brier/n_groups": 12.8125,
      "aux_brier/n_step_records": 56.40625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5161845730027548,
      "calib/avg_num_step_conf": 7.0703125,
      "calib/ece": 0.4563083003952569,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 5.922865013774023e-05,
      "calib/mean_conf": 0.021952569169960474,
      "calib/mu_c": 0.021983471074380166,
      "calib/mu_w": 0.021924242424242426,
      "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": 0.01776787424143487,
      "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": 2346.0,
      "completions/max_terminated_length": 2346.0,
      "completions/mean_length": 410.5625,
      "completions/mean_terminated_length": 410.5625,
      "completions/min_length": 110.0,
      "completions/min_terminated_length": 110.0,
      "epoch": 0.0896,
      "grad_norm": 0.006935984827578068,
      "learning_rate": 3.2222222222222227e-06,
      "loss": 0.0606,
      "num_tokens": 18306764.0,
      "reward": 1.1007013320922852,
      "reward_std": 0.2433372586965561,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.5356179475784302,
      "rewards/format_reward_step": 0.98828125,
      "step": 84
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.41343224375046e-06,
      "aux_brier/mean_group_std": 0.0658378648923139,
      "aux_brier/mean_r": 0.956898890554232,
      "aux_brier/n_active_tok": 251.375,
      "aux_brier/n_groups": 16.5625,
      "aux_brier/n_step_records": 62.84375,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5936186974789915,
      "calib/avg_num_step_conf": 8.078125,
      "calib/ece": 0.4594254655870446,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.004048582995951417,
      "calib/gap": 0.004857058167016801,
      "calib/mean_conf": 0.028995587044534414,
      "calib/mu_c": 0.031512605042016806,
      "calib/mu_w": 0.026655546875000005,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.003319838056680162,
      "calib/std_conf": 0.08079508902204688,
      "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": 2880.0,
      "completions/max_terminated_length": 2880.0,
      "completions/mean_length": 468.23828125,
      "completions/mean_terminated_length": 471.9252014160156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 110.0,
      "epoch": 0.09066666666666667,
      "grad_norm": 0.010801228694617748,
      "learning_rate": 3.1944444444444443e-06,
      "loss": 0.0932,
      "num_tokens": 18534457.0,
      "reward": 1.071960210800171,
      "reward_std": 0.23767264187335968,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5144033432006836,
      "rewards/format_reward_step": 0.95703125,
      "step": 85
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 7.5741469434120745e-06,
      "aux_brier/mean_group_std": 0.059585150501863936,
      "aux_brier/mean_r": 0.9586120814670223,
      "aux_brier/n_active_tok": 232.875,
      "aux_brier/n_groups": 14.46875,
      "aux_brier/n_step_records": 58.21875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.512226347458712,
      "calib/avg_num_step_conf": 7.27734375,
      "calib/ece": 0.4018181818181818,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0010203559083344013,
      "calib/mean_conf": 0.021897233201581028,
      "calib/mu_c": 0.02130841121495327,
      "calib/mu_w": 0.02232876712328767,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0003952569169960474,
      "calib/std_conf": 0.015126518878198464,
      "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": 1749.0,
      "completions/max_terminated_length": 1749.0,
      "completions/mean_length": 438.4609375,
      "completions/mean_terminated_length": 440.180419921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.09173333333333333,
      "grad_norm": 0.1242508813738823,
      "learning_rate": 3.1666666666666667e-06,
      "loss": 0.0276,
      "num_tokens": 18752215.0,
      "reward": 1.0550609827041626,
      "reward_std": 0.28246423602104187,
      "rewards/accuracy_reward_step": 0.421875,
      "rewards/final_brier_reward_step": 0.5796191692352295,
      "rewards/format_reward_step": 0.9765625,
      "step": 86
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 7.154137164969154e-06,
      "aux_brier/mean_group_std": 0.07578767623087967,
      "aux_brier/mean_r": 0.9541173446474975,
      "aux_brier/n_active_tok": 227.125,
      "aux_brier/n_groups": 14.96875,
      "aux_brier/n_step_records": 56.78125,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.540683023872679,
      "calib/avg_num_step_conf": 7.453125,
      "calib/ece": 0.5616465863453816,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.001502652519893899,
      "calib/mean_conf": 0.02068273092369478,
      "calib/mu_c": 0.021310344827586206,
      "calib/mu_w": 0.019807692307692307,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.013530262932848557,
      "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": 0.015625,
      "completions/max_length": 2245.0,
      "completions/max_terminated_length": 2245.0,
      "completions/mean_length": 398.46484375,
      "completions/mean_terminated_length": 404.7897033691406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 88.0,
      "epoch": 0.0928,
      "grad_norm": 0.5288179516792297,
      "learning_rate": 3.138888888888889e-06,
      "loss": 0.0143,
      "num_tokens": 18959718.0,
      "reward": 1.1551870107650757,
      "reward_std": 0.286136269569397,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.42543554306030273,
      "rewards/format_reward_step": 0.96484375,
      "step": 87
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.120458337002585e-07,
      "aux_brier/mean_group_std": 0.055331605233758414,
      "aux_brier/mean_r": 0.9598075037022034,
      "aux_brier/n_active_tok": 240.25,
      "aux_brier/n_groups": 14.625,
      "aux_brier/n_step_records": 60.0625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5096358403203604,
      "calib/avg_num_step_conf": 7.80078125,
      "calib/ece": 0.45916600790513834,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0018713552746840272,
      "calib/mean_conf": 0.023047430830039527,
      "calib/mu_c": 0.024016393442622956,
      "calib/mu_w": 0.02214503816793893,
      "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": 0.0158490197238091,
      "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": 2910.0,
      "completions/max_terminated_length": 2910.0,
      "completions/mean_length": 448.9921875,
      "completions/mean_terminated_length": 450.7529602050781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 153.0,
      "epoch": 0.09386666666666667,
      "grad_norm": 0.1636967957019806,
      "learning_rate": 3.1111111111111116e-06,
      "loss": 0.0467,
      "num_tokens": 19184508.0,
      "reward": 1.1021703481674194,
      "reward_std": 0.20122531056404114,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.5336815118789673,
      "rewards/format_reward_step": 0.984375,
      "step": 88
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.4262653590678696e-06,
      "aux_brier/mean_group_std": 0.07005360666593853,
      "aux_brier/mean_r": 0.9535259534192914,
      "aux_brier/n_active_tok": 230.75,
      "aux_brier/n_groups": 13.125,
      "aux_brier/n_step_records": 57.6875,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5213982660025825,
      "calib/avg_num_step_conf": 7.23046875,
      "calib/ece": 0.4325390625,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0005417204697780245,
      "calib/mean_conf": 0.0244921875,
      "calib/mu_c": 0.02478632478632479,
      "calib/mu_w": 0.024244604316546764,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.01429788888139937,
      "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": 1462.0,
      "completions/max_terminated_length": 1462.0,
      "completions/mean_length": 445.78125,
      "completions/mean_terminated_length": 447.5294494628906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.09493333333333333,
      "grad_norm": 0.061887916177511215,
      "learning_rate": 3.0833333333333336e-06,
      "loss": 0.0064,
      "num_tokens": 19407516.0,
      "reward": 1.0940849781036377,
      "reward_std": 0.17070043087005615,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.5638397932052612,
      "rewards/format_reward_step": 0.9921875,
      "step": 89
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 4.512320092034394e-06,
      "aux_brier/mean_group_std": 0.10073570940345282,
      "aux_brier/mean_r": 0.9266705959774814,
      "aux_brier/n_active_tok": 250.75,
      "aux_brier/n_groups": 14.53125,
      "aux_brier/n_step_records": 62.6875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5735608345902463,
      "calib/avg_num_step_conf": 7.84375,
      "calib/ece": 0.5133201581027668,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004210658622423331,
      "calib/mean_conf": 0.02422924901185771,
      "calib/mu_c": 0.026176470588235294,
      "calib/mu_w": 0.021965811965811963,
      "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": 0.015085357080634389,
      "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": 2271.0,
      "completions/max_terminated_length": 2271.0,
      "completions/mean_length": 448.97265625,
      "completions/mean_terminated_length": 448.97265625,
      "completions/min_length": 100.0,
      "completions/min_terminated_length": 100.0,
      "epoch": 0.096,
      "grad_norm": 0.006909631192684174,
      "learning_rate": 3.055555555555556e-06,
      "loss": 0.0318,
      "num_tokens": 19625773.0,
      "reward": 1.1444084644317627,
      "reward_std": 0.22809094190597534,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.48388397693634033,
      "rewards/format_reward_step": 0.984375,
      "step": 90
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.808361265706829e-06,
      "aux_brier/mean_group_std": 0.05787001523135353,
      "aux_brier/mean_r": 0.966620922965862,
      "aux_brier/n_active_tok": 244.625,
      "aux_brier/n_groups": 13.375,
      "aux_brier/n_step_records": 61.15625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5891415283250436,
      "calib/avg_num_step_conf": 7.67578125,
      "calib/ece": 0.52256,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0033059879852722594,
      "calib/mean_conf": 0.025439999999999997,
      "calib/mu_c": 0.02693430656934306,
      "calib/mu_w": 0.0236283185840708,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.014338981832752283,
      "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": 2203.0,
      "completions/max_terminated_length": 2203.0,
      "completions/mean_length": 448.3203125,
      "completions/mean_terminated_length": 450.0784606933594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 160.0,
      "epoch": 0.09706666666666666,
      "grad_norm": 0.014485384337604046,
      "learning_rate": 3.0277777777777776e-06,
      "loss": 0.0269,
      "num_tokens": 19848255.0,
      "reward": 1.1486003398895264,
      "reward_std": 0.2162783443927765,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.4694015681743622,
      "rewards/format_reward_step": 0.9765625,
      "step": 91
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -8.149870861284825e-06,
      "aux_brier/mean_group_std": 0.027356117382574357,
      "aux_brier/mean_r": 0.9824328463670696,
      "aux_brier/n_active_tok": 229.0,
      "aux_brier/n_groups": 14.375,
      "aux_brier/n_step_records": 57.25,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5204615384615384,
      "calib/avg_num_step_conf": 7.33203125,
      "calib/ece": 0.46485098039215683,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004266153846153847,
      "calib/mean_conf": 0.025345098039215686,
      "calib/mu_c": 0.02752,
      "calib/mu_w": 0.023253846153846153,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.03323143039604012,
      "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": 1029.0,
      "completions/max_terminated_length": 1029.0,
      "completions/mean_length": 393.69140625,
      "completions/mean_terminated_length": 395.2353210449219,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.09813333333333334,
      "grad_norm": 0.025894714519381523,
      "learning_rate": 3e-06,
      "loss": -0.0019,
      "num_tokens": 20055760.0,
      "reward": 1.1168794631958008,
      "reward_std": 0.24395212531089783,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.5300179719924927,
      "rewards/format_reward_step": 0.9921875,
      "step": 92
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 3.4370384449644575e-06,
      "aux_brier/mean_group_std": 0.056055496449707885,
      "aux_brier/mean_r": 0.9675858516883075,
      "aux_brier/n_active_tok": 269.75,
      "aux_brier/n_groups": 18.03125,
      "aux_brier/n_step_records": 67.4375,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.47361735391831156,
      "calib/avg_num_step_conf": 8.45703125,
      "calib/ece": 0.4544233870967742,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0021065077193668236,
      "calib/mean_conf": 0.026221774193548392,
      "calib/mu_c": 0.02512605042016806,
      "calib/mu_w": 0.027232558139534884,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0004032258064516129,
      "calib/std_conf": 0.01307682054483867,
      "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": 2995.0,
      "completions/max_terminated_length": 2995.0,
      "completions/mean_length": 489.22265625,
      "completions/mean_terminated_length": 491.1412048339844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 132.0,
      "epoch": 0.0992,
      "grad_norm": 0.03542132303118706,
      "learning_rate": 2.9722222222222225e-06,
      "loss": 0.0628,
      "num_tokens": 20286777.0,
      "reward": 1.080827236175537,
      "reward_std": 0.26389071345329285,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5264338850975037,
      "rewards/format_reward_step": 0.96875,
      "step": 93
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.616237748026551e-06,
      "aux_brier/mean_group_std": 0.06548742352883224,
      "aux_brier/mean_r": 0.9555905112583005,
      "aux_brier/n_active_tok": 228.75,
      "aux_brier/n_groups": 14.5,
      "aux_brier/n_step_records": 57.1875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5670480120030008,
      "calib/avg_num_step_conf": 7.15625,
      "calib/ece": 0.48303952569169967,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0039129157289322335,
      "calib/mean_conf": 0.026841897233201583,
      "calib/mu_c": 0.02875968992248062,
      "calib/mu_w": 0.024846774193548388,
      "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": 0.014088606655990703,
      "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": 2454.0,
      "completions/max_terminated_length": 2454.0,
      "completions/mean_length": 426.56640625,
      "completions/mean_terminated_length": 426.56640625,
      "completions/min_length": 127.0,
      "completions/min_terminated_length": 127.0,
      "epoch": 0.10026666666666667,
      "grad_norm": 0.006841793190687895,
      "learning_rate": 2.944444444444445e-06,
      "loss": 0.073,
      "num_tokens": 20504658.0,
      "reward": 1.12615966796875,
      "reward_std": 0.19835515320301056,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.512451171875,
      "rewards/format_reward_step": 0.98828125,
      "step": 94
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.7462224184139341e-06,
      "aux_brier/mean_group_std": 0.07639294718765116,
      "aux_brier/mean_r": 0.9508508828057519,
      "aux_brier/n_active_tok": 246.75,
      "aux_brier/n_groups": 14.5,
      "aux_brier/n_step_records": 61.6875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4827979409656611,
      "calib/avg_num_step_conf": 7.75,
      "calib/ece": 0.5620238095238095,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0010835994005343116,
      "calib/mean_conf": 0.03091269841269841,
      "calib/mu_c": 0.030469798657718115,
      "calib/mu_w": 0.03155339805825243,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0008333333333333333,
      "calib/std_conf": 0.019668894079380023,
      "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": 2502.0,
      "completions/max_terminated_length": 2502.0,
      "completions/mean_length": 452.68359375,
      "completions/mean_terminated_length": 454.4588623046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 88.0,
      "epoch": 0.10133333333333333,
      "grad_norm": 0.009349067695438862,
      "learning_rate": 2.916666666666667e-06,
      "loss": 0.0141,
      "num_tokens": 20726673.0,
      "reward": 1.1833415031433105,
      "reward_std": 0.22875583171844482,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.4364910125732422,
      "rewards/format_reward_step": 0.984375,
      "step": 95
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 7.968594871093915e-06,
      "aux_brier/mean_group_std": 0.06872664148503413,
      "aux_brier/mean_r": 0.9523272980180933,
      "aux_brier/n_active_tok": 224.125,
      "aux_brier/n_groups": 12.125,
      "aux_brier/n_step_records": 56.03125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5447864945382324,
      "calib/avg_num_step_conf": 7.0546875,
      "calib/ece": 0.5976594488188977,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0019566037735849057,
      "calib/mean_conf": 0.0283248031496063,
      "calib/mu_c": 0.0290566037735849,
      "calib/mu_w": 0.027099999999999996,
      "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": 0.01253930101094713,
      "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": 1992.0,
      "completions/max_terminated_length": 1992.0,
      "completions/mean_length": 412.9375,
      "completions/mean_terminated_length": 412.9375,
      "completions/min_length": 140.0,
      "completions/min_terminated_length": 140.0,
      "epoch": 0.1024,
      "grad_norm": 0.009241505526006222,
      "learning_rate": 2.888888888888889e-06,
      "loss": -0.0,
      "num_tokens": 20938201.0,
      "reward": 1.2167165279388428,
      "reward_std": 0.17452289164066315,
      "rewards/accuracy_reward_step": 0.62109375,
      "rewards/final_brier_reward_step": 0.40592920780181885,
      "rewards/format_reward_step": 0.98828125,
      "step": 96
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.801826086642258e-06,
      "aux_brier/mean_group_std": 0.04182444767990009,
      "aux_brier/mean_r": 0.9757240453047343,
      "aux_brier/n_active_tok": 249.125,
      "aux_brier/n_groups": 15.90625,
      "aux_brier/n_step_records": 62.28125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6404476670870114,
      "calib/avg_num_step_conf": 7.78515625,
      "calib/ece": 0.48678571428571427,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005648171500630513,
      "calib/mean_conf": 0.029880952380952383,
      "calib/mu_c": 0.03261538461538461,
      "calib/mu_w": 0.026967213114754095,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0003968253968253968,
      "calib/std_conf": 0.013612615030987228,
      "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": 2387.0,
      "completions/max_terminated_length": 2387.0,
      "completions/mean_length": 429.21484375,
      "completions/mean_terminated_length": 430.8980712890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.10346666666666667,
      "grad_norm": 0.14593623578548431,
      "learning_rate": 2.861111111111111e-06,
      "loss": 0.0241,
      "num_tokens": 21153152.0,
      "reward": 1.1242269277572632,
      "reward_std": 0.2638818025588989,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5047203302383423,
      "rewards/format_reward_step": 0.98046875,
      "step": 97
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 4.996638026888434e-06,
      "aux_brier/mean_group_std": 0.08093589092060835,
      "aux_brier/mean_r": 0.9443325923255121,
      "aux_brier/n_active_tok": 254.25,
      "aux_brier/n_groups": 16.375,
      "aux_brier/n_step_records": 63.5625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5582205029013539,
      "calib/avg_num_step_conf": 8.14453125,
      "calib/ece": 0.5277689243027888,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0025609284332688553,
      "calib/mean_conf": 0.03398406374501992,
      "calib/mu_c": 0.0351063829787234,
      "calib/mu_w": 0.032545454545454544,
      "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": 0.014059624917069644,
      "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": 2711.0,
      "completions/max_terminated_length": 2711.0,
      "completions/mean_length": 456.859375,
      "completions/mean_terminated_length": 458.6510009765625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 122.0,
      "epoch": 0.10453333333333334,
      "grad_norm": 0.03403199464082718,
      "learning_rate": 2.8333333333333335e-06,
      "loss": 0.0727,
      "num_tokens": 21376292.0,
      "reward": 1.1577739715576172,
      "reward_std": 0.2801635265350342,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.4670332074165344,
      "rewards/format_reward_step": 0.98046875,
      "step": 98
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.6811854322007491e-06,
      "aux_brier/mean_group_std": 0.03440397797157451,
      "aux_brier/mean_r": 0.9778653646278308,
      "aux_brier/n_active_tok": 267.25,
      "aux_brier/n_groups": 17.375,
      "aux_brier/n_step_records": 66.8125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4395824239778845,
      "calib/avg_num_step_conf": 8.37109375,
      "calib/ece": 0.27311857707509885,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0022383966244725675,
      "calib/mean_conf": 0.039134387351778654,
      "calib/mu_c": 0.037594936708860764,
      "calib/mu_w": 0.03983333333333333,
      "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": 0.017172851866403924,
      "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": 2594.0,
      "completions/max_terminated_length": 2594.0,
      "completions/mean_length": 506.23046875,
      "completions/mean_terminated_length": 506.23046875,
      "completions/min_length": 128.0,
      "completions/min_terminated_length": 128.0,
      "epoch": 0.1056,
      "grad_norm": 0.009494793601334095,
      "learning_rate": 2.805555555555556e-06,
      "loss": 0.0316,
      "num_tokens": 21611687.0,
      "reward": 0.9780057668685913,
      "reward_std": 0.2294953614473343,
      "rewards/accuracy_reward_step": 0.30859375,
      "rewards/final_brier_reward_step": 0.7010855674743652,
      "rewards/format_reward_step": 0.98828125,
      "step": 99
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 4.011528835834444e-06,
      "aux_brier/mean_group_std": 0.08616607958938509,
      "aux_brier/mean_r": 0.9386765459259507,
      "aux_brier/n_active_tok": 257.5,
      "aux_brier/n_groups": 15.78125,
      "aux_brier/n_step_records": 64.375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5979155944415853,
      "calib/avg_num_step_conf": 8.171875,
      "calib/ece": 0.42842,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005672928461142562,
      "calib/mean_conf": 0.03558,
      "calib/mu_c": 0.03862068965517241,
      "calib/mu_w": 0.03294776119402985,
      "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": 0.014716100026841351,
      "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": 2438.0,
      "completions/max_terminated_length": 2438.0,
      "completions/mean_length": 461.5,
      "completions/mean_terminated_length": 463.3098449707031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 110.0,
      "epoch": 0.10666666666666667,
      "grad_norm": 0.05984112247824669,
      "learning_rate": 2.7777777777777783e-06,
      "loss": 0.0861,
      "num_tokens": 21837239.0,
      "reward": 1.0806536674499512,
      "reward_std": 0.26517045497894287,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.5569897294044495,
      "rewards/format_reward_step": 0.9765625,
      "step": 100
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.2493385881939822e-06,
      "aux_brier/mean_group_std": 0.041907449987532125,
      "aux_brier/mean_r": 0.9751555650940253,
      "aux_brier/n_active_tok": 276.5,
      "aux_brier/n_groups": 16.40625,
      "aux_brier/n_step_records": 69.125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5376787495843033,
      "calib/avg_num_step_conf": 8.640625,
      "calib/ece": 0.3486904761904763,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002274027269704028,
      "calib/mean_conf": 0.036230158730158724,
      "calib/mu_c": 0.03762886597938145,
      "calib/mu_w": 0.03535483870967742,
      "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": 0.016513420563437593,
      "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": 2925.0,
      "completions/max_terminated_length": 2925.0,
      "completions/mean_length": 490.14453125,
      "completions/mean_terminated_length": 490.14453125,
      "completions/min_length": 112.0,
      "completions/min_terminated_length": 112.0,
      "epoch": 0.10773333333333333,
      "grad_norm": 0.011581323109567165,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.0924,
      "num_tokens": 22069708.0,
      "reward": 1.0301841497421265,
      "reward_std": 0.23364655673503876,
      "rewards/accuracy_reward_step": 0.3828125,
      "rewards/final_brier_reward_step": 0.6285492181777954,
      "rewards/format_reward_step": 0.98046875,
      "step": 101
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -7.214122758822383e-07,
      "aux_brier/mean_group_std": 0.06669447807464005,
      "aux_brier/mean_r": 0.9520474488679199,
      "aux_brier/n_active_tok": 234.375,
      "aux_brier/n_groups": 13.59375,
      "aux_brier/n_step_records": 58.59375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5959164292497625,
      "calib/avg_num_step_conf": 7.38671875,
      "calib/ece": 0.6086205533596838,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0045531135531135455,
      "calib/mean_conf": 0.03169565217391304,
      "calib/mu_c": 0.03333333333333333,
      "calib/mu_w": 0.028780219780219787,
      "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": 0.01405150959177948,
      "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": 2497.0,
      "completions/max_terminated_length": 2497.0,
      "completions/mean_length": 397.5703125,
      "completions/mean_terminated_length": 399.1294250488281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 72.0,
      "epoch": 0.1088,
      "grad_norm": 0.00953032448887825,
      "learning_rate": 2.7222222222222224e-06,
      "loss": -0.0035,
      "num_tokens": 22278182.0,
      "reward": 1.2299764156341553,
      "reward_std": 0.15418517589569092,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.39646828174591064,
      "rewards/format_reward_step": 0.98828125,
      "step": 102
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.623730196635535e-06,
      "aux_brier/mean_group_std": 0.06268301511030247,
      "aux_brier/mean_r": 0.962447104480166,
      "aux_brier/n_active_tok": 266.0,
      "aux_brier/n_groups": 17.375,
      "aux_brier/n_step_records": 66.5,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5176478083588175,
      "calib/avg_num_step_conf": 8.3359375,
      "calib/ece": 0.5344545454545454,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0007089067278287509,
      "calib/mean_conf": 0.034715415019762844,
      "calib/mu_c": 0.035020833333333334,
      "calib/mu_w": 0.034311926605504584,
      "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": 0.014682342636720714,
      "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": 2412.0,
      "completions/max_terminated_length": 2412.0,
      "completions/mean_length": 501.39453125,
      "completions/mean_terminated_length": 503.3608093261719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 124.0,
      "epoch": 0.10986666666666667,
      "grad_norm": 0.06882353872060776,
      "learning_rate": 2.6944444444444444e-06,
      "loss": -0.0031,
      "num_tokens": 22511091.0,
      "reward": 1.1725845336914062,
      "reward_std": 0.20618131756782532,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.4637756049633026,
      "rewards/format_reward_step": 0.98828125,
      "step": 103
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.1690783303652807e-05,
      "aux_brier/mean_group_std": 0.05171970853851588,
      "aux_brier/mean_r": 0.9686342693300826,
      "aux_brier/n_active_tok": 247.25,
      "aux_brier/n_groups": 14.625,
      "aux_brier/n_step_records": 61.8125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5647796352583586,
      "calib/avg_num_step_conf": 7.95703125,
      "calib/ece": 0.4087747035573123,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0021105623100303936,
      "calib/mean_conf": 0.03391304347826087,
      "calib/mu_c": 0.03508928571428572,
      "calib/mu_w": 0.032978723404255325,
      "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": 0.014367575340746352,
      "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": 2381.0,
      "completions/max_terminated_length": 2381.0,
      "completions/mean_length": 449.234375,
      "completions/mean_terminated_length": 452.7716369628906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 159.0,
      "epoch": 0.11093333333333333,
      "grad_norm": 0.07084021717309952,
      "learning_rate": 2.666666666666667e-06,
      "loss": 0.0049,
      "num_tokens": 22732775.0,
      "reward": 1.0746469497680664,
      "reward_std": 0.2366853654384613,
      "rewards/accuracy_reward_step": 0.4375,
      "rewards/final_brier_reward_step": 0.5798375010490417,
      "rewards/format_reward_step": 0.984375,
      "step": 104
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 4.3962614099113395e-07,
      "aux_brier/mean_group_std": 0.03452205577012924,
      "aux_brier/mean_r": 0.97974184310751,
      "aux_brier/n_active_tok": 255.875,
      "aux_brier/n_groups": 17.125,
      "aux_brier/n_step_records": 63.96875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.42881355932203397,
      "calib/avg_num_step_conf": 8.06640625,
      "calib/ece": 0.432298814229249,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004502272441933471,
      "calib/mean_conf": 0.03655494071146245,
      "calib/mu_c": 0.03415254237288135,
      "calib/mu_w": 0.03865481481481482,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0012252964426877472,
      "calib/std_conf": 0.015366233491518097,
      "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": 2626.0,
      "completions/max_terminated_length": 2626.0,
      "completions/mean_length": 461.1796875,
      "completions/mean_terminated_length": 461.1796875,
      "completions/min_length": 172.0,
      "completions/min_terminated_length": 172.0,
      "epoch": 0.112,
      "grad_norm": 0.009009039960801601,
      "learning_rate": 2.6388888888888893e-06,
      "loss": 0.0727,
      "num_tokens": 22956597.0,
      "reward": 1.0923669338226318,
      "reward_std": 0.24024072289466858,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.5569679141044617,
      "rewards/format_reward_step": 0.984375,
      "step": 105
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -8.847564145791909e-06,
      "aux_brier/mean_group_std": 0.06365965600698384,
      "aux_brier/mean_r": 0.958427444866737,
      "aux_brier/n_active_tok": 241.75,
      "aux_brier/n_groups": 13.59375,
      "aux_brier/n_step_records": 60.4375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5625705329153605,
      "calib/avg_num_step_conf": 7.5703125,
      "calib/ece": 0.3918156862745098,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003337304075235116,
      "calib/mean_conf": 0.03955686274509804,
      "calib/mu_c": 0.04145454545454545,
      "calib/mu_w": 0.038117241379310336,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.013752279738311352,
      "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": 2198.0,
      "completions/max_terminated_length": 2198.0,
      "completions/mean_length": 438.31640625,
      "completions/mean_terminated_length": 438.31640625,
      "completions/min_length": 163.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.11306666666666666,
      "grad_norm": 0.007643554359674454,
      "learning_rate": 2.6111111111111113e-06,
      "loss": 0.0252,
      "num_tokens": 23173390.0,
      "reward": 1.0748772621154785,
      "reward_std": 0.22851277887821198,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.5963842272758484,
      "rewards/format_reward_step": 0.9921875,
      "step": 106
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 4.34117809899881e-06,
      "aux_brier/mean_group_std": 0.05989411908267228,
      "aux_brier/mean_r": 0.9628471575134833,
      "aux_brier/n_active_tok": 251.375,
      "aux_brier/n_groups": 14.5625,
      "aux_brier/n_step_records": 62.84375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6133786141575275,
      "calib/avg_num_step_conf": 7.921875,
      "calib/ece": 0.495984251968504,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005618145563310056,
      "calib/mean_conf": 0.039448818897637794,
      "calib/mu_c": 0.04205882352941176,
      "calib/mu_w": 0.0364406779661017,
      "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": 0.014644720992074445,
      "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": 2444.0,
      "completions/max_terminated_length": 2444.0,
      "completions/mean_length": 449.125,
      "completions/mean_terminated_length": 449.125,
      "completions/min_length": 153.0,
      "completions/min_terminated_length": 153.0,
      "epoch": 0.11413333333333334,
      "grad_norm": 0.0080488296225667,
      "learning_rate": 2.5833333333333337e-06,
      "loss": 0.0252,
      "num_tokens": 23392982.0,
      "reward": 1.1503819227218628,
      "reward_std": 0.25812166929244995,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.4999653995037079,
      "rewards/format_reward_step": 0.98828125,
      "step": 107
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.11153050233459e-06,
      "aux_brier/mean_group_std": 0.0811696201758042,
      "aux_brier/mean_r": 0.9475855806344488,
      "aux_brier/n_active_tok": 267.75,
      "aux_brier/n_groups": 15.21875,
      "aux_brier/n_step_records": 66.9375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6236570247933885,
      "calib/avg_num_step_conf": 8.39453125,
      "calib/ece": 0.6093517786561264,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.007393939393939397,
      "calib/mean_conf": 0.04282213438735178,
      "calib/mu_c": 0.045393939393939396,
      "calib/mu_w": 0.038,
      "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": 0.017201548331752234,
      "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": 1900.0,
      "completions/max_terminated_length": 1900.0,
      "completions/mean_length": 477.796875,
      "completions/mean_terminated_length": 477.796875,
      "completions/min_length": 111.0,
      "completions/min_terminated_length": 111.0,
      "epoch": 0.1152,
      "grad_norm": 0.0073010846972465515,
      "learning_rate": 2.5555555555555557e-06,
      "loss": 0.0032,
      "num_tokens": 23618530.0,
      "reward": 1.2426183223724365,
      "reward_std": 0.2211112380027771,
      "rewards/accuracy_reward_step": 0.6484375,
      "rewards/final_brier_reward_step": 0.4001609683036804,
      "rewards/format_reward_step": 0.98828125,
      "step": 108
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.1203846021510273e-06,
      "aux_brier/mean_group_std": 0.04565737504777914,
      "aux_brier/mean_r": 0.9744196508989337,
      "aux_brier/n_active_tok": 271.375,
      "aux_brier/n_groups": 15.3125,
      "aux_brier/n_step_records": 67.84375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5828637615827699,
      "calib/avg_num_step_conf": 8.79296875,
      "calib/ece": 0.4364268774703557,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0036446280991735552,
      "calib/mean_conf": 0.04183399209486167,
      "calib/mu_c": 0.04373553719008264,
      "calib/mu_w": 0.04009090909090909,
      "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": 0.016523671804807218,
      "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": 2665.0,
      "completions/max_terminated_length": 2665.0,
      "completions/mean_length": 469.89453125,
      "completions/mean_terminated_length": 473.594482421875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 136.0,
      "epoch": 0.11626666666666667,
      "grad_norm": 0.2555139362812042,
      "learning_rate": 2.5277777777777778e-06,
      "loss": -0.0407,
      "num_tokens": 23843423.0,
      "reward": 1.1055392026901245,
      "reward_std": 0.1596943885087967,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.554969310760498,
      "rewards/format_reward_step": 0.98828125,
      "step": 109
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.658392951106816e-06,
      "aux_brier/mean_group_std": 0.053848758224179144,
      "aux_brier/mean_r": 0.9650792494076447,
      "aux_brier/n_active_tok": 223.625,
      "aux_brier/n_groups": 12.75,
      "aux_brier/n_step_records": 55.90625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.43342462180490354,
      "calib/avg_num_step_conf": 7.4140625,
      "calib/ece": 0.394044,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004108763693270727,
      "calib/mean_conf": 0.039555999999999994,
      "calib/mu_c": 0.037222222222222226,
      "calib/mu_w": 0.04133098591549295,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0008,
      "calib/std_conf": 0.013373363974707337,
      "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": 3053.0,
      "completions/max_terminated_length": 3053.0,
      "completions/mean_length": 449.1015625,
      "completions/mean_terminated_length": 452.6377868652344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 114.0,
      "epoch": 0.11733333333333333,
      "grad_norm": 0.16124384105205536,
      "learning_rate": 2.5e-06,
      "loss": 0.0469,
      "num_tokens": 24063313.0,
      "reward": 1.051352620124817,
      "reward_std": 0.29947662353515625,
      "rewards/accuracy_reward_step": 0.421875,
      "rewards/final_brier_reward_step": 0.5804105997085571,
      "rewards/format_reward_step": 0.96875,
      "step": 110
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.069663242178834e-06,
      "aux_brier/mean_group_std": 0.05618775918637486,
      "aux_brier/mean_r": 0.9636342183906583,
      "aux_brier/n_active_tok": 264.5,
      "aux_brier/n_groups": 17.9375,
      "aux_brier/n_step_records": 66.125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.560864972348226,
      "calib/avg_num_step_conf": 8.29296875,
      "calib/ece": 0.4334409448818898,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.003937007874015748,
      "calib/gap": 0.010332069844031573,
      "calib/mean_conf": 0.04293700787401575,
      "calib/mu_c": 0.048347107438016526,
      "calib/mu_w": 0.03801503759398495,
      "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": 0.06149799920328963,
      "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": 2971.0,
      "completions/max_terminated_length": 2971.0,
      "completions/mean_length": 486.4765625,
      "completions/mean_terminated_length": 486.4765625,
      "completions/min_length": 105.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.1184,
      "grad_norm": 0.008153988048434258,
      "learning_rate": 2.4722222222222226e-06,
      "loss": 0.018,
      "num_tokens": 24295259.0,
      "reward": 1.102805495262146,
      "reward_std": 0.2584152817726135,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.5518468618392944,
      "rewards/format_reward_step": 0.984375,
      "step": 111
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.3048599009655177e-06,
      "aux_brier/mean_group_std": 0.10069022026117411,
      "aux_brier/mean_r": 0.9255774714505626,
      "aux_brier/n_active_tok": 234.625,
      "aux_brier/n_groups": 14.0625,
      "aux_brier/n_step_records": 58.65625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6405714285714286,
      "calib/avg_num_step_conf": 7.46484375,
      "calib/ece": 0.4636565254098406,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005920413667698403,
      "calib/mean_conf": 0.03921200048657371,
      "calib/mu_c": 0.042160413667698404,
      "calib/mu_w": 0.03624,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.00043824701195219125,
      "calib/std_conf": 0.014251539851051709,
      "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": 2529.0,
      "completions/max_terminated_length": 2529.0,
      "completions/mean_length": 516.875,
      "completions/mean_terminated_length": 516.875,
      "completions/min_length": 176.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.11946666666666667,
      "grad_norm": 0.008149411529302597,
      "learning_rate": 2.4444444444444447e-06,
      "loss": 0.0401,
      "num_tokens": 24535499.0,
      "reward": 1.1104958057403564,
      "reward_std": 0.2235538810491562,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.5279207229614258,
      "rewards/format_reward_step": 0.97265625,
      "step": 112
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.7379988755950038e-06,
      "aux_brier/mean_group_std": 0.07673422997697633,
      "aux_brier/mean_r": 0.9488373491836689,
      "aux_brier/n_active_tok": 235.0,
      "aux_brier/n_groups": 14.34375,
      "aux_brier/n_step_records": 58.75,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5268595041322314,
      "calib/avg_num_step_conf": 7.34375,
      "calib/ece": 0.49160784313725486,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.007038361909460972,
      "calib/mean_conf": 0.04423529411764706,
      "calib/mu_c": 0.040895522388059706,
      "calib/mu_w": 0.04793388429752068,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0051764705882352945,
      "calib/std_conf": 0.047404222627870515,
      "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": 2072.0,
      "completions/max_terminated_length": 2072.0,
      "completions/mean_length": 435.609375,
      "completions/mean_terminated_length": 437.31768798828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 140.0,
      "epoch": 0.12053333333333334,
      "grad_norm": 0.009362783282995224,
      "learning_rate": 2.4166666666666667e-06,
      "loss": 0.0282,
      "num_tokens": 24752215.0,
      "reward": 1.1440508365631104,
      "reward_std": 0.27415621280670166,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.505890965461731,
      "rewards/format_reward_step": 0.98828125,
      "step": 113
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.150451800051137e-06,
      "aux_brier/mean_group_std": 0.05664586972073767,
      "aux_brier/mean_r": 0.9660372742649933,
      "aux_brier/n_active_tok": 231.875,
      "aux_brier/n_groups": 13.3125,
      "aux_brier/n_step_records": 57.96875,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5644085753458806,
      "calib/avg_num_step_conf": 7.390625,
      "calib/ece": 0.5190745098039216,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0025438115418172716,
      "calib/mean_conf": 0.03778823529411765,
      "calib/mu_c": 0.038915492957746475,
      "calib/mu_w": 0.036371681415929204,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.01065843396249524,
      "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": 1483.0,
      "completions/max_terminated_length": 1483.0,
      "completions/mean_length": 420.94921875,
      "completions/mean_terminated_length": 422.60003662109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 140.0,
      "epoch": 0.1216,
      "grad_norm": 0.17683005332946777,
      "learning_rate": 2.388888888888889e-06,
      "loss": 0.0022,
      "num_tokens": 24965002.0,
      "reward": 1.1705669164657593,
      "reward_std": 0.23685584962368011,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.47914260625839233,
      "rewards/format_reward_step": 0.9921875,
      "step": 114
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -7.781708822873767e-07,
      "aux_brier/mean_group_std": 0.05817280359278826,
      "aux_brier/mean_r": 0.9644236579121487,
      "aux_brier/n_active_tok": 226.125,
      "aux_brier/n_groups": 11.75,
      "aux_brier/n_step_records": 56.53125,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5555972013993004,
      "calib/avg_num_step_conf": 7.11328125,
      "calib/ece": 0.5100669291338583,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.003937007874015748,
      "calib/gap": -0.006791229385307358,
      "calib/mean_conf": 0.04331889763779528,
      "calib/mu_c": 0.04021739130434782,
      "calib/mu_w": 0.04700862068965518,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.005039370078740157,
      "calib/std_conf": 0.06247176273279215,
      "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": 0.00390625,
      "completions/max_length": 948.0,
      "completions/max_terminated_length": 948.0,
      "completions/mean_length": 418.890625,
      "completions/mean_terminated_length": 420.5333557128906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 126.0,
      "epoch": 0.12266666666666666,
      "grad_norm": 0.3346615731716156,
      "learning_rate": 2.361111111111111e-06,
      "loss": 0.007,
      "num_tokens": 25177502.0,
      "reward": 1.1499089002609253,
      "reward_std": 0.2871477007865906,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.48244792222976685,
      "rewards/format_reward_step": 0.97265625,
      "step": 115
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.519956423893913e-07,
      "aux_brier/mean_group_std": 0.03897763365044267,
      "aux_brier/mean_r": 0.9761643599604232,
      "aux_brier/n_active_tok": 254.5,
      "aux_brier/n_groups": 16.4375,
      "aux_brier/n_step_records": 63.625,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5380801140782991,
      "calib/avg_num_step_conf": 8.26171875,
      "calib/ece": 0.4993574297188755,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00061317085817994,
      "calib/mean_conf": 0.03670682730923695,
      "calib/mu_c": 0.03699248120300752,
      "calib/mu_w": 0.03637931034482758,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0009638554216867468,
      "calib/std_conf": 0.017596503668048828,
      "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": 2504.0,
      "completions/max_terminated_length": 2504.0,
      "completions/mean_length": 499.73046875,
      "completions/mean_terminated_length": 503.66534423828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 103.0,
      "epoch": 0.12373333333333333,
      "grad_norm": 0.0661558285355568,
      "learning_rate": 2.3333333333333336e-06,
      "loss": 0.0408,
      "num_tokens": 25409953.0,
      "reward": 1.1254181861877441,
      "reward_std": 0.27198296785354614,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.48604804277420044,
      "rewards/format_reward_step": 0.96875,
      "step": 116
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 9.08533166477099e-06,
      "aux_brier/mean_group_std": 0.047897951509321336,
      "aux_brier/mean_r": 0.9690990044847806,
      "aux_brier/n_active_tok": 259.5,
      "aux_brier/n_groups": 16.8125,
      "aux_brier/n_step_records": 64.875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5697837150127226,
      "calib/avg_num_step_conf": 8.11328125,
      "calib/ece": 0.4408764940239044,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002727099236641238,
      "calib/mean_conf": 0.03832669322709164,
      "calib/mu_c": 0.039750000000000014,
      "calib/mu_w": 0.037022900763358776,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0005577689243027889,
      "calib/std_conf": 0.028333332555256656,
      "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": 2331.0,
      "completions/max_terminated_length": 2331.0,
      "completions/mean_length": 491.10546875,
      "completions/mean_terminated_length": 491.10546875,
      "completions/min_length": 174.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.1248,
      "grad_norm": 0.007189909461885691,
      "learning_rate": 2.305555555555556e-06,
      "loss": 0.0419,
      "num_tokens": 25642276.0,
      "reward": 1.0869762897491455,
      "reward_std": 0.2557705044746399,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5354054570198059,
      "rewards/format_reward_step": 0.96875,
      "step": 117
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.610709073314066e-06,
      "aux_brier/mean_group_std": 0.046083838386966454,
      "aux_brier/mean_r": 0.9714296156016924,
      "aux_brier/n_active_tok": 237.125,
      "aux_brier/n_groups": 14.28125,
      "aux_brier/n_step_records": 59.28125,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5148148148148148,
      "calib/avg_num_step_conf": 7.671875,
      "calib/ece": 0.5082208835341365,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00016257309941521136,
      "calib/mean_conf": 0.03394779116465863,
      "calib/mu_c": 0.034022222222222225,
      "calib/mu_w": 0.03385964912280701,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0,
      "calib/std_conf": 0.010271088851985645,
      "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": 1521.0,
      "completions/max_terminated_length": 1521.0,
      "completions/mean_length": 442.80078125,
      "completions/mean_terminated_length": 449.8293762207031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.12586666666666665,
      "grad_norm": 0.0722954198718071,
      "learning_rate": 2.277777777777778e-06,
      "loss": -0.0293,
      "num_tokens": 25859641.0,
      "reward": 1.133664846420288,
      "reward_std": 0.23743316531181335,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.4799717664718628,
      "rewards/format_reward_step": 0.97265625,
      "step": 118
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.457912561966812e-06,
      "aux_brier/mean_group_std": 0.06911768021232345,
      "aux_brier/mean_r": 0.9609109672958739,
      "aux_brier/n_active_tok": 262.875,
      "aux_brier/n_groups": 17.40625,
      "aux_brier/n_step_records": 65.71875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5227099236641222,
      "calib/avg_num_step_conf": 8.21875,
      "calib/ece": 0.44505976095617533,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0010642493638676892,
      "calib/mean_conf": 0.033027888446215146,
      "calib/mu_c": 0.03358333333333333,
      "calib/mu_w": 0.032519083969465644,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.010197024169621382,
      "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": 2926.0,
      "completions/max_terminated_length": 2926.0,
      "completions/mean_length": 522.71875,
      "completions/mean_terminated_length": 522.71875,
      "completions/min_length": 102.0,
      "completions/min_terminated_length": 102.0,
      "epoch": 0.12693333333333334,
      "grad_norm": 0.006843086332082748,
      "learning_rate": 2.25e-06,
      "loss": 0.0582,
      "num_tokens": 26098521.0,
      "reward": 1.0915626287460327,
      "reward_std": 0.23342382907867432,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.53812575340271,
      "rewards/format_reward_step": 0.9765625,
      "step": 119
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.5438382932053685e-06,
      "aux_brier/mean_group_std": 0.07403760023497923,
      "aux_brier/mean_r": 0.9592139927628702,
      "aux_brier/n_active_tok": 218.875,
      "aux_brier/n_groups": 11.90625,
      "aux_brier/n_step_records": 54.71875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5660766400203019,
      "calib/avg_num_step_conf": 6.8515625,
      "calib/ece": 0.529300395256917,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.001622002283974116,
      "calib/mean_conf": 0.031964426877470364,
      "calib/mu_c": 0.03267605633802817,
      "calib/mu_w": 0.031054054054054054,
      "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": 0.011632155215101388,
      "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": 2531.0,
      "completions/max_terminated_length": 2531.0,
      "completions/mean_length": 430.33203125,
      "completions/mean_terminated_length": 430.33203125,
      "completions/min_length": 170.0,
      "completions/min_terminated_length": 170.0,
      "epoch": 0.128,
      "grad_norm": 0.008278531953692436,
      "learning_rate": 2.222222222222222e-06,
      "loss": 0.0438,
      "num_tokens": 26315374.0,
      "reward": 1.1601471900939941,
      "reward_std": 0.26189303398132324,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.46090108156204224,
      "rewards/format_reward_step": 0.98046875,
      "step": 120
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.3198501480626845e-06,
      "aux_brier/mean_group_std": 0.05460861669709895,
      "aux_brier/mean_r": 0.9674395889680963,
      "aux_brier/n_active_tok": 249.625,
      "aux_brier/n_groups": 14.125,
      "aux_brier/n_step_records": 62.40625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5341512775107513,
      "calib/avg_num_step_conf": 8.28125,
      "calib/ece": 0.4364126984126984,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0021791044776119373,
      "calib/mean_conf": 0.031841269841269844,
      "calib/mu_c": 0.032999999999999995,
      "calib/mu_w": 0.030820895522388057,
      "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": 0.011353212746061988,
      "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": 2538.0,
      "completions/max_terminated_length": 2538.0,
      "completions/mean_length": 478.0234375,
      "completions/mean_terminated_length": 483.69171142578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 166.0,
      "epoch": 0.12906666666666666,
      "grad_norm": 0.0893382579088211,
      "learning_rate": 2.1944444444444445e-06,
      "loss": -0.0206,
      "num_tokens": 26542804.0,
      "reward": 1.09130859375,
      "reward_std": 0.27117180824279785,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.5527344942092896,
      "rewards/format_reward_step": 0.984375,
      "step": 121
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.561933296034823e-07,
      "aux_brier/mean_group_std": 0.06950377476941044,
      "aux_brier/mean_r": 0.9614083428039205,
      "aux_brier/n_active_tok": 228.25,
      "aux_brier/n_groups": 11.75,
      "aux_brier/n_step_records": 57.0625,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.6495856074388519,
      "calib/avg_num_step_conf": 7.16015625,
      "calib/ece": 0.580296,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004283538845091306,
      "calib/mean_conf": 0.031704,
      "calib/mu_c": 0.03336601307189543,
      "calib/mu_w": 0.02908247422680412,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.010392323320605456,
      "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": 2416.0,
      "completions/max_terminated_length": 2416.0,
      "completions/mean_length": 439.09375,
      "completions/mean_terminated_length": 442.5511779785156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 121.0,
      "epoch": 0.13013333333333332,
      "grad_norm": 0.089215949177742,
      "learning_rate": 2.166666666666667e-06,
      "loss": 0.0221,
      "num_tokens": 26762556.0,
      "reward": 1.1824750900268555,
      "reward_std": 0.22523635625839233,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.4095883369445801,
      "rewards/format_reward_step": 0.96484375,
      "step": 122
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 6.48899481078824e-06,
      "aux_brier/mean_group_std": 0.07317168488485834,
      "aux_brier/mean_r": 0.9473232174387858,
      "aux_brier/n_active_tok": 249.375,
      "aux_brier/n_groups": 14.3125,
      "aux_brier/n_step_records": 62.34375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4970003749531309,
      "calib/avg_num_step_conf": 8.23828125,
      "calib/ece": 0.46628458498023717,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00017185351831021378,
      "calib/mean_conf": 0.03173913043478261,
      "calib/mu_c": 0.031825396825396826,
      "calib/mu_w": 0.03165354330708661,
      "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": 0.00945862272785462,
      "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": 1597.0,
      "completions/max_terminated_length": 1597.0,
      "completions/mean_length": 479.8125,
      "completions/mean_terminated_length": 485.5019836425781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 135.0,
      "epoch": 0.1312,
      "grad_norm": 0.25065913796424866,
      "learning_rate": 2.138888888888889e-06,
      "loss": -0.0364,
      "num_tokens": 26990676.0,
      "reward": 1.1179126501083374,
      "reward_std": 0.2727106213569641,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.5263378620147705,
      "rewards/format_reward_step": 0.98828125,
      "step": 123
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.8334267029018747e-06,
      "aux_brier/mean_group_std": 0.053535013414634715,
      "aux_brier/mean_r": 0.963970846064989,
      "aux_brier/n_active_tok": 228.125,
      "aux_brier/n_groups": 12.78125,
      "aux_brier/n_step_records": 57.03125,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.4924968866749688,
      "calib/avg_num_step_conf": 7.22265625,
      "calib/ece": 0.5387265625,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.00018405977584059346,
      "calib/mean_conf": 0.0315859375,
      "calib/mu_c": 0.03150684931506849,
      "calib/mu_w": 0.03169090909090908,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.009656626079852825,
      "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": 1428.0,
      "completions/max_terminated_length": 1428.0,
      "completions/mean_length": 443.515625,
      "completions/mean_terminated_length": 445.2549133300781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.13226666666666667,
      "grad_norm": 0.09617814421653748,
      "learning_rate": 2.1111111111111114e-06,
      "loss": 0.0235,
      "num_tokens": 27211032.0,
      "reward": 1.186445951461792,
      "reward_std": 0.13561290502548218,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.46453407406806946,
      "rewards/format_reward_step": 1.0,
      "step": 124
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.3602710374770464e-06,
      "aux_brier/mean_group_std": 0.06598502881847493,
      "aux_brier/mean_r": 0.9600569149164065,
      "aux_brier/n_active_tok": 233.875,
      "aux_brier/n_groups": 16.34375,
      "aux_brier/n_step_records": 58.46875,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5076048186284406,
      "calib/avg_num_step_conf": 7.5390625,
      "calib/ece": 0.4482456557377049,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00013171411265899824,
      "calib/mean_conf": 0.0320822131147541,
      "calib/mu_c": 0.032150769230769226,
      "calib/mu_w": 0.03201905511811023,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00040983606557377055,
      "calib/std_conf": 0.011573074071428855,
      "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": 2710.0,
      "completions/max_terminated_length": 2710.0,
      "completions/mean_length": 487.9296875,
      "completions/mean_terminated_length": 489.8431701660156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 155.0,
      "epoch": 0.13333333333333333,
      "grad_norm": 0.09179168194532394,
      "learning_rate": 2.0833333333333334e-06,
      "loss": 0.0881,
      "num_tokens": 27440750.0,
      "reward": 1.0617573261260986,
      "reward_std": 0.3050077557563782,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.5204670429229736,
      "rewards/format_reward_step": 0.94921875,
      "step": 125
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.091867457229782e-06,
      "aux_brier/mean_group_std": 0.07556235001658002,
      "aux_brier/mean_r": 0.9444273987982837,
      "aux_brier/n_active_tok": 233.625,
      "aux_brier/n_groups": 12.8125,
      "aux_brier/n_step_records": 58.40625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5751915708812261,
      "calib/avg_num_step_conf": 7.69921875,
      "calib/ece": 0.5035936254980079,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0028943167305236286,
      "calib/mean_conf": 0.03425498007968127,
      "calib/mu_c": 0.03559259259259259,
      "calib/mu_w": 0.032698275862068964,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.012785337916420253,
      "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": 2462.0,
      "completions/max_terminated_length": 2462.0,
      "completions/mean_length": 461.37109375,
      "completions/mean_terminated_length": 465.0039367675781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 154.0,
      "epoch": 0.1344,
      "grad_norm": 0.03541954979300499,
      "learning_rate": 2.0555555555555555e-06,
      "loss": -0.0137,
      "num_tokens": 27664325.0,
      "reward": 1.1399164199829102,
      "reward_std": 0.23595809936523438,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.4893532991409302,
      "rewards/format_reward_step": 0.98046875,
      "step": 126
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.4350087719217086e-06,
      "aux_brier/mean_group_std": 0.04903680719944231,
      "aux_brier/mean_r": 0.9662162115789407,
      "aux_brier/n_active_tok": 259.75,
      "aux_brier/n_groups": 17.3125,
      "aux_brier/n_step_records": 64.9375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5781046360081922,
      "calib/avg_num_step_conf": 8.1484375,
      "calib/ece": 0.482207874015748,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0029358158009060997,
      "calib/mean_conf": 0.03354015748031496,
      "calib/mu_c": 0.034961832061068704,
      "calib/mu_w": 0.032026016260162604,
      "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": 0.00994752578617108,
      "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": 2371.0,
      "completions/max_terminated_length": 2371.0,
      "completions/mean_length": 456.46484375,
      "completions/mean_terminated_length": 456.46484375,
      "completions/min_length": 168.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.13546666666666668,
      "grad_norm": 0.007647297345101833,
      "learning_rate": 2.027777777777778e-06,
      "loss": 0.04,
      "num_tokens": 27884852.0,
      "reward": 1.1365714073181152,
      "reward_std": 0.23692110180854797,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.5150356292724609,
      "rewards/format_reward_step": 0.9921875,
      "step": 127
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.3018596861518192e-06,
      "aux_brier/mean_group_std": 0.041965678466232094,
      "aux_brier/mean_r": 0.9697844503544641,
      "aux_brier/n_active_tok": 223.25,
      "aux_brier/n_groups": 14.75,
      "aux_brier/n_step_records": 55.8125,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6206041625888954,
      "calib/avg_num_step_conf": 7.02734375,
      "calib/ece": 0.4380483870967742,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004920206172114583,
      "calib/mean_conf": 0.033725806451612904,
      "calib/mu_c": 0.036324786324786335,
      "calib/mu_w": 0.03140458015267175,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.012262806807629674,
      "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": 2567.0,
      "completions/max_terminated_length": 2567.0,
      "completions/mean_length": 450.3203125,
      "completions/mean_terminated_length": 455.66009521484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 99.0,
      "epoch": 0.13653333333333334,
      "grad_norm": 0.02723911590874195,
      "learning_rate": 2.0000000000000003e-06,
      "loss": -0.0055,
      "num_tokens": 28106798.0,
      "reward": 1.0753717422485352,
      "reward_std": 0.26309192180633545,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.5436742901802063,
      "rewards/format_reward_step": 0.96484375,
      "step": 128
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.6262359635310641e-06,
      "aux_brier/mean_group_std": 0.07015446827119359,
      "aux_brier/mean_r": 0.9520226037816187,
      "aux_brier/n_active_tok": 253.125,
      "aux_brier/n_groups": 15.71875,
      "aux_brier/n_step_records": 63.28125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.497678916827853,
      "calib/avg_num_step_conf": 7.984375,
      "calib/ece": 0.5247808764940238,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0002114764667956251,
      "calib/mean_conf": 0.03697211155378487,
      "calib/mu_c": 0.036879432624113466,
      "calib/mu_w": 0.03709090909090909,
      "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": 0.011237884982858392,
      "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": 2672.0,
      "completions/max_terminated_length": 2672.0,
      "completions/mean_length": 460.859375,
      "completions/mean_terminated_length": 462.66668701171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 129.0,
      "epoch": 0.1376,
      "grad_norm": 0.0317411795258522,
      "learning_rate": 1.9722222222222224e-06,
      "loss": 0.0439,
      "num_tokens": 28327162.0,
      "reward": 1.1582276821136475,
      "reward_std": 0.1795579493045807,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.46884840726852417,
      "rewards/format_reward_step": 0.98046875,
      "step": 129
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.841466538480944e-06,
      "aux_brier/mean_group_std": 0.05399584990671969,
      "aux_brier/mean_r": 0.9670969175815727,
      "aux_brier/n_active_tok": 229.5,
      "aux_brier/n_groups": 12.8125,
      "aux_brier/n_step_records": 57.375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5582240226337448,
      "calib/avg_num_step_conf": 7.37890625,
      "calib/ece": 0.5330555555555556,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0034722222222222238,
      "calib/mean_conf": 0.03837301587301588,
      "calib/mu_c": 0.03986111111111111,
      "calib/mu_w": 0.03638888888888889,
      "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": 0.01294569495580044,
      "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": 2298.0,
      "completions/max_terminated_length": 2298.0,
      "completions/mean_length": 432.0078125,
      "completions/mean_terminated_length": 433.7019958496094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.13866666666666666,
      "grad_norm": 0.05528915300965309,
      "learning_rate": 1.944444444444445e-06,
      "loss": 0.0211,
      "num_tokens": 28543044.0,
      "reward": 1.1688967943191528,
      "reward_std": 0.19122979044914246,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.46464961767196655,
      "rewards/format_reward_step": 0.98046875,
      "step": 130
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.6664377451292012e-06,
      "aux_brier/mean_group_std": 0.04716095616572432,
      "aux_brier/mean_r": 0.9729032156268617,
      "aux_brier/n_active_tok": 228.5,
      "aux_brier/n_groups": 14.0625,
      "aux_brier/n_step_records": 57.125,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.632483660130719,
      "calib/avg_num_step_conf": 7.2734375,
      "calib/ece": 0.36559523809523814,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005039215686274509,
      "calib/mean_conf": 0.04003968253968254,
      "calib/mu_c": 0.0430392156862745,
      "calib/mu_w": 0.03799999999999999,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0004365079365079365,
      "calib/std_conf": 0.01142602209220689,
      "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": 2262.0,
      "completions/max_terminated_length": 2262.0,
      "completions/mean_length": 409.79296875,
      "completions/mean_terminated_length": 411.4000244140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 152.0,
      "epoch": 0.13973333333333332,
      "grad_norm": 0.009943638928234577,
      "learning_rate": 1.916666666666667e-06,
      "loss": 0.0278,
      "num_tokens": 28754159.0,
      "reward": 1.0452568531036377,
      "reward_std": 0.18828809261322021,
      "rewards/accuracy_reward_step": 0.3984375,
      "rewards/final_brier_reward_step": 0.6185277700424194,
      "rewards/format_reward_step": 0.984375,
      "step": 131
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.3497392188653876e-06,
      "aux_brier/mean_group_std": 0.10650228961541859,
      "aux_brier/mean_r": 0.9199625817435015,
      "aux_brier/n_active_tok": 246.75,
      "aux_brier/n_groups": 15.0625,
      "aux_brier/n_step_records": 61.6875,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5904807692307691,
      "calib/avg_num_step_conf": 7.71484375,
      "calib/ece": 0.5504330708661417,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003293589743589738,
      "calib/mean_conf": 0.04011811023622047,
      "calib/mu_c": 0.04146666666666666,
      "calib/mu_w": 0.03817307692307692,
      "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": 0.012374799241656916,
      "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": 3012.0,
      "completions/max_terminated_length": 3012.0,
      "completions/mean_length": 476.90234375,
      "completions/mean_terminated_length": 476.90234375,
      "completions/min_length": 115.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.1408,
      "grad_norm": 0.007062196731567383,
      "learning_rate": 1.888888888888889e-06,
      "loss": 0.0843,
      "num_tokens": 28981838.0,
      "reward": 1.1953049898147583,
      "reward_std": 0.24086421728134155,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.4530948996543884,
      "rewards/format_reward_step": 0.9921875,
      "step": 132
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.0968403521857155e-06,
      "aux_brier/mean_group_std": 0.044143214320325724,
      "aux_brier/mean_r": 0.9686348367300469,
      "aux_brier/n_active_tok": 266.75,
      "aux_brier/n_groups": 15.96875,
      "aux_brier/n_step_records": 66.6875,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5717361585782638,
      "calib/avg_num_step_conf": 8.484375,
      "calib/ece": 0.3443373493975903,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.004016064257028112,
      "calib/gap": -0.0015037593984962294,
      "calib/mean_conf": 0.045140562248995986,
      "calib/mu_c": 0.04421052631578948,
      "calib/mu_w": 0.04571428571428571,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.003975903614457831,
      "calib/std_conf": 0.06135092554760138,
      "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": 2814.0,
      "completions/max_terminated_length": 2814.0,
      "completions/mean_length": 533.01171875,
      "completions/mean_terminated_length": 535.1019897460938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 89.0,
      "epoch": 0.14186666666666667,
      "grad_norm": 0.047625601291656494,
      "learning_rate": 1.8611111111111113e-06,
      "loss": 0.0934,
      "num_tokens": 29224633.0,
      "reward": 1.0097026824951172,
      "reward_std": 0.29551905393600464,
      "rewards/accuracy_reward_step": 0.37109375,
      "rewards/final_brier_reward_step": 0.6247484087944031,
      "rewards/format_reward_step": 0.96484375,
      "step": 133
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.0788852103946773e-06,
      "aux_brier/mean_group_std": 0.05374224382897951,
      "aux_brier/mean_r": 0.9622196910151494,
      "aux_brier/n_active_tok": 247.75,
      "aux_brier/n_groups": 16.5,
      "aux_brier/n_step_records": 61.9375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5610791826309068,
      "calib/avg_num_step_conf": 7.79296875,
      "calib/ece": 0.42318725099601595,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00256704980842911,
      "calib/mean_conf": 0.03896414342629481,
      "calib/mu_c": 0.04034482758620689,
      "calib/mu_w": 0.03777777777777778,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.012389861761354913,
      "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": 2980.0,
      "completions/max_terminated_length": 2980.0,
      "completions/mean_length": 508.6328125,
      "completions/mean_terminated_length": 512.6378173828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.14293333333333333,
      "grad_norm": 0.05960475653409958,
      "learning_rate": 1.8333333333333333e-06,
      "loss": 0.0179,
      "num_tokens": 29463795.0,
      "reward": 1.0839262008666992,
      "reward_std": 0.2569059729576111,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.5622671842575073,
      "rewards/format_reward_step": 0.98046875,
      "step": 134
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.640086092384422e-06,
      "aux_brier/mean_group_std": 0.03808248767728147,
      "aux_brier/mean_r": 0.9762806866309728,
      "aux_brier/n_active_tok": 242.75,
      "aux_brier/n_groups": 15.03125,
      "aux_brier/n_step_records": 60.6875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4719419924337957,
      "calib/avg_num_step_conf": 7.625,
      "calib/ece": 0.47575,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0008102143757881541,
      "calib/mean_conf": 0.040123015873015866,
      "calib/mu_c": 0.03973076923076922,
      "calib/mu_w": 0.040540983606557376,
      "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": 0.012658077681348367,
      "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": 2734.0,
      "completions/max_terminated_length": 2734.0,
      "completions/mean_length": 508.1484375,
      "completions/mean_terminated_length": 508.1484375,
      "completions/min_length": 147.0,
      "completions/min_terminated_length": 147.0,
      "epoch": 0.144,
      "grad_norm": 0.00830479059368372,
      "learning_rate": 1.8055555555555557e-06,
      "loss": 0.0416,
      "num_tokens": 29699761.0,
      "reward": 1.1277220249176025,
      "reward_std": 0.23875916004180908,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.5108880400657654,
      "rewards/format_reward_step": 0.9765625,
      "step": 135
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.942148501108079e-06,
      "aux_brier/mean_group_std": 0.0384968094462189,
      "aux_brier/mean_r": 0.9768957754704042,
      "aux_brier/n_active_tok": 246.875,
      "aux_brier/n_groups": 14.25,
      "aux_brier/n_step_records": 61.71875,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5098223812213645,
      "calib/avg_num_step_conf": 7.73828125,
      "calib/ece": 0.41082283464566927,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.003937007874015748,
      "calib/gap": -0.010541831419067338,
      "calib/mean_conf": 0.04744488188976377,
      "calib/mu_c": 0.0415929203539823,
      "calib/mu_w": 0.05213475177304964,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0066929133858267716,
      "calib/std_conf": 0.07385564576126066,
      "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": 2488.0,
      "completions/max_terminated_length": 2488.0,
      "completions/mean_length": 475.51953125,
      "completions/mean_terminated_length": 475.51953125,
      "completions/min_length": 129.0,
      "completions/min_terminated_length": 129.0,
      "epoch": 0.14506666666666668,
      "grad_norm": 0.007522990461438894,
      "learning_rate": 1.777777777777778e-06,
      "loss": 0.0359,
      "num_tokens": 29929982.0,
      "reward": 1.077108383178711,
      "reward_std": 0.22018541395664215,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.5740584135055542,
      "rewards/format_reward_step": 0.984375,
      "step": 136
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.3910668150707295e-07,
      "aux_brier/mean_group_std": 0.05392120868715749,
      "aux_brier/mean_r": 0.9598089070251012,
      "aux_brier/n_active_tok": 248.0,
      "aux_brier/n_groups": 17.46875,
      "aux_brier/n_step_records": 62.0,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5445488721804511,
      "calib/avg_num_step_conf": 7.76171875,
      "calib/ece": 0.4275810276679842,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.003952569169960474,
      "calib/gap": 0.010537343358395998,
      "calib/mean_conf": 0.04672727272727272,
      "calib/mu_c": 0.05226666666666667,
      "calib/mu_w": 0.04172932330827067,
      "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": 0.06051015600309179,
      "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": 2810.0,
      "completions/max_terminated_length": 2810.0,
      "completions/mean_length": 452.89453125,
      "completions/mean_terminated_length": 454.6706237792969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.14613333333333334,
      "grad_norm": 0.008079344406723976,
      "learning_rate": 1.75e-06,
      "loss": 0.0117,
      "num_tokens": 30152907.0,
      "reward": 1.1006499528884888,
      "reward_std": 0.2560088634490967,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5588501691818237,
      "rewards/format_reward_step": 0.984375,
      "step": 137
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.5199230675454487e-06,
      "aux_brier/mean_group_std": 0.041636905255926476,
      "aux_brier/mean_r": 0.9728434556229701,
      "aux_brier/n_active_tok": 242.0,
      "aux_brier/n_groups": 14.53125,
      "aux_brier/n_step_records": 60.5,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.531087577962578,
      "calib/avg_num_step_conf": 7.69140625,
      "calib/ece": 0.5457936507936508,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0008082120582120558,
      "calib/mean_conf": 0.042301587301587305,
      "calib/mu_c": 0.04263513513513513,
      "calib/mu_w": 0.041826923076923074,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0003968253968253968,
      "calib/std_conf": 0.010091855777607468,
      "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": 2870.0,
      "completions/max_terminated_length": 2870.0,
      "completions/mean_length": 445.3828125,
      "completions/mean_terminated_length": 448.8897705078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 94.0,
      "epoch": 0.1472,
      "grad_norm": 0.016710827127099037,
      "learning_rate": 1.7222222222222224e-06,
      "loss": 0.0385,
      "num_tokens": 30371261.0,
      "reward": 1.1837337017059326,
      "reward_std": 0.2773284316062927,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.45368513464927673,
      "rewards/format_reward_step": 0.984375,
      "step": 138
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.9383198082988393e-06,
      "aux_brier/mean_group_std": 0.05748919481822279,
      "aux_brier/mean_r": 0.9611986358883595,
      "aux_brier/n_active_tok": 214.375,
      "aux_brier/n_groups": 11.84375,
      "aux_brier/n_step_records": 53.59375,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5597033528605261,
      "calib/avg_num_step_conf": 6.81640625,
      "calib/ece": 0.5160823529411764,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002291349869126269,
      "calib/mean_conf": 0.04078039215686275,
      "calib/mu_c": 0.04179577464788733,
      "calib/mu_w": 0.03950442477876106,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.009377842462847076,
      "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": 1362.0,
      "completions/max_terminated_length": 1362.0,
      "completions/mean_length": 402.2421875,
      "completions/mean_terminated_length": 403.81964111328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 124.0,
      "epoch": 0.14826666666666666,
      "grad_norm": 0.013816448859870434,
      "learning_rate": 1.6944444444444446e-06,
      "loss": -0.0254,
      "num_tokens": 30577331.0,
      "reward": 1.1713128089904785,
      "reward_std": 0.2394649088382721,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.4821265935897827,
      "rewards/format_reward_step": 0.9921875,
      "step": 139
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.4136152465270566e-06,
      "aux_brier/mean_group_std": 0.048222059957680787,
      "aux_brier/mean_r": 0.9646903379686798,
      "aux_brier/n_active_tok": 226.125,
      "aux_brier/n_groups": 13.1875,
      "aux_brier/n_step_records": 56.53125,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5741913517194416,
      "calib/avg_num_step_conf": 7.20703125,
      "calib/ece": 0.6096102362204724,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002519237316990122,
      "calib/mean_conf": 0.039996062992125984,
      "calib/mu_c": 0.040878787878787876,
      "calib/mu_w": 0.03835955056179775,
      "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": 0.00967442363600775,
      "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": 1158.0,
      "completions/max_terminated_length": 1158.0,
      "completions/mean_length": 442.96875,
      "completions/mean_terminated_length": 444.7059020996094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 139.0,
      "epoch": 0.14933333333333335,
      "grad_norm": 0.12129943072795868,
      "learning_rate": 1.6666666666666667e-06,
      "loss": 0.0224,
      "num_tokens": 30795747.0,
      "reward": 1.2441990375518799,
      "reward_std": 0.2039608657360077,
      "rewards/accuracy_reward_step": 0.6484375,
      "rewards/final_brier_reward_step": 0.3986715078353882,
      "rewards/format_reward_step": 0.9921875,
      "step": 140
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.430010737104361e-07,
      "aux_brier/mean_group_std": 0.05504474775615245,
      "aux_brier/mean_r": 0.9557840337264398,
      "aux_brier/n_active_tok": 222.5,
      "aux_brier/n_groups": 13.53125,
      "aux_brier/n_step_records": 55.625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6478922125457529,
      "calib/avg_num_step_conf": 7.171875,
      "calib/ece": 0.5089486166007905,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.006007572889057178,
      "calib/mean_conf": 0.040458498023715414,
      "calib/mu_c": 0.04316546762589928,
      "calib/mu_w": 0.037157894736842105,
      "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": 0.011003559292867749,
      "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": 2629.0,
      "completions/max_terminated_length": 2629.0,
      "completions/mean_length": 462.74609375,
      "completions/mean_terminated_length": 464.5608215332031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 126.0,
      "epoch": 0.1504,
      "grad_norm": 0.018802516162395477,
      "learning_rate": 1.638888888888889e-06,
      "loss": 0.0292,
      "num_tokens": 31021306.0,
      "reward": 1.156792402267456,
      "reward_std": 0.2195369005203247,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.48654428124427795,
      "rewards/format_reward_step": 0.984375,
      "step": 141
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.2907357074840604e-06,
      "aux_brier/mean_group_std": 0.06699872421413317,
      "aux_brier/mean_r": 0.9522982126509846,
      "aux_brier/n_active_tok": 234.0,
      "aux_brier/n_groups": 12.875,
      "aux_brier/n_step_records": 58.5,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6051772962324552,
      "calib/avg_num_step_conf": 7.3125,
      "calib/ece": 0.472313725490196,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0038469588771238503,
      "calib/mean_conf": 0.04141176470588235,
      "calib/mu_c": 0.0432824427480916,
      "calib/mu_w": 0.03943548387096775,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.010231011048859815,
      "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": 1366.0,
      "completions/max_terminated_length": 1366.0,
      "completions/mean_length": 443.5859375,
      "completions/mean_terminated_length": 445.3255310058594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 138.0,
      "epoch": 0.15146666666666667,
      "grad_norm": 0.03870021551847458,
      "learning_rate": 1.6111111111111113e-06,
      "loss": -0.0109,
      "num_tokens": 31240024.0,
      "reward": 1.1394506692886353,
      "reward_std": 0.18925997614860535,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.5265531539916992,
      "rewards/format_reward_step": 0.9921875,
      "step": 142
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.100163481371208e-06,
      "aux_brier/mean_group_std": 0.06091011977021592,
      "aux_brier/mean_r": 0.9659451125676692,
      "aux_brier/n_active_tok": 231.5,
      "aux_brier/n_groups": 13.78125,
      "aux_brier/n_step_records": 57.875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4878180661577608,
      "calib/avg_num_step_conf": 7.2578125,
      "calib/ece": 0.4822231075697211,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.000595419847328249,
      "calib/mean_conf": 0.039689243027888454,
      "calib/mu_c": 0.03940458015267176,
      "calib/mu_w": 0.04000000000000001,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.010282418553020646,
      "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": 2547.0,
      "completions/max_terminated_length": 2547.0,
      "completions/mean_length": 483.70703125,
      "completions/mean_terminated_length": 485.60394287109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 155.0,
      "epoch": 0.15253333333333333,
      "grad_norm": 0.00964946299791336,
      "learning_rate": 1.5833333333333333e-06,
      "loss": 0.0459,
      "num_tokens": 31471189.0,
      "reward": 1.118970274925232,
      "reward_std": 0.23116764426231384,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.49931907653808594,
      "rewards/format_reward_step": 0.96484375,
      "step": 143
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.7418115696443692e-06,
      "aux_brier/mean_group_std": 0.02911257219086772,
      "aux_brier/mean_r": 0.9812782295986435,
      "aux_brier/n_active_tok": 225.875,
      "aux_brier/n_groups": 13.0,
      "aux_brier/n_step_records": 56.46875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6049480194762469,
      "calib/avg_num_step_conf": 7.09375,
      "calib/ece": 0.5510916334661354,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005094222924068963,
      "calib/mean_conf": 0.04253386454183267,
      "calib/mu_c": 0.044604026845637586,
      "calib/mu_w": 0.03950980392156862,
      "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": 0.013390319115745215,
      "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": 2756.0,
      "completions/max_terminated_length": 2756.0,
      "completions/mean_length": 468.78515625,
      "completions/mean_terminated_length": 468.78515625,
      "completions/min_length": 120.0,
      "completions/min_terminated_length": 120.0,
      "epoch": 0.1536,
      "grad_norm": 0.0072034187614917755,
      "learning_rate": 1.5555555555555558e-06,
      "loss": 0.0611,
      "num_tokens": 31695326.0,
      "reward": 1.1843680143356323,
      "reward_std": 0.22442063689231873,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.4484097957611084,
      "rewards/format_reward_step": 0.98046875,
      "step": 144
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.663540706882017e-06,
      "aux_brier/mean_group_std": 0.0773822811106855,
      "aux_brier/mean_r": 0.9442634775455325,
      "aux_brier/n_active_tok": 232.875,
      "aux_brier/n_groups": 13.71875,
      "aux_brier/n_step_records": 58.21875,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5897197420634921,
      "calib/avg_num_step_conf": 7.31640625,
      "calib/ece": 0.5188671875,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0032837301587301665,
      "calib/mean_conf": 0.04363281250000001,
      "calib/mu_c": 0.04506944444444445,
      "calib/mu_w": 0.04178571428571428,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.010883165823410198,
      "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": 1474.0,
      "completions/max_terminated_length": 1474.0,
      "completions/mean_length": 425.484375,
      "completions/mean_terminated_length": 427.1529541015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 128.0,
      "epoch": 0.15466666666666667,
      "grad_norm": 0.00815511029213667,
      "learning_rate": 1.527777777777778e-06,
      "loss": 0.0387,
      "num_tokens": 31906954.0,
      "reward": 1.1840451955795288,
      "reward_std": 0.2767128646373749,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.48618084192276,
      "rewards/format_reward_step": 1.0,
      "step": 145
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.1936076177483415e-06,
      "aux_brier/mean_group_std": 0.0266679472405245,
      "aux_brier/mean_r": 0.9837740362043433,
      "aux_brier/n_active_tok": 224.5,
      "aux_brier/n_groups": 12.34375,
      "aux_brier/n_step_records": 56.125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6357101782633697,
      "calib/avg_num_step_conf": 7.3203125,
      "calib/ece": 0.39953492063492063,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005112440099674133,
      "calib/mean_conf": 0.04094126984126984,
      "calib/mu_c": 0.04380180180180179,
      "calib/mu_w": 0.03868936170212766,
      "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": 0.011170996103439421,
      "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": 1512.0,
      "completions/max_terminated_length": 1512.0,
      "completions/mean_length": 445.8671875,
      "completions/mean_terminated_length": 449.3779602050781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.15573333333333333,
      "grad_norm": 0.16069373488426208,
      "learning_rate": 1.5e-06,
      "loss": 0.0081,
      "num_tokens": 32128312.0,
      "reward": 1.069602131843567,
      "reward_std": 0.2339491844177246,
      "rewards/accuracy_reward_step": 0.43359375,
      "rewards/final_brier_reward_step": 0.5830962657928467,
      "rewards/format_reward_step": 0.98046875,
      "step": 146
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -6.122242581813042e-06,
      "aux_brier/mean_group_std": 0.04190042669157262,
      "aux_brier/mean_r": 0.9669998645399763,
      "aux_brier/n_active_tok": 228.25,
      "aux_brier/n_groups": 12.3125,
      "aux_brier/n_step_records": 57.0625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5327541827541827,
      "calib/avg_num_step_conf": 7.30859375,
      "calib/ece": 0.3733201581027668,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0006274131274131289,
      "calib/mean_conf": 0.042490118577075096,
      "calib/mu_c": 0.04285714285714286,
      "calib/mu_w": 0.04222972972972973,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0003952569169960474,
      "calib/std_conf": 0.010693580745283591,
      "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": 1305.0,
      "completions/max_terminated_length": 1305.0,
      "completions/mean_length": 433.19140625,
      "completions/mean_terminated_length": 438.3280944824219,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 137.0,
      "epoch": 0.1568,
      "grad_norm": 0.0881151482462883,
      "learning_rate": 1.4722222222222225e-06,
      "loss": -0.0246,
      "num_tokens": 32342889.0,
      "reward": 1.0542166233062744,
      "reward_std": 0.187119722366333,
      "rewards/accuracy_reward_step": 0.41015625,
      "rewards/final_brier_reward_step": 0.6074917912483215,
      "rewards/format_reward_step": 0.984375,
      "step": 147
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.676872145587986e-06,
      "aux_brier/mean_group_std": 0.05869642259014696,
      "aux_brier/mean_r": 0.9598940281989762,
      "aux_brier/n_active_tok": 231.125,
      "aux_brier/n_groups": 13.71875,
      "aux_brier/n_step_records": 57.78125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6284667713239142,
      "calib/avg_num_step_conf": 7.22265625,
      "calib/ece": 0.5698425196850394,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004559131344845639,
      "calib/mean_conf": 0.04433070866141732,
      "calib/mu_c": 0.04608974358974359,
      "calib/mu_w": 0.04153061224489795,
      "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": 0.011977048338358404,
      "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": 2889.0,
      "completions/max_terminated_length": 2889.0,
      "completions/mean_length": 441.85546875,
      "completions/mean_terminated_length": 441.85546875,
      "completions/min_length": 88.0,
      "completions/min_terminated_length": 88.0,
      "epoch": 0.15786666666666666,
      "grad_norm": 0.008789680898189545,
      "learning_rate": 1.4444444444444445e-06,
      "loss": 0.06,
      "num_tokens": 32561116.0,
      "reward": 1.2146917581558228,
      "reward_std": 0.2064843475818634,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.43689218163490295,
      "rewards/format_reward_step": 0.9921875,
      "step": 148
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 3.879356427954894e-06,
      "aux_brier/mean_group_std": 0.060673587415830385,
      "aux_brier/mean_r": 0.9645449808870379,
      "aux_brier/n_active_tok": 254.625,
      "aux_brier/n_groups": 14.8125,
      "aux_brier/n_step_records": 63.65625,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.54015503875969,
      "calib/avg_num_step_conf": 8.06640625,
      "calib/ece": 0.447988188976378,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003981705426356605,
      "calib/mean_conf": 0.04413779527559055,
      "calib/mu_c": 0.04616000000000001,
      "calib/mu_w": 0.0421782945736434,
      "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": 0.024315201910699562,
      "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": 2418.0,
      "completions/max_terminated_length": 2418.0,
      "completions/mean_length": 493.984375,
      "completions/mean_terminated_length": 493.984375,
      "completions/min_length": 102.0,
      "completions/min_terminated_length": 102.0,
      "epoch": 0.15893333333333334,
      "grad_norm": 0.0068074800074100494,
      "learning_rate": 1.4166666666666667e-06,
      "loss": 0.0289,
      "num_tokens": 32792032.0,
      "reward": 1.1180649995803833,
      "reward_std": 0.2500190734863281,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.5425726771354675,
      "rewards/format_reward_step": 0.98828125,
      "step": 149
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.052330016186435e-06,
      "aux_brier/mean_group_std": 0.05836092531492694,
      "aux_brier/mean_r": 0.9614388214260573,
      "aux_brier/n_active_tok": 224.375,
      "aux_brier/n_groups": 14.21875,
      "aux_brier/n_step_records": 56.09375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5908130601792574,
      "calib/avg_num_step_conf": 7.01171875,
      "calib/ece": 0.5207142857142857,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 3.329065300895773e-05,
      "calib/mean_conf": 0.04674603174603175,
      "calib/mu_c": 0.04676056338028168,
      "calib/mu_w": 0.04672727272727272,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.001984126984126984,
      "calib/std_conf": 0.030869139186155324,
      "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": 2596.0,
      "completions/max_terminated_length": 2596.0,
      "completions/mean_length": 432.80859375,
      "completions/mean_terminated_length": 432.80859375,
      "completions/min_length": 118.0,
      "completions/min_terminated_length": 118.0,
      "epoch": 0.16,
      "grad_norm": 0.007508590817451477,
      "learning_rate": 1.3888888888888892e-06,
      "loss": 0.0385,
      "num_tokens": 33007791.0,
      "reward": 1.1638078689575195,
      "reward_std": 0.21474945545196533,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.475543737411499,
      "rewards/format_reward_step": 0.98046875,
      "step": 150
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.030056679203085e-06,
      "aux_brier/mean_group_std": 0.05431678431607821,
      "aux_brier/mean_r": 0.9656270469246786,
      "aux_brier/n_active_tok": 235.75,
      "aux_brier/n_groups": 15.125,
      "aux_brier/n_step_records": 58.9375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6165965046562062,
      "calib/avg_num_step_conf": 7.70703125,
      "calib/ece": 0.42398406374501985,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": -0.0010352085725219909,
      "calib/mean_conf": 0.0500398406374502,
      "calib/mu_c": 0.0494871794871795,
      "calib/mu_w": 0.05052238805970149,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.003944223107569721,
      "calib/std_conf": 0.06177002720842194,
      "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": 2549.0,
      "completions/max_terminated_length": 2549.0,
      "completions/mean_length": 454.0703125,
      "completions/mean_terminated_length": 459.4545593261719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.16106666666666666,
      "grad_norm": 0.18628118932247162,
      "learning_rate": 1.3611111111111112e-06,
      "loss": 0.0335,
      "num_tokens": 33231057.0,
      "reward": 1.0839394330978394,
      "reward_std": 0.23412418365478516,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.5623202919960022,
      "rewards/format_reward_step": 0.97265625,
      "step": 151
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.4813556452608623e-06,
      "aux_brier/mean_group_std": 0.024609692455337367,
      "aux_brier/mean_r": 0.9845774417567617,
      "aux_brier/n_active_tok": 232.375,
      "aux_brier/n_groups": 14.53125,
      "aux_brier/n_step_records": 58.09375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5571491754596576,
      "calib/avg_num_step_conf": 7.4296875,
      "calib/ece": 0.4276984126984127,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0010260946483856767,
      "calib/mean_conf": 0.04634920634920635,
      "calib/mu_c": 0.046890756302521014,
      "calib/mu_w": 0.04586466165413534,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0009126984126984126,
      "calib/std_conf": 0.012286473049864192,
      "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": 2503.0,
      "completions/max_terminated_length": 2503.0,
      "completions/mean_length": 440.078125,
      "completions/mean_terminated_length": 443.5433044433594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 128.0,
      "epoch": 0.16213333333333332,
      "grad_norm": 0.06021711230278015,
      "learning_rate": 1.3333333333333334e-06,
      "loss": 0.0013,
      "num_tokens": 33449109.0,
      "reward": 1.0943185091018677,
      "reward_std": 0.253381609916687,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.556961715221405,
      "rewards/format_reward_step": 0.98046875,
      "step": 152
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 9.954473118162221e-07,
      "aux_brier/mean_group_std": 0.07337225726323636,
      "aux_brier/mean_r": 0.9445023351861707,
      "aux_brier/n_active_tok": 233.0,
      "aux_brier/n_groups": 14.96875,
      "aux_brier/n_step_records": 58.25,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5616624525916561,
      "calib/avg_num_step_conf": 7.39453125,
      "calib/ece": 0.5073517786561265,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0011826801517066943,
      "calib/mean_conf": 0.04711462450592886,
      "calib/mu_c": 0.04764285714285714,
      "calib/mu_w": 0.046460176991150445,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0005533596837944664,
      "calib/std_conf": 0.011557328474167147,
      "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": 3051.0,
      "completions/max_terminated_length": 3051.0,
      "completions/mean_length": 458.6484375,
      "completions/mean_terminated_length": 460.44708251953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.1632,
      "grad_norm": 0.14926986396312714,
      "learning_rate": 1.3055555555555556e-06,
      "loss": 0.0176,
      "num_tokens": 33673843.0,
      "reward": 1.1638129949569702,
      "reward_std": 0.24860437214374542,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.4911898374557495,
      "rewards/format_reward_step": 0.98828125,
      "step": 153
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.602123753334354e-06,
      "aux_brier/mean_group_std": 0.0533491072908468,
      "aux_brier/mean_r": 0.9626451918972292,
      "aux_brier/n_active_tok": 216.375,
      "aux_brier/n_groups": 12.21875,
      "aux_brier/n_step_records": 54.09375,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5833640383198232,
      "calib/avg_num_step_conf": 6.7890625,
      "calib/ece": 0.41314453125,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.001814664701547529,
      "calib/mean_conf": 0.04779296875000001,
      "calib/mu_c": 0.04877118644067796,
      "calib/mu_w": 0.04695652173913043,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.013546509488112554,
      "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": 930.0,
      "completions/max_terminated_length": 930.0,
      "completions/mean_length": 410.5625,
      "completions/mean_terminated_length": 412.1725769042969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 128.0,
      "epoch": 0.16426666666666667,
      "grad_norm": 0.007773758377879858,
      "learning_rate": 1.2777777777777779e-06,
      "loss": 0.017,
      "num_tokens": 33883387.0,
      "reward": 1.1004674434661865,
      "reward_std": 0.20809802412986755,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.5737451910972595,
      "rewards/format_reward_step": 0.9921875,
      "step": 154
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -6.664061979189118e-07,
      "aux_brier/mean_group_std": 0.06166461096641661,
      "aux_brier/mean_r": 0.9576725326485513,
      "aux_brier/n_active_tok": 214.0,
      "aux_brier/n_groups": 12.125,
      "aux_brier/n_step_records": 53.5,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5149309443822323,
      "calib/avg_num_step_conf": 6.703125,
      "calib/ece": 0.4006156862745098,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0006366181410974378,
      "calib/mean_conf": 0.04730588235294117,
      "calib/mu_c": 0.047657894736842114,
      "calib/mu_w": 0.04702127659574468,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.00043137254901960784,
      "calib/std_conf": 0.01128562180456048,
      "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": 1221.0,
      "completions/max_terminated_length": 1221.0,
      "completions/mean_length": 390.046875,
      "completions/mean_terminated_length": 391.5765075683594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.16533333333333333,
      "grad_norm": 0.009233204647898674,
      "learning_rate": 1.25e-06,
      "loss": -0.0112,
      "num_tokens": 34090455.0,
      "reward": 1.0910768508911133,
      "reward_std": 0.21764743328094482,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.5908705592155457,
      "rewards/format_reward_step": 0.99609375,
      "step": 155
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.325620095946281e-06,
      "aux_brier/mean_group_std": 0.054480400654208366,
      "aux_brier/mean_r": 0.9674517185710153,
      "aux_brier/n_active_tok": 243.875,
      "aux_brier/n_groups": 14.5,
      "aux_brier/n_step_records": 60.96875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.539719329438659,
      "calib/avg_num_step_conf": 7.68359375,
      "calib/ece": 0.4561354581673307,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.000800736601473187,
      "calib/mean_conf": 0.04984063745019921,
      "calib/mu_c": 0.05023622047244094,
      "calib/mu_w": 0.04943548387096775,
      "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": 0.012717216347358796,
      "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": 2768.0,
      "completions/max_terminated_length": 2768.0,
      "completions/mean_length": 465.54296875,
      "completions/mean_terminated_length": 467.36865234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 138.0,
      "epoch": 0.1664,
      "grad_norm": 0.036862220615148544,
      "learning_rate": 1.2222222222222223e-06,
      "loss": 0.0194,
      "num_tokens": 34314394.0,
      "reward": 1.1163079738616943,
      "reward_std": 0.23989202082157135,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5277324318885803,
      "rewards/format_reward_step": 0.9765625,
      "step": 156
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 6.117947797312784e-06,
      "aux_brier/mean_group_std": 0.0758693357565563,
      "aux_brier/mean_r": 0.953219320369304,
      "aux_brier/n_active_tok": 241.25,
      "aux_brier/n_groups": 14.40625,
      "aux_brier/n_step_records": 60.3125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5507772680937819,
      "calib/avg_num_step_conf": 7.7734375,
      "calib/ece": 0.5174031620553359,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0019758537206931784,
      "calib/mean_conf": 0.05176679841897234,
      "calib/mu_c": 0.05261805555555556,
      "calib/mu_w": 0.05064220183486238,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.016348386088374075,
      "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": 2498.0,
      "completions/max_terminated_length": 2498.0,
      "completions/mean_length": 427.2734375,
      "completions/mean_terminated_length": 428.94903564453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 122.0,
      "epoch": 0.16746666666666668,
      "grad_norm": 0.01833532564342022,
      "learning_rate": 1.1944444444444446e-06,
      "loss": 0.0152,
      "num_tokens": 34527504.0,
      "reward": 1.1730599403381348,
      "reward_std": 0.275108277797699,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.48130229115486145,
      "rewards/format_reward_step": 0.98046875,
      "step": 157
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.509500303924053e-06,
      "aux_brier/mean_group_std": 0.05750336788565342,
      "aux_brier/mean_r": 0.9604100471300262,
      "aux_brier/n_active_tok": 247.375,
      "aux_brier/n_groups": 16.125,
      "aux_brier/n_step_records": 61.84375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4988127326402259,
      "calib/avg_num_step_conf": 7.73046875,
      "calib/ece": 0.5304347826086955,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00037094082916183657,
      "calib/mean_conf": 0.05059288537549407,
      "calib/mu_c": 0.05074829931972789,
      "calib/mu_w": 0.05037735849056605,
      "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": 0.012512612955692191,
      "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": 3052.0,
      "completions/max_terminated_length": 3052.0,
      "completions/mean_length": 475.828125,
      "completions/mean_terminated_length": 475.828125,
      "completions/min_length": 126.0,
      "completions/min_terminated_length": 126.0,
      "epoch": 0.16853333333333334,
      "grad_norm": 0.008424046449363232,
      "learning_rate": 1.1666666666666668e-06,
      "loss": 0.1205,
      "num_tokens": 34754556.0,
      "reward": 1.1857742071151733,
      "reward_std": 0.2292053997516632,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.46965938806533813,
      "rewards/format_reward_step": 0.98828125,
      "step": 158
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.003362758770891e-06,
      "aux_brier/mean_group_std": 0.03255680191022221,
      "aux_brier/mean_r": 0.9800528170120831,
      "aux_brier/n_active_tok": 215.0,
      "aux_brier/n_groups": 11.78125,
      "aux_brier/n_step_records": 53.75,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5679563492063492,
      "calib/avg_num_step_conf": 7.04296875,
      "calib/ece": 0.4506889763779528,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0008302331349206368,
      "calib/mean_conf": 0.049783464566929135,
      "calib/mu_c": 0.04936507936507937,
      "calib/mu_w": 0.050195312500000006,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.002204724409448819,
      "calib/std_conf": 0.0193701095480198,
      "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": 1813.0,
      "completions/max_terminated_length": 1813.0,
      "completions/mean_length": 407.72265625,
      "completions/mean_terminated_length": 409.32159423828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 114.0,
      "epoch": 0.1696,
      "grad_norm": 0.1121128648519516,
      "learning_rate": 1.138888888888889e-06,
      "loss": -0.012,
      "num_tokens": 34963717.0,
      "reward": 1.1217937469482422,
      "reward_std": 0.2411247342824936,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.541862428188324,
      "rewards/format_reward_step": 0.98828125,
      "step": 159
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 7.517531570189195e-06,
      "aux_brier/mean_group_std": 0.03846520621942318,
      "aux_brier/mean_r": 0.9789132501042819,
      "aux_brier/n_active_tok": 238.125,
      "aux_brier/n_groups": 15.40625,
      "aux_brier/n_step_records": 59.53125,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6025040650406505,
      "calib/avg_num_step_conf": 7.9921875,
      "calib/ece": 0.4452016129032258,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004286829268292679,
      "calib/mean_conf": 0.05076612903225808,
      "calib/mu_c": 0.052926829268292695,
      "calib/mu_w": 0.04864000000000002,
      "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": 0.013992284089735556,
      "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": 2795.0,
      "completions/max_terminated_length": 2795.0,
      "completions/mean_length": 416.046875,
      "completions/mean_terminated_length": 422.65081787109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 144.0,
      "epoch": 0.17066666666666666,
      "grad_norm": 0.25963467359542847,
      "learning_rate": 1.111111111111111e-06,
      "loss": 0.0256,
      "num_tokens": 35175065.0,
      "reward": 1.0989573001861572,
      "reward_std": 0.26471462845802307,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.5364543199539185,
      "rewards/format_reward_step": 0.96875,
      "step": 160
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.7469933693291537e-06,
      "aux_brier/mean_group_std": 0.07084469799028036,
      "aux_brier/mean_r": 0.9581898140432695,
      "aux_brier/n_active_tok": 203.125,
      "aux_brier/n_groups": 11.1875,
      "aux_brier/n_step_records": 50.78125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.557933122333838,
      "calib/avg_num_step_conf": 6.3515625,
      "calib/ece": 0.6153333333333333,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.00392156862745098,
      "calib/gap": -0.007453557176276326,
      "calib/mean_conf": 0.05517647058823529,
      "calib/mu_c": 0.05266272189349112,
      "calib/mu_w": 0.06011627906976744,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0038823529411764705,
      "calib/std_conf": 0.06054546367695088,
      "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": 2293.0,
      "completions/max_terminated_length": 2293.0,
      "completions/mean_length": 397.84765625,
      "completions/mean_terminated_length": 397.84765625,
      "completions/min_length": 130.0,
      "completions/min_terminated_length": 130.0,
      "epoch": 0.17173333333333332,
      "grad_norm": 0.008866028860211372,
      "learning_rate": 1.0833333333333335e-06,
      "loss": 0.0211,
      "num_tokens": 35380834.0,
      "reward": 1.2528870105743408,
      "reward_std": 0.21473273634910583,
      "rewards/accuracy_reward_step": 0.66015625,
      "rewards/final_brier_reward_step": 0.3943605422973633,
      "rewards/format_reward_step": 0.98828125,
      "step": 161
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.1794646693495956e-06,
      "aux_brier/mean_group_std": 0.05350575704744897,
      "aux_brier/mean_r": 0.9654285859149274,
      "aux_brier/n_active_tok": 220.0,
      "aux_brier/n_groups": 12.40625,
      "aux_brier/n_step_records": 55.0,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5192154520111508,
      "calib/avg_num_step_conf": 7.015625,
      "calib/ece": 0.5847058823529412,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0005874153723616216,
      "calib/mean_conf": 0.050588235294117656,
      "calib/mu_c": 0.05080246913580247,
      "calib/mu_w": 0.05021505376344085,
      "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": 0.009943322659617565,
      "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": 1134.0,
      "completions/max_terminated_length": 1134.0,
      "completions/mean_length": 404.54296875,
      "completions/mean_terminated_length": 406.1294250488281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.1728,
      "grad_norm": 0.02492787502706051,
      "learning_rate": 1.0555555555555557e-06,
      "loss": -0.0009,
      "num_tokens": 35588541.0,
      "reward": 1.2370920181274414,
      "reward_std": 0.23188012838363647,
      "rewards/accuracy_reward_step": 0.6328125,
      "rewards/final_brier_reward_step": 0.4249304533004761,
      "rewards/format_reward_step": 0.99609375,
      "step": 162
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -7.77822889486135e-07,
      "aux_brier/mean_group_std": 0.052049492159757335,
      "aux_brier/mean_r": 0.9677971015727546,
      "aux_brier/n_active_tok": 225.125,
      "aux_brier/n_groups": 13.03125,
      "aux_brier/n_step_records": 56.28125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6053039702233252,
      "calib/avg_num_step_conf": 7.04296875,
      "calib/ece": 0.4588582677165355,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0048238213399503654,
      "calib/mean_conf": 0.05295275590551181,
      "calib/mu_c": 0.0553076923076923,
      "calib/mu_w": 0.050483870967741935,
      "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": 0.01644430268604241,
      "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": 2672.0,
      "completions/max_terminated_length": 2672.0,
      "completions/mean_length": 443.4921875,
      "completions/mean_terminated_length": 443.4921875,
      "completions/min_length": 141.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.17386666666666667,
      "grad_norm": 0.008522797375917435,
      "learning_rate": 1.0277777777777777e-06,
      "loss": 0.0339,
      "num_tokens": 35806907.0,
      "reward": 1.1382803916931152,
      "reward_std": 0.20851092040538788,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5374965071678162,
      "rewards/format_reward_step": 0.9921875,
      "step": 163
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.820982779625172e-06,
      "aux_brier/mean_group_std": 0.03242034688768444,
      "aux_brier/mean_r": 0.9802378489130346,
      "aux_brier/n_active_tok": 240.375,
      "aux_brier/n_groups": 15.84375,
      "aux_brier/n_step_records": 60.09375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.582537560478737,
      "calib/avg_num_step_conf": 7.74609375,
      "calib/ece": 0.47485258964143423,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0027198879551820695,
      "calib/mean_conf": 0.05104382470119522,
      "calib/mu_c": 0.05233333333333333,
      "calib/mu_w": 0.04961344537815126,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.012386620131417099,
      "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": 2792.0,
      "completions/max_terminated_length": 2792.0,
      "completions/mean_length": 466.140625,
      "completions/mean_terminated_length": 467.9686584472656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 124.0,
      "epoch": 0.17493333333333333,
      "grad_norm": 0.015697116032242775,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0351,
      "num_tokens": 36032375.0,
      "reward": 1.1348862648010254,
      "reward_std": 0.26142996549606323,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.516107439994812,
      "rewards/format_reward_step": 0.98046875,
      "step": 164
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 3.5277262844246593e-07,
      "aux_brier/mean_group_std": 0.050297317082726646,
      "aux_brier/mean_r": 0.9672013091528888,
      "aux_brier/n_active_tok": 243.625,
      "aux_brier/n_groups": 14.53125,
      "aux_brier/n_step_records": 60.90625,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.61565260669404,
      "calib/avg_num_step_conf": 7.67578125,
      "calib/ece": 0.3954117647058823,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.00784313725490196,
      "calib/gap": 0.006390444195595382,
      "calib/mean_conf": 0.05909803921568627,
      "calib/mu_c": 0.06263157894736843,
      "calib/mu_w": 0.056241134751773045,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003725490196078431,
      "calib/std_conf": 0.08267803248129973,
      "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": 1356.0,
      "completions/max_terminated_length": 1356.0,
      "completions/mean_length": 438.484375,
      "completions/mean_terminated_length": 440.2039489746094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.176,
      "grad_norm": 0.047477055341005325,
      "learning_rate": 9.722222222222224e-07,
      "loss": 0.0018,
      "num_tokens": 36250203.0,
      "reward": 1.0924279689788818,
      "reward_std": 0.187387153506279,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.5962746143341064,
      "rewards/format_reward_step": 0.99609375,
      "step": 165
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.171900063494217e-07,
      "aux_brier/mean_group_std": 0.06750884783215277,
      "aux_brier/mean_r": 0.9568796035838543,
      "aux_brier/n_active_tok": 246.625,
      "aux_brier/n_groups": 14.75,
      "aux_brier/n_step_records": 61.65625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5955882352941176,
      "calib/avg_num_step_conf": 7.70703125,
      "calib/ece": 0.5528853754940711,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00303267973856209,
      "calib/mean_conf": 0.053833992094861664,
      "calib/mu_c": 0.055032679738562094,
      "calib/mu_w": 0.052000000000000005,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0009881422924901185,
      "calib/std_conf": 0.01852047554729525,
      "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": 2983.0,
      "completions/max_terminated_length": 2983.0,
      "completions/mean_length": 486.08203125,
      "completions/mean_terminated_length": 486.08203125,
      "completions/min_length": 179.0,
      "completions/min_terminated_length": 179.0,
      "epoch": 0.17706666666666668,
      "grad_norm": 0.009201133623719215,
      "learning_rate": 9.444444444444445e-07,
      "loss": 0.0739,
      "num_tokens": 36480824.0,
      "reward": 1.20216965675354,
      "reward_std": 0.2176104187965393,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.44930312037467957,
      "rewards/format_reward_step": 0.984375,
      "step": 166
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.838929182349183e-08,
      "aux_brier/mean_group_std": 0.06980713138458353,
      "aux_brier/mean_r": 0.9489904823940546,
      "aux_brier/n_active_tok": 220.0,
      "aux_brier/n_groups": 13.375,
      "aux_brier/n_step_records": 55.0,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5848094471282876,
      "calib/avg_num_step_conf": 6.875,
      "calib/ece": 0.5824409448818898,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005668276972624793,
      "calib/mean_conf": 0.05535433070866142,
      "calib/mu_c": 0.0574074074074074,
      "calib/mu_w": 0.05173913043478261,
      "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": 0.01705919640640684,
      "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": 1325.0,
      "completions/max_terminated_length": 1325.0,
      "completions/mean_length": 422.1796875,
      "completions/mean_terminated_length": 423.8353271484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 135.0,
      "epoch": 0.17813333333333334,
      "grad_norm": 0.0106651084497571,
      "learning_rate": 9.166666666666666e-07,
      "loss": -0.0091,
      "num_tokens": 36694510.0,
      "reward": 1.236081838607788,
      "reward_std": 0.18263085186481476,
      "rewards/accuracy_reward_step": 0.6328125,
      "rewards/final_brier_reward_step": 0.42870235443115234,
      "rewards/format_reward_step": 0.9921875,
      "step": 167
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.2900859430176048e-06,
      "aux_brier/mean_group_std": 0.07416423545452544,
      "aux_brier/mean_r": 0.9448001503083485,
      "aux_brier/n_active_tok": 240.25,
      "aux_brier/n_groups": 14.03125,
      "aux_brier/n_step_records": 60.0625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4915302626570231,
      "calib/avg_num_step_conf": 7.59765625,
      "calib/ece": 0.5072964426877471,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0011799264052785169,
      "calib/mean_conf": 0.05396837944664032,
      "calib/mu_c": 0.053450704225352114,
      "calib/mu_w": 0.05463063063063063,
      "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": 0.01717058646832708,
      "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": 2539.0,
      "completions/max_terminated_length": 2539.0,
      "completions/mean_length": 491.74609375,
      "completions/mean_terminated_length": 493.6745300292969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 147.0,
      "epoch": 0.1792,
      "grad_norm": 0.09920989722013474,
      "learning_rate": 8.88888888888889e-07,
      "loss": 0.0508,
      "num_tokens": 36925069.0,
      "reward": 1.1712583303451538,
      "reward_std": 0.26346373558044434,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.48972082138061523,
      "rewards/format_reward_step": 0.98828125,
      "step": 168
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 9.291881220008236e-07,
      "aux_brier/mean_group_std": 0.046320013962909065,
      "aux_brier/mean_r": 0.9691474285849738,
      "aux_brier/n_active_tok": 222.375,
      "aux_brier/n_groups": 12.90625,
      "aux_brier/n_step_records": 55.59375,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5444451317580107,
      "calib/avg_num_step_conf": 6.9609375,
      "calib/ece": 0.48094117647058826,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002918470864777936,
      "calib/mean_conf": 0.05631372549019607,
      "calib/mu_c": 0.05766423357664234,
      "calib/mu_w": 0.054745762711864404,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.017727225318511404,
      "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": 1188.0,
      "completions/max_terminated_length": 1188.0,
      "completions/mean_length": 428.91796875,
      "completions/mean_terminated_length": 430.60003662109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 128.0,
      "epoch": 0.18026666666666666,
      "grad_norm": 0.03275342658162117,
      "learning_rate": 8.611111111111112e-07,
      "loss": 0.0182,
      "num_tokens": 37139056.0,
      "reward": 1.1580235958099365,
      "reward_std": 0.22460977733135223,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.5149070024490356,
      "rewards/format_reward_step": 0.98828125,
      "step": 169
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.7450650141740427e-06,
      "aux_brier/mean_group_std": 0.047810629905576524,
      "aux_brier/mean_r": 0.9700219712996427,
      "aux_brier/n_active_tok": 250.125,
      "aux_brier/n_groups": 13.96875,
      "aux_brier/n_step_records": 62.53125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5746103896103896,
      "calib/avg_num_step_conf": 7.94140625,
      "calib/ece": 0.50284,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0033701298701298496,
      "calib/mean_conf": 0.05716000000000001,
      "calib/mu_c": 0.05864285714285712,
      "calib/mu_w": 0.05527272727272727,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.017399264352265013,
      "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": 1858.0,
      "completions/max_terminated_length": 1858.0,
      "completions/mean_length": 453.69921875,
      "completions/mean_terminated_length": 457.2716369628906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.18133333333333335,
      "grad_norm": 0.03828566521406174,
      "learning_rate": 8.333333333333333e-07,
      "loss": -0.0102,
      "num_tokens": 37359355.0,
      "reward": 1.157741665840149,
      "reward_std": 0.20567163825035095,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.49034178256988525,
      "rewards/format_reward_step": 0.9765625,
      "step": 170
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -7.616850770109096e-07,
      "aux_brier/mean_group_std": 0.06064033757296603,
      "aux_brier/mean_r": 0.9528602629857266,
      "aux_brier/n_active_tok": 229.625,
      "aux_brier/n_groups": 14.75,
      "aux_brier/n_step_records": 57.40625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5076825396825397,
      "calib/avg_num_step_conf": 7.1875,
      "calib/ece": 0.44635059760956175,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": -0.007371428571428598,
      "calib/mean_conf": 0.0609003984063745,
      "calib/mu_c": 0.0572,
      "calib/mu_w": 0.0645714285714286,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0046215139442231075,
      "calib/std_conf": 0.06310405338507291,
      "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": 3020.0,
      "completions/max_terminated_length": 3020.0,
      "completions/mean_length": 434.984375,
      "completions/mean_terminated_length": 436.6902160644531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 99.0,
      "epoch": 0.1824,
      "grad_norm": 0.02412530779838562,
      "learning_rate": 8.055555555555557e-07,
      "loss": 0.0243,
      "num_tokens": 37577607.0,
      "reward": 1.1136420965194702,
      "reward_std": 0.26265227794647217,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.54050612449646,
      "rewards/format_reward_step": 0.98046875,
      "step": 171
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.4629824352796472e-06,
      "aux_brier/mean_group_std": 0.050449672669728565,
      "aux_brier/mean_r": 0.972808238036711,
      "aux_brier/n_active_tok": 214.375,
      "aux_brier/n_groups": 12.125,
      "aux_brier/n_step_records": 53.59375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5078609986504723,
      "calib/avg_num_step_conf": 7.1171875,
      "calib/ece": 0.5616334661354581,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002009446693657234,
      "calib/mean_conf": 0.05988047808764941,
      "calib/mu_c": 0.060641025641025637,
      "calib/mu_w": 0.0586315789473684,
      "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": 0.018840535931701704,
      "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": 2404.0,
      "completions/max_terminated_length": 2404.0,
      "completions/mean_length": 401.13671875,
      "completions/mean_terminated_length": 407.5039978027344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 123.0,
      "epoch": 0.18346666666666667,
      "grad_norm": 0.4477880299091339,
      "learning_rate": 7.777777777777779e-07,
      "loss": -0.0563,
      "num_tokens": 37783650.0,
      "reward": 1.2098934650421143,
      "reward_std": 0.2365962117910385,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.44113630056381226,
      "rewards/format_reward_step": 0.98046875,
      "step": 172
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.0819355003052227e-07,
      "aux_brier/mean_group_std": 0.06625295315938617,
      "aux_brier/mean_r": 0.9609529933912402,
      "aux_brier/n_active_tok": 236.5,
      "aux_brier/n_groups": 14.625,
      "aux_brier/n_step_records": 59.125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5375490196078432,
      "calib/avg_num_step_conf": 7.70703125,
      "calib/ece": 0.5448221343873517,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0020189542483660325,
      "calib/mean_conf": 0.05992094861660079,
      "calib/mu_c": 0.06071895424836603,
      "calib/mu_w": 0.058699999999999995,
      "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": 0.019048349027701958,
      "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": 1312.0,
      "completions/max_terminated_length": 1312.0,
      "completions/mean_length": 430.2421875,
      "completions/mean_terminated_length": 433.6299133300781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 96.0,
      "epoch": 0.18453333333333333,
      "grad_norm": 0.014841507188975811,
      "learning_rate": 7.5e-07,
      "loss": -0.0398,
      "num_tokens": 37996952.0,
      "reward": 1.2026280164718628,
      "reward_std": 0.2117912322282791,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.4511374831199646,
      "rewards/format_reward_step": 0.9765625,
      "step": 173
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.120774879019407e-06,
      "aux_brier/mean_group_std": 0.061135886942515615,
      "aux_brier/mean_r": 0.9600173147391435,
      "aux_brier/n_active_tok": 238.5,
      "aux_brier/n_groups": 14.625,
      "aux_brier/n_step_records": 59.625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.535721257075932,
      "calib/avg_num_step_conf": 7.6328125,
      "calib/ece": 0.37390000000000007,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.004,
      "calib/gap": 0.01181111327997919,
      "calib/mean_conf": 0.06210000000000001,
      "calib/mu_c": 0.06876146788990826,
      "calib/mu_w": 0.056950354609929074,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.062023302072688774,
      "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": 2635.0,
      "completions/max_terminated_length": 2635.0,
      "completions/mean_length": 499.42578125,
      "completions/mean_terminated_length": 501.38433837890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 190.0,
      "epoch": 0.1856,
      "grad_norm": 0.07152820378541946,
      "learning_rate": 7.222222222222222e-07,
      "loss": 0.0406,
      "num_tokens": 38229037.0,
      "reward": 1.0576062202453613,
      "reward_std": 0.26821964979171753,
      "rewards/accuracy_reward_step": 0.42578125,
      "rewards/final_brier_reward_step": 0.5976123809814453,
      "rewards/format_reward_step": 0.96484375,
      "step": 174
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.7371069238880654e-06,
      "aux_brier/mean_group_std": 0.05884674309013523,
      "aux_brier/mean_r": 0.9630362532326565,
      "aux_brier/n_active_tok": 227.75,
      "aux_brier/n_groups": 14.96875,
      "aux_brier/n_step_records": 56.9375,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5827347879116411,
      "calib/avg_num_step_conf": 7.3828125,
      "calib/ece": 0.3154701195219124,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00540852419026968,
      "calib/mean_conf": 0.05903187250996017,
      "calib/mu_c": 0.062414893617021266,
      "calib/mu_w": 0.05700636942675159,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.01866391206647686,
      "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": 2398.0,
      "completions/max_terminated_length": 2398.0,
      "completions/mean_length": 458.15625,
      "completions/mean_terminated_length": 459.9529724121094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 118.0,
      "epoch": 0.18666666666666668,
      "grad_norm": 0.07808632403612137,
      "learning_rate": 6.944444444444446e-07,
      "loss": -0.0022,
      "num_tokens": 38452149.0,
      "reward": 1.0183327198028564,
      "reward_std": 0.2497301697731018,
      "rewards/accuracy_reward_step": 0.3671875,
      "rewards/final_brier_reward_step": 0.6514562368392944,
      "rewards/format_reward_step": 0.9765625,
      "step": 175
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.5985905837956338e-06,
      "aux_brier/mean_group_std": 0.0884663780686098,
      "aux_brier/mean_r": 0.9432563578126447,
      "aux_brier/n_active_tok": 237.875,
      "aux_brier/n_groups": 13.84375,
      "aux_brier/n_step_records": 59.46875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5465293668954996,
      "calib/avg_num_step_conf": 7.44140625,
      "calib/ece": 0.39305555555555555,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00283371472158659,
      "calib/mean_conf": 0.05932539682539682,
      "calib/mu_c": 0.060877192982456155,
      "calib/mu_w": 0.058043478260869565,
      "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": 0.018514828619535843,
      "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": 2286.0,
      "completions/max_terminated_length": 2286.0,
      "completions/mean_length": 443.7734375,
      "completions/mean_terminated_length": 447.2677001953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 151.0,
      "epoch": 0.18773333333333334,
      "grad_norm": 0.06450923532247543,
      "learning_rate": 6.666666666666667e-07,
      "loss": -0.0179,
      "num_tokens": 38669819.0,
      "reward": 1.0819435119628906,
      "reward_std": 0.21873226761817932,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.5855870842933655,
      "rewards/format_reward_step": 0.98046875,
      "step": 176
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -5.160022302330702e-06,
      "aux_brier/mean_group_std": 0.05379418406876082,
      "aux_brier/mean_r": 0.96433970289857,
      "aux_brier/n_active_tok": 254.25,
      "aux_brier/n_groups": 16.15625,
      "aux_brier/n_step_records": 63.5625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6234283186958995,
      "calib/avg_num_step_conf": 7.9453125,
      "calib/ece": 0.41174603174603175,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.003968253968253968,
      "calib/gap": 0.012313135780628051,
      "calib/mean_conf": 0.060476190476190475,
      "calib/mu_c": 0.06697478991596639,
      "calib/mu_w": 0.05466165413533834,
      "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": 0.06021597561130154,
      "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": 3054.0,
      "completions/max_terminated_length": 3054.0,
      "completions/mean_length": 480.9765625,
      "completions/mean_terminated_length": 480.9765625,
      "completions/min_length": 139.0,
      "completions/min_terminated_length": 139.0,
      "epoch": 0.1888,
      "grad_norm": 0.007034329231828451,
      "learning_rate": 6.388888888888889e-07,
      "loss": 0.0926,
      "num_tokens": 38896781.0,
      "reward": 1.0977587699890137,
      "reward_std": 0.20862998068332672,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5707226395606995,
      "rewards/format_reward_step": 0.98046875,
      "step": 177
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -9.732071380996743e-07,
      "aux_brier/mean_group_std": 0.06649860599247942,
      "aux_brier/mean_r": 0.9601318942695777,
      "aux_brier/n_active_tok": 228.5,
      "aux_brier/n_groups": 12.90625,
      "aux_brier/n_step_records": 57.125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5752877237851662,
      "calib/avg_num_step_conf": 7.23046875,
      "calib/ece": 0.48394422310756974,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0031643222506393884,
      "calib/mean_conf": 0.057888446215139444,
      "calib/mu_c": 0.05933823529411765,
      "calib/mu_w": 0.05617391304347826,
      "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": 0.01753549348745854,
      "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": 2423.0,
      "completions/max_terminated_length": 2423.0,
      "completions/mean_length": 431.89453125,
      "completions/mean_terminated_length": 435.2952880859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.18986666666666666,
      "grad_norm": 0.045542120933532715,
      "learning_rate": 6.111111111111112e-07,
      "loss": 0.0288,
      "num_tokens": 39113418.0,
      "reward": 1.1467008590698242,
      "reward_std": 0.2595553398132324,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.5086784958839417,
      "rewards/format_reward_step": 0.9765625,
      "step": 178
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.311839944202365e-06,
      "aux_brier/mean_group_std": 0.055502624836597944,
      "aux_brier/mean_r": 0.9660979821513774,
      "aux_brier/n_active_tok": 238.125,
      "aux_brier/n_groups": 13.5,
      "aux_brier/n_step_records": 59.53125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.509009009009009,
      "calib/avg_num_step_conf": 7.46484375,
      "calib/ece": 0.504110671936759,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0007080319756376033,
      "calib/mean_conf": 0.057154150197628456,
      "calib/mu_c": 0.057464788732394356,
      "calib/mu_w": 0.05675675675675675,
      "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": 0.016267403346250885,
      "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": 2083.0,
      "completions/max_terminated_length": 2083.0,
      "completions/mean_length": 464.83203125,
      "completions/mean_terminated_length": 464.83203125,
      "completions/min_length": 154.0,
      "completions/min_terminated_length": 154.0,
      "epoch": 0.19093333333333334,
      "grad_norm": 0.009458055719733238,
      "learning_rate": 5.833333333333334e-07,
      "loss": 0.0128,
      "num_tokens": 39338679.0,
      "reward": 1.1722915172576904,
      "reward_std": 0.20855097472667694,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.493853896856308,
      "rewards/format_reward_step": 0.98828125,
      "step": 179
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.2989589782408384e-06,
      "aux_brier/mean_group_std": 0.0686843718934011,
      "aux_brier/mean_r": 0.9488519305117243,
      "aux_brier/n_active_tok": 269.625,
      "aux_brier/n_groups": 15.6875,
      "aux_brier/n_step_records": 67.40625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5707132704483784,
      "calib/avg_num_step_conf": 8.44140625,
      "calib/ece": 0.5084462151394422,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": -0.0028240082698023083,
      "calib/mean_conf": 0.06509960159362552,
      "calib/mu_c": 0.06387323943661972,
      "calib/mu_w": 0.06669724770642203,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003904382470119522,
      "calib/std_conf": 0.061094678987928926,
      "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": 2913.0,
      "completions/max_terminated_length": 2913.0,
      "completions/mean_length": 524.140625,
      "completions/mean_terminated_length": 526.1961059570312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.192,
      "grad_norm": 0.0794568806886673,
      "learning_rate": 5.555555555555555e-07,
      "loss": 0.048,
      "num_tokens": 39576715.0,
      "reward": 1.1642001867294312,
      "reward_std": 0.24491116404533386,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.484925776720047,
      "rewards/format_reward_step": 0.9765625,
      "step": 180
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 3.351923299532711e-06,
      "aux_brier/mean_group_std": 0.04639050225856281,
      "aux_brier/mean_r": 0.9709694630155832,
      "aux_brier/n_active_tok": 231.5,
      "aux_brier/n_groups": 14.03125,
      "aux_brier/n_step_records": 57.875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5408491577818435,
      "calib/avg_num_step_conf": 7.25390625,
      "calib/ece": 0.41938492063492055,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0004987067062015224,
      "calib/mean_conf": 0.06077380952380953,
      "calib/mu_c": 0.061033057851239685,
      "calib/mu_w": 0.06053435114503816,
      "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": 0.017903412278161224,
      "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": 2711.0,
      "completions/max_terminated_length": 2711.0,
      "completions/mean_length": 442.34765625,
      "completions/mean_terminated_length": 444.0823669433594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.19306666666666666,
      "grad_norm": 0.008770574815571308,
      "learning_rate": 5.277777777777779e-07,
      "loss": 0.066,
      "num_tokens": 39796220.0,
      "reward": 1.1041061878204346,
      "reward_std": 0.27390772104263306,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.5648627877235413,
      "rewards/format_reward_step": 0.98046875,
      "step": 181
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.4054107825864932e-06,
      "aux_brier/mean_group_std": 0.06660856474245658,
      "aux_brier/mean_r": 0.9614253927732934,
      "aux_brier/n_active_tok": 255.625,
      "aux_brier/n_groups": 15.34375,
      "aux_brier/n_step_records": 63.90625,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.579108617131063,
      "calib/avg_num_step_conf": 7.98828125,
      "calib/ece": 0.539763779527559,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004971620227038204,
      "calib/mean_conf": 0.05866141732283465,
      "calib/mu_c": 0.06065789473684211,
      "calib/mu_w": 0.05568627450980391,
      "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": 0.015616577338039712,
      "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": 2135.0,
      "completions/max_terminated_length": 2135.0,
      "completions/mean_length": 478.85546875,
      "completions/mean_terminated_length": 478.85546875,
      "completions/min_length": 177.0,
      "completions/min_terminated_length": 177.0,
      "epoch": 0.19413333333333332,
      "grad_norm": 0.007696555461734533,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0217,
      "num_tokens": 40024967.0,
      "reward": 1.2044800519943237,
      "reward_std": 0.18606093525886536,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.4663578271865845,
      "rewards/format_reward_step": 0.98828125,
      "step": 182
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -6.094010174262365e-07,
      "aux_brier/mean_group_std": 0.05322795634776855,
      "aux_brier/mean_r": 0.9653928837478363,
      "aux_brier/n_active_tok": 257.25,
      "aux_brier/n_groups": 16.375,
      "aux_brier/n_step_records": 64.3125,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5117647058823529,
      "calib/avg_num_step_conf": 8.1953125,
      "calib/ece": 0.4218072289156627,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0015061409179056226,
      "calib/mean_conf": 0.056104417670682735,
      "calib/mu_c": 0.056890756302521,
      "calib/mu_w": 0.05538461538461538,
      "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": 0.016637076359789144,
      "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": 2666.0,
      "completions/max_terminated_length": 2666.0,
      "completions/mean_length": 516.234375,
      "completions/mean_terminated_length": 520.2991943359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 172.0,
      "epoch": 0.1952,
      "grad_norm": 0.2508448362350464,
      "learning_rate": 4.7222222222222226e-07,
      "loss": 0.014,
      "num_tokens": 40263803.0,
      "reward": 1.0914976596832275,
      "reward_std": 0.29345816373825073,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5534909963607788,
      "rewards/format_reward_step": 0.96875,
      "step": 183
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 5.572338486192763e-06,
      "aux_brier/mean_group_std": 0.06606226210886477,
      "aux_brier/mean_r": 0.9525124889424592,
      "aux_brier/n_active_tok": 250.875,
      "aux_brier/n_groups": 15.75,
      "aux_brier/n_step_records": 62.71875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5611282248748555,
      "calib/avg_num_step_conf": 7.83984375,
      "calib/ece": 0.5181818181818181,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003923758182518285,
      "calib/mean_conf": 0.06284584980237155,
      "calib/mu_c": 0.06448979591836734,
      "calib/mu_w": 0.06056603773584906,
      "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": 0.020560427486509943,
      "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": 2429.0,
      "completions/max_terminated_length": 2429.0,
      "completions/mean_length": 465.7734375,
      "completions/mean_terminated_length": 465.7734375,
      "completions/min_length": 130.0,
      "completions/min_terminated_length": 130.0,
      "epoch": 0.19626666666666667,
      "grad_norm": 0.008814026601612568,
      "learning_rate": 4.444444444444445e-07,
      "loss": 0.0705,
      "num_tokens": 40488321.0,
      "reward": 1.1893103122711182,
      "reward_std": 0.25937968492507935,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.483803927898407,
      "rewards/format_reward_step": 0.98828125,
      "step": 184
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.1588724361244473e-06,
      "aux_brier/mean_group_std": 0.0680926022584669,
      "aux_brier/mean_r": 0.9577335660485276,
      "aux_brier/n_active_tok": 266.125,
      "aux_brier/n_groups": 18.21875,
      "aux_brier/n_step_records": 66.53125,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5301294982755255,
      "calib/avg_num_step_conf": 8.625,
      "calib/ece": 0.45564516129032256,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0006962972603631185,
      "calib/mean_conf": 0.05895161290322582,
      "calib/mu_c": 0.059291338582677176,
      "calib/mu_w": 0.05859504132231406,
      "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": 0.02424067989794577,
      "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": 2904.0,
      "completions/max_terminated_length": 2904.0,
      "completions/mean_length": 486.45703125,
      "completions/mean_terminated_length": 490.28741455078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 180.0,
      "epoch": 0.19733333333333333,
      "grad_norm": 0.06516099721193314,
      "learning_rate": 4.1666666666666667e-07,
      "loss": 0.0267,
      "num_tokens": 40719774.0,
      "reward": 1.1123558282852173,
      "reward_std": 0.2039007544517517,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5275484323501587,
      "rewards/format_reward_step": 0.96875,
      "step": 185
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.4122980576125599e-06,
      "aux_brier/mean_group_std": 0.05753200831765791,
      "aux_brier/mean_r": 0.9665061358973027,
      "aux_brier/n_active_tok": 241.625,
      "aux_brier/n_groups": 14.59375,
      "aux_brier/n_step_records": 60.40625,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5528299555116164,
      "calib/avg_num_step_conf": 7.55078125,
      "calib/ece": 0.4765490196078431,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0023739495798319465,
      "calib/mean_conf": 0.05874509803921569,
      "calib/mu_c": 0.059852941176470595,
      "calib/mu_w": 0.05747899159663865,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.000980392156862745,
      "calib/std_conf": 0.02178310474751863,
      "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": 2429.0,
      "completions/max_terminated_length": 2429.0,
      "completions/mean_length": 467.29296875,
      "completions/mean_terminated_length": 467.29296875,
      "completions/min_length": 154.0,
      "completions/min_terminated_length": 154.0,
      "epoch": 0.1984,
      "grad_norm": 0.009285667911171913,
      "learning_rate": 3.8888888888888895e-07,
      "loss": 0.0191,
      "num_tokens": 40944441.0,
      "reward": 1.1604286432266235,
      "reward_std": 0.22991374135017395,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.5245273113250732,
      "rewards/format_reward_step": 0.99609375,
      "step": 186
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 4.875297361373043e-07,
      "aux_brier/mean_group_std": 0.05111810074396666,
      "aux_brier/mean_r": 0.9716391110350127,
      "aux_brier/n_active_tok": 291.75,
      "aux_brier/n_groups": 22.875,
      "aux_brier/n_step_records": 72.9375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5274491094147583,
      "calib/avg_num_step_conf": 9.12109375,
      "calib/ece": 0.46223107569721117,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0016469465648855097,
      "calib/mean_conf": 0.062390438247011966,
      "calib/mu_c": 0.061603053435114505,
      "calib/mu_w": 0.06325000000000001,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.001354581673306773,
      "calib/std_conf": 0.025213489800807613,
      "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": 2771.0,
      "completions/max_terminated_length": 2771.0,
      "completions/mean_length": 499.36328125,
      "completions/mean_terminated_length": 503.2952880859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.19946666666666665,
      "grad_norm": 0.0630650669336319,
      "learning_rate": 3.611111111111111e-07,
      "loss": 0.006,
      "num_tokens": 41173822.0,
      "reward": 1.1337924003601074,
      "reward_std": 0.2610511779785156,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.5273569822311401,
      "rewards/format_reward_step": 0.98046875,
      "step": 187
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 3.1933831217134667e-06,
      "aux_brier/mean_group_std": 0.0658192235539247,
      "aux_brier/mean_r": 0.959594664071407,
      "aux_brier/n_active_tok": 255.625,
      "aux_brier/n_groups": 16.90625,
      "aux_brier/n_step_records": 63.90625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5583759718253705,
      "calib/avg_num_step_conf": 8.32421875,
      "calib/ece": 0.5389079999999999,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0030518306864243602,
      "calib/mean_conf": 0.06309200000000001,
      "calib/mu_c": 0.06185906040268455,
      "calib/mu_w": 0.06491089108910891,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.003,
      "calib/std_conf": 0.048295543645350957,
      "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": 3059.0,
      "completions/max_terminated_length": 3059.0,
      "completions/mean_length": 512.6796875,
      "completions/mean_terminated_length": 514.6902465820312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.20053333333333334,
      "grad_norm": 0.020949432626366615,
      "learning_rate": 3.3333333333333335e-07,
      "loss": 0.0827,
      "num_tokens": 41409140.0,
      "reward": 1.1771697998046875,
      "reward_std": 0.21333910524845123,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.4508669376373291,
      "rewards/format_reward_step": 0.96484375,
      "step": 188
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -4.662979643244025e-07,
      "aux_brier/mean_group_std": 0.06215085901039336,
      "aux_brier/mean_r": 0.9365707571128276,
      "aux_brier/n_active_tok": 234.0,
      "aux_brier/n_groups": 13.28125,
      "aux_brier/n_step_records": 58.5,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6068052930056712,
      "calib/avg_num_step_conf": 7.3125,
      "calib/ece": 0.48778656126482206,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004202898550724622,
      "calib/mean_conf": 0.05924901185770752,
      "calib/mu_c": 0.06115942028985507,
      "calib/mu_w": 0.05695652173913045,
      "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": 0.01812967260233465,
      "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": 1738.0,
      "completions/max_terminated_length": 1738.0,
      "completions/mean_length": 445.37109375,
      "completions/mean_terminated_length": 445.37109375,
      "completions/min_length": 125.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.2016,
      "grad_norm": 0.009667615406215191,
      "learning_rate": 3.055555555555556e-07,
      "loss": 0.0073,
      "num_tokens": 41630923.0,
      "reward": 1.161043643951416,
      "reward_std": 0.223192036151886,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5113620758056641,
      "rewards/format_reward_step": 0.98828125,
      "step": 189
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.4906425744865004e-06,
      "aux_brier/mean_group_std": 0.06953513418144637,
      "aux_brier/mean_r": 0.957203299778975,
      "aux_brier/n_active_tok": 278.25,
      "aux_brier/n_groups": 15.4375,
      "aux_brier/n_step_records": 69.5625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6428019323671498,
      "calib/avg_num_step_conf": 9.20703125,
      "calib/ece": 0.4777199999999999,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.004,
      "calib/gap": 0.015977455716586146,
      "calib/mean_conf": 0.06228,
      "calib/mu_c": 0.06962962962962962,
      "calib/mu_w": 0.05365217391304348,
      "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": 0.06275031155301143,
      "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": 2023.0,
      "completions/max_terminated_length": 2023.0,
      "completions/mean_length": 510.2109375,
      "completions/mean_terminated_length": 518.3095703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 180.0,
      "epoch": 0.20266666666666666,
      "grad_norm": 0.017394818365573883,
      "learning_rate": 2.7777777777777776e-07,
      "loss": -0.0192,
      "num_tokens": 41867145.0,
      "reward": 1.148287057876587,
      "reward_std": 0.2366393804550171,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.5150230526924133,
      "rewards/format_reward_step": 0.9765625,
      "step": 190
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -2.85897069242369e-06,
      "aux_brier/mean_group_std": 0.0496645300252347,
      "aux_brier/mean_r": 0.9716731173971757,
      "aux_brier/n_active_tok": 270.75,
      "aux_brier/n_groups": 18.25,
      "aux_brier/n_step_records": 67.6875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5701904761904762,
      "calib/avg_num_step_conf": 8.51171875,
      "calib/ece": 0.4376494023904382,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.008969523809523816,
      "calib/mean_conf": 0.06434262948207173,
      "calib/mu_c": 0.06880952380952382,
      "calib/mu_w": 0.059840000000000004,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.03304991600091439,
      "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": 3015.0,
      "completions/max_terminated_length": 3015.0,
      "completions/mean_length": 449.8984375,
      "completions/mean_terminated_length": 453.4409484863281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 117.0,
      "epoch": 0.20373333333333332,
      "grad_norm": 0.036954063922166824,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0236,
      "num_tokens": 42086487.0,
      "reward": 1.1174577474594116,
      "reward_std": 0.21388424932956696,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.5479558706283569,
      "rewards/format_reward_step": 0.9765625,
      "step": 191
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.8904280134623974e-06,
      "aux_brier/mean_group_std": 0.06223107335754051,
      "aux_brier/mean_r": 0.9547934977486502,
      "aux_brier/n_active_tok": 262.375,
      "aux_brier/n_groups": 18.125,
      "aux_brier/n_step_records": 65.59375,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6210419130206281,
      "calib/avg_num_step_conf": 8.609375,
      "calib/ece": 0.42072874493927126,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.008008146104322704,
      "calib/mean_conf": 0.05700404858299596,
      "calib/mu_c": 0.06118644067796611,
      "calib/mu_w": 0.053178294573643405,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.019297688807874764,
      "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": 0.015625,
      "completions/max_length": 2921.0,
      "completions/max_terminated_length": 2921.0,
      "completions/mean_length": 510.125,
      "completions/mean_terminated_length": 518.2222290039062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 153.0,
      "epoch": 0.2048,
      "grad_norm": 0.04598285257816315,
      "learning_rate": 2.2222222222222224e-07,
      "loss": 0.0224,
      "num_tokens": 42322055.0,
      "reward": 1.086470127105713,
      "reward_std": 0.23417028784751892,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5568180084228516,
      "rewards/format_reward_step": 0.96484375,
      "step": 192
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -9.376260637905531e-07,
      "aux_brier/mean_group_std": 0.04323480778009911,
      "aux_brier/mean_r": 0.9706668702024487,
      "aux_brier/n_active_tok": 289.875,
      "aux_brier/n_groups": 19.59375,
      "aux_brier/n_step_records": 72.46875,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5318066157760815,
      "calib/avg_num_step_conf": 9.2890625,
      "calib/ece": 0.4664112903225806,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0018170548704899842,
      "calib/mean_conf": 0.06181451612903227,
      "calib/mu_c": 0.06267175572519083,
      "calib/mu_w": 0.06085470085470085,
      "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": 0.02339039882139536,
      "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": 0.01171875,
      "completions/max_length": 2701.0,
      "completions/max_terminated_length": 2701.0,
      "completions/mean_length": 506.421875,
      "completions/mean_terminated_length": 512.4268798828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.20586666666666667,
      "grad_norm": 0.047434333711862564,
      "learning_rate": 1.9444444444444447e-07,
      "loss": 0.075,
      "num_tokens": 42557411.0,
      "reward": 1.122399091720581,
      "reward_std": 0.30837562680244446,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.5130344033241272,
      "rewards/format_reward_step": 0.96484375,
      "step": 193
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.8079859654427395e-07,
      "aux_brier/mean_group_std": 0.07902714387319085,
      "aux_brier/mean_r": 0.9493729210776587,
      "aux_brier/n_active_tok": 221.0,
      "aux_brier/n_groups": 11.375,
      "aux_brier/n_step_records": 55.25,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6139731827805222,
      "calib/avg_num_step_conf": 7.15234375,
      "calib/ece": 0.5078968253968255,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009739526528517353,
      "calib/mean_conf": 0.05956349206349207,
      "calib/mu_c": 0.06377622377622377,
      "calib/mu_w": 0.05403669724770642,
      "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": 0.03417113699715753,
      "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": 2643.0,
      "completions/max_terminated_length": 2643.0,
      "completions/mean_length": 415.90234375,
      "completions/mean_terminated_length": 419.1771545410156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 170.0,
      "epoch": 0.20693333333333333,
      "grad_norm": 0.04716826602816582,
      "learning_rate": 1.6666666666666668e-07,
      "loss": 0.0392,
      "num_tokens": 42769826.0,
      "reward": 1.1682668924331665,
      "reward_std": 0.2654598355293274,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.485567569732666,
      "rewards/format_reward_step": 0.9765625,
      "step": 194
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 1.943249178779083e-06,
      "aux_brier/mean_group_std": 0.06785813019765813,
      "aux_brier/mean_r": 0.9545375591300149,
      "aux_brier/n_active_tok": 236.125,
      "aux_brier/n_groups": 13.46875,
      "aux_brier/n_step_records": 59.03125,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6068923076923077,
      "calib/avg_num_step_conf": 7.47265625,
      "calib/ece": 0.4565098039215686,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.00392156862745098,
      "calib/gap": 0.00082153846153845,
      "calib/mean_conf": 0.06105882352941176,
      "calib/mu_c": 0.061461538461538456,
      "calib/mu_w": 0.060640000000000006,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0038823529411764705,
      "calib/std_conf": 0.061248901315382034,
      "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": 1289.0,
      "completions/max_terminated_length": 1289.0,
      "completions/mean_length": 453.4296875,
      "completions/mean_terminated_length": 455.2078552246094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 151.0,
      "epoch": 0.208,
      "grad_norm": 0.01760878600180149,
      "learning_rate": 1.3888888888888888e-07,
      "loss": 0.0039,
      "num_tokens": 42991888.0,
      "reward": 1.139699935913086,
      "reward_std": 0.23205438256263733,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5431749820709229,
      "rewards/format_reward_step": 0.9921875,
      "step": 195
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -3.2844297479517337e-06,
      "aux_brier/mean_group_std": 0.06294050563975621,
      "aux_brier/mean_r": 0.9594141956263865,
      "aux_brier/n_active_tok": 220.5,
      "aux_brier/n_groups": 11.21875,
      "aux_brier/n_step_records": 55.125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5544950544950545,
      "calib/avg_num_step_conf": 7.08203125,
      "calib/ece": 0.5030314960629921,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003930573930573902,
      "calib/mean_conf": 0.05996062992125985,
      "calib/mu_c": 0.06167832167832166,
      "calib/mu_w": 0.057747747747747755,
      "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": 0.016848047890287928,
      "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": 1180.0,
      "completions/max_terminated_length": 1180.0,
      "completions/mean_length": 384.16015625,
      "completions/mean_terminated_length": 387.1850280761719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 132.0,
      "epoch": 0.20906666666666668,
      "grad_norm": 0.019564451649785042,
      "learning_rate": 1.1111111111111112e-07,
      "loss": -0.0211,
      "num_tokens": 43192777.0,
      "reward": 1.1793503761291504,
      "reward_std": 0.18318453431129456,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.4986511766910553,
      "rewards/format_reward_step": 0.9921875,
      "step": 196
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.362222672551816e-06,
      "aux_brier/mean_group_std": 0.08313946287186962,
      "aux_brier/mean_r": 0.9432585961029891,
      "aux_brier/n_active_tok": 259.5,
      "aux_brier/n_groups": 16.59375,
      "aux_brier/n_step_records": 64.875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5997772536687631,
      "calib/avg_num_step_conf": 8.5390625,
      "calib/ece": 0.36132000000000003,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004682914046121568,
      "calib/mean_conf": 0.06268000000000001,
      "calib/mu_c": 0.06537735849056603,
      "calib/mu_w": 0.06069444444444446,
      "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": 0.01942209051569887,
      "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": 2798.0,
      "completions/max_terminated_length": 2798.0,
      "completions/mean_length": 446.7109375,
      "completions/mean_terminated_length": 452.0079345703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.21013333333333334,
      "grad_norm": 0.02286953665316105,
      "learning_rate": 8.333333333333334e-08,
      "loss": 0.0234,
      "num_tokens": 43412191.0,
      "reward": 1.055452585220337,
      "reward_std": 0.2655121088027954,
      "rewards/accuracy_reward_step": 0.4140625,
      "rewards/final_brier_reward_step": 0.6124355792999268,
      "rewards/format_reward_step": 0.9765625,
      "step": 197
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.6347661127014135e-06,
      "aux_brier/mean_group_std": 0.055640640710336864,
      "aux_brier/mean_r": 0.9689238525694506,
      "aux_brier/n_active_tok": 252.375,
      "aux_brier/n_groups": 15.5,
      "aux_brier/n_step_records": 63.09375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6123111582295255,
      "calib/avg_num_step_conf": 8.05078125,
      "calib/ece": 0.5529761904761904,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00546382189239332,
      "calib/mean_conf": 0.058134920634920643,
      "calib/mu_c": 0.06025974025974027,
      "calib/mu_w": 0.05479591836734695,
      "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": 0.015510290068518198,
      "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": 2896.0,
      "completions/max_terminated_length": 2896.0,
      "completions/mean_length": 456.9375,
      "completions/mean_terminated_length": 460.5354309082031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 144.0,
      "epoch": 0.2112,
      "grad_norm": 0.015106615610420704,
      "learning_rate": 5.555555555555556e-08,
      "loss": 0.0456,
      "num_tokens": 43634551.0,
      "reward": 1.2046021223068237,
      "reward_std": 0.23903879523277283,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.4512211084365845,
      "rewards/format_reward_step": 0.98046875,
      "step": 198
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": -1.486556655438953e-06,
      "aux_brier/mean_group_std": 0.06384287553559435,
      "aux_brier/mean_r": 0.9551035533752109,
      "aux_brier/n_active_tok": 250.875,
      "aux_brier/n_groups": 13.4375,
      "aux_brier/n_step_records": 62.71875,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5826656243892907,
      "calib/avg_num_step_conf": 8.23046875,
      "calib/ece": 0.4610725806451613,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0026099928343430376,
      "calib/mean_conf": 0.05908870967741936,
      "calib/mu_c": 0.060341085271317825,
      "calib/mu_w": 0.05773109243697479,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.017337965100291862,
      "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": 2482.0,
      "completions/max_terminated_length": 2482.0,
      "completions/mean_length": 499.125,
      "completions/mean_terminated_length": 505.0434875488281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 116.0,
      "epoch": 0.21226666666666666,
      "grad_norm": 0.1626821905374527,
      "learning_rate": 2.777777777777778e-08,
      "loss": 0.0458,
      "num_tokens": 43866527.0,
      "reward": 1.1197582483291626,
      "reward_std": 0.3326494097709656,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5180955529212952,
      "rewards/format_reward_step": 0.96484375,
      "step": 199
    },
    {
      "aux_brier/lambda": 0.25,
      "aux_brier/loss": 2.888534441636281e-07,
      "aux_brier/mean_group_std": 0.05886517239074551,
      "aux_brier/mean_r": 0.9617580087783129,
      "aux_brier/n_active_tok": 239.0,
      "aux_brier/n_groups": 14.5,
      "aux_brier/n_step_records": 59.75,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6468071519795658,
      "calib/avg_num_step_conf": 7.6953125,
      "calib/ece": 0.5157312252964427,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0078882503192848,
      "calib/mean_conf": 0.057391304347826085,
      "calib/mu_c": 0.06075862068965517,
      "calib/mu_w": 0.05287037037037037,
      "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": 0.02087779811258606,
      "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": 1631.0,
      "completions/max_terminated_length": 1631.0,
      "completions/mean_length": 469.28515625,
      "completions/mean_terminated_length": 472.9803161621094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.21333333333333335,
      "grad_norm": 0.24779270589351654,
      "learning_rate": 0.0,
      "loss": 0.0147,
      "num_tokens": 44094712.0,
      "reward": 1.1823010444641113,
      "reward_std": 0.2039198875427246,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.4870171844959259,
      "rewards/format_reward_step": 0.98828125,
      "step": 200
    },
    {
      "epoch": 0.21333333333333335,
      "step": 200,
      "total_flos": 0.0,
      "train_loss": 0.03735336397076026,
      "train_runtime": 15822.0969,
      "train_samples_per_second": 3.236,
      "train_steps_per_second": 0.013
    }
  ],
  "logging_steps": 1,
  "max_steps": 200,
  "num_input_tokens_seen": 44094712,
  "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
}