{
  "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": [
    {
      "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": 0.004053546581417322,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0319,
      "num_tokens": 264685.0,
      "reward": 0.045387499034404755,
      "reward_std": 0.09252828359603882,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.01655624993145466,
      "rewards/format_reward_step": 0.04296875,
      "step": 1
    },
    {
      "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": 0.006135514006018639,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0653,
      "num_tokens": 533467.0,
      "reward": 0.09099707007408142,
      "reward_std": 0.17441941797733307,
      "rewards/accuracy_reward_step": 0.03125,
      "rewards/final_brier_reward_step": 0.02965039201080799,
      "rewards/format_reward_step": 0.08984375,
      "step": 2
    },
    {
      "calib/answer_extract_rate": 0.08203125,
      "calib/auroc": 0.8676470588235294,
      "calib/avg_num_step_conf": 0.296875,
      "calib/ece": 0.8660526315789473,
      "calib/final_conf_rate": 0.07421875,
      "calib/format_rate": 0.05859375,
      "calib/frac_conf_gt_0.9": 0.9473684210526315,
      "calib/gap": 0.026470588235294135,
      "calib/mean_conf": 0.9713157894736841,
      "calib/mu_c": 0.995,
      "calib/mu_w": 0.9685294117647059,
      "calib/nonempty_final_conf_rate": 0.07421875,
      "calib/nonempty_reasoning_rate": 0.08984375,
      "calib/nonempty_step_conf_rate": 0.0703125,
      "calib/pce": 0.8660526315789473,
      "calib/std_conf": 0.024270520631934384,
      "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": 0.08203125,
      "completions/max_length": 2903.0,
      "completions/max_terminated_length": 2903.0,
      "completions/mean_length": 690.5546875,
      "completions/mean_terminated_length": 752.2637939453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0032,
      "grad_norm": 0.0056653269566595554,
      "learning_rate": 7.5e-07,
      "loss": 0.0359,
      "num_tokens": 815505.0,
      "reward": 0.042592138051986694,
      "reward_std": 0.09321815520524979,
      "rewards/accuracy_reward_step": 0.0078125,
      "rewards/final_brier_reward_step": 0.010965527966618538,
      "rewards/format_reward_step": 0.05859375,
      "step": 3
    },
    {
      "calib/answer_extract_rate": 0.078125,
      "calib/avg_num_step_conf": 0.140625,
      "calib/ece": 0.8799999999999999,
      "calib/final_conf_rate": 0.03125,
      "calib/format_rate": 0.015625,
      "calib/frac_conf_gt_0.9": 0.75,
      "calib/mean_conf": 0.88,
      "calib/mu_c": NaN,
      "calib/mu_w": 0.88,
      "calib/nonempty_final_conf_rate": 0.03125,
      "calib/nonempty_reasoning_rate": 0.08203125,
      "calib/nonempty_step_conf_rate": 0.03515625,
      "calib/pce": 0.8799999999999999,
      "calib/std_conf": 0.20266968199511246,
      "calib/step_conf_rate": 0.03515625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2918.0,
      "completions/max_terminated_length": 2918.0,
      "completions/mean_length": 714.84375,
      "completions/mean_terminated_length": 778.723388671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.004266666666666667,
      "grad_norm": 0.0025851440150290728,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0194,
      "num_tokens": 1104673.0,
      "reward": 0.009871484711766243,
      "reward_std": 0.02792077511548996,
      "rewards/accuracy_reward_step": 0.0,
      "rewards/final_brier_reward_step": 0.004117968957871199,
      "rewards/format_reward_step": 0.015625,
      "step": 4
    },
    {
      "calib/answer_extract_rate": 0.04296875,
      "calib/auroc": 0.2222222222222222,
      "calib/avg_num_step_conf": 0.234375,
      "calib/ece": 0.7763636363636364,
      "calib/final_conf_rate": 0.04296875,
      "calib/format_rate": 0.03125,
      "calib/frac_conf_gt_0.9": 0.9090909090909091,
      "calib/gap": -0.0344444444444445,
      "calib/mean_conf": 0.9581818181818181,
      "calib/mu_c": 0.9299999999999999,
      "calib/mu_w": 0.9644444444444444,
      "calib/nonempty_final_conf_rate": 0.04296875,
      "calib/nonempty_reasoning_rate": 0.0625,
      "calib/nonempty_step_conf_rate": 0.0546875,
      "calib/pce": 0.7763636363636364,
      "calib/std_conf": 0.026906633794452246,
      "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": 0.078125,
      "completions/max_length": 2915.0,
      "completions/max_terminated_length": 2915.0,
      "completions/mean_length": 659.07421875,
      "completions/mean_terminated_length": 714.927978515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.005333333333333333,
      "grad_norm": 0.004935156553983688,
      "learning_rate": 1.25e-06,
      "loss": 0.0127,
      "num_tokens": 1380084.0,
      "reward": 0.026415038853883743,
      "reward_std": 0.07068032026290894,
      "rewards/accuracy_reward_step": 0.0078125,
      "rewards/final_brier_reward_step": 0.005955078173428774,
      "rewards/format_reward_step": 0.03125,
      "step": 5
    },
    {
      "calib/answer_extract_rate": 0.11328125,
      "calib/auroc": 0.5343137254901962,
      "calib/avg_num_step_conf": 0.6796875,
      "calib/ece": 0.7021739130434783,
      "calib/final_conf_rate": 0.08984375,
      "calib/format_rate": 0.078125,
      "calib/frac_conf_gt_0.9": 0.9565217391304348,
      "calib/gap": 0.002647058823529669,
      "calib/mean_conf": 0.9630434782608696,
      "calib/mu_c": 0.9650000000000002,
      "calib/mu_w": 0.9623529411764705,
      "calib/nonempty_final_conf_rate": 0.08984375,
      "calib/nonempty_reasoning_rate": 0.13671875,
      "calib/nonempty_step_conf_rate": 0.11328125,
      "calib/pce": 0.7021739130434783,
      "calib/std_conf": 0.03469554320566958,
      "calib/step_conf_rate": 0.11328125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2996.0,
      "completions/max_terminated_length": 2996.0,
      "completions/mean_length": 605.66015625,
      "completions/mean_terminated_length": 665.4463500976562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0064,
      "grad_norm": 0.009108630008995533,
      "learning_rate": 1.5e-06,
      "loss": 0.0105,
      "num_tokens": 1641085.0,
      "reward": 0.07385234534740448,
      "reward_std": 0.15033572912216187,
      "rewards/accuracy_reward_step": 0.0234375,
      "rewards/final_brier_reward_step": 0.02270468696951866,
      "rewards/format_reward_step": 0.078125,
      "step": 6
    },
    {
      "calib/answer_extract_rate": 0.13671875,
      "calib/auroc": 0.5735294117647058,
      "calib/avg_num_step_conf": 0.32421875,
      "calib/ece": 0.7004761904761903,
      "calib/final_conf_rate": 0.08203125,
      "calib/format_rate": 0.04296875,
      "calib/frac_conf_gt_0.9": 0.7142857142857143,
      "calib/gap": 0.03897058823529409,
      "calib/mean_conf": 0.8909523809523809,
      "calib/mu_c": 0.9225,
      "calib/mu_w": 0.8835294117647059,
      "calib/nonempty_final_conf_rate": 0.08203125,
      "calib/nonempty_reasoning_rate": 0.15625,
      "calib/nonempty_step_conf_rate": 0.078125,
      "calib/pce": 0.7004761904761903,
      "calib/std_conf": 0.1668733412446116,
      "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": 0.08203125,
      "completions/max_length": 2924.0,
      "completions/max_terminated_length": 2924.0,
      "completions/mean_length": 691.96484375,
      "completions/mean_terminated_length": 753.7999877929688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.007466666666666667,
      "grad_norm": 0.005322422366589308,
      "learning_rate": 1.75e-06,
      "loss": 0.0253,
      "num_tokens": 1925652.0,
      "reward": 0.04469941556453705,
      "reward_std": 0.10981818288564682,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.015180077403783798,
      "rewards/format_reward_step": 0.04296875,
      "step": 7
    },
    {
      "calib/answer_extract_rate": 0.078125,
      "calib/auroc": 0.5185185185185185,
      "calib/avg_num_step_conf": 0.30078125,
      "calib/ece": 0.36933333333333324,
      "calib/final_conf_rate": 0.05859375,
      "calib/format_rate": 0.05078125,
      "calib/frac_conf_gt_0.9": 0.8,
      "calib/gap": -0.017777777777777892,
      "calib/mean_conf": 0.9293333333333335,
      "calib/mu_c": 0.922222222222222,
      "calib/mu_w": 0.94,
      "calib/nonempty_final_conf_rate": 0.05859375,
      "calib/nonempty_reasoning_rate": 0.0859375,
      "calib/nonempty_step_conf_rate": 0.0703125,
      "calib/pce": 0.3493333333333333,
      "calib/std_conf": 0.07361763073853679,
      "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": 0.06640625,
      "completions/max_length": 2974.0,
      "completions/max_terminated_length": 2974.0,
      "completions/mean_length": 687.71484375,
      "completions/mean_terminated_length": 736.6317749023438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.008533333333333334,
      "grad_norm": 0.004791423212736845,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0516,
      "num_tokens": 2208219.0,
      "reward": 0.08494491875171661,
      "reward_std": 0.19156719744205475,
      "rewards/accuracy_reward_step": 0.04296875,
      "rewards/final_brier_reward_step": 0.033171091228723526,
      "rewards/format_reward_step": 0.05078125,
      "step": 8
    },
    {
      "calib/answer_extract_rate": 0.09375,
      "calib/auroc": 0.07692307692307693,
      "calib/avg_num_step_conf": 0.40625,
      "calib/ece": 0.9007142857142856,
      "calib/final_conf_rate": 0.0546875,
      "calib/format_rate": 0.05078125,
      "calib/frac_conf_gt_0.9": 1.0,
      "calib/gap": -0.02384615384615374,
      "calib/mean_conf": 0.972142857142857,
      "calib/mu_c": 0.95,
      "calib/mu_w": 0.9738461538461537,
      "calib/nonempty_final_conf_rate": 0.0546875,
      "calib/nonempty_reasoning_rate": 0.1171875,
      "calib/nonempty_step_conf_rate": 0.07421875,
      "calib/pce": 0.9007142857142856,
      "calib/std_conf": 0.01739399380269419,
      "calib/step_conf_rate": 0.07421875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2936.0,
      "completions/max_terminated_length": 2936.0,
      "completions/mean_length": 563.11328125,
      "completions/mean_terminated_length": 629.5065307617188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 10.0,
      "epoch": 0.0096,
      "grad_norm": 0.004804661963135004,
      "learning_rate": 2.25e-06,
      "loss": 0.0143,
      "num_tokens": 2459912.0,
      "reward": 0.03250976651906967,
      "reward_std": 0.07374005764722824,
      "rewards/accuracy_reward_step": 0.00390625,
      "rewards/final_brier_reward_step": 0.006425781175494194,
      "rewards/format_reward_step": 0.05078125,
      "step": 9
    },
    {
      "calib/answer_extract_rate": 0.109375,
      "calib/auroc": 0.3999999999999999,
      "calib/avg_num_step_conf": 0.56640625,
      "calib/ece": 0.732813043478261,
      "calib/final_conf_rate": 0.08984375,
      "calib/format_rate": 0.078125,
      "calib/frac_conf_gt_0.9": 0.6086956521739131,
      "calib/gap": -0.15443166666666674,
      "calib/mean_conf": 0.777621739130435,
      "calib/mu_c": 0.6433333333333333,
      "calib/mu_w": 0.7977650000000001,
      "calib/nonempty_final_conf_rate": 0.08984375,
      "calib/nonempty_reasoning_rate": 0.14453125,
      "calib/nonempty_step_conf_rate": 0.1171875,
      "calib/pce": 0.6900000000000002,
      "calib/std_conf": 0.32825864663802473,
      "calib/step_conf_rate": 0.1171875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2877.0,
      "completions/max_terminated_length": 2877.0,
      "completions/mean_length": 586.82421875,
      "completions/mean_terminated_length": 656.0131225585938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.010666666666666666,
      "grad_norm": 0.0062477379105985165,
      "learning_rate": 2.5e-06,
      "loss": 0.0259,
      "num_tokens": 2716939.0,
      "reward": 0.062019336968660355,
      "reward_std": 0.13213272392749786,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.02247617021203041,
      "rewards/format_reward_step": 0.078125,
      "step": 10
    },
    {
      "calib/answer_extract_rate": 0.1640625,
      "calib/auroc": 0.35625,
      "calib/avg_num_step_conf": 0.79296875,
      "calib/ece": 0.6326470588235293,
      "calib/final_conf_rate": 0.1328125,
      "calib/format_rate": 0.12109375,
      "calib/frac_conf_gt_0.9": 0.8529411764705882,
      "calib/gap": 0.034083333333333354,
      "calib/mean_conf": 0.9179411764705881,
      "calib/mu_c": 0.942,
      "calib/mu_w": 0.9079166666666666,
      "calib/nonempty_final_conf_rate": 0.1328125,
      "calib/nonempty_reasoning_rate": 0.1953125,
      "calib/nonempty_step_conf_rate": 0.15625,
      "calib/pce": 0.628235294117647,
      "calib/std_conf": 0.1566896410060854,
      "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": 0.07421875,
      "completions/max_length": 2780.0,
      "completions/max_terminated_length": 2780.0,
      "completions/mean_length": 656.1328125,
      "completions/mean_terminated_length": 708.734130859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.011733333333333333,
      "grad_norm": 0.005708031821995974,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.0538,
      "num_tokens": 2989389.0,
      "reward": 0.12179140746593475,
      "reward_std": 0.22675037384033203,
      "rewards/accuracy_reward_step": 0.0390625,
      "rewards/final_brier_reward_step": 0.04436406493186951,
      "rewards/format_reward_step": 0.12109375,
      "step": 11
    },
    {
      "calib/answer_extract_rate": 0.203125,
      "calib/auroc": 0.6475095785440612,
      "calib/avg_num_step_conf": 0.85546875,
      "calib/ece": 0.6549999999999999,
      "calib/final_conf_rate": 0.1484375,
      "calib/format_rate": 0.12890625,
      "calib/frac_conf_gt_0.9": 0.7631578947368421,
      "calib/gap": 0.09367816091954018,
      "calib/mean_conf": 0.8918421052631578,
      "calib/mu_c": 0.9633333333333334,
      "calib/mu_w": 0.8696551724137932,
      "calib/nonempty_final_conf_rate": 0.1484375,
      "calib/nonempty_reasoning_rate": 0.2421875,
      "calib/nonempty_step_conf_rate": 0.19140625,
      "calib/pce": 0.6549999999999999,
      "calib/std_conf": 0.18988715850206647,
      "calib/step_conf_rate": 0.19140625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0859375,
      "completions/max_length": 2910.0,
      "completions/max_terminated_length": 2910.0,
      "completions/mean_length": 600.0234375,
      "completions/mean_terminated_length": 656.4359130859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0128,
      "grad_norm": 0.009454714134335518,
      "learning_rate": 3e-06,
      "loss": 0.0693,
      "num_tokens": 3247171.0,
      "reward": 0.13060665130615234,
      "reward_std": 0.2470502257347107,
      "rewards/accuracy_reward_step": 0.0390625,
      "rewards/final_brier_reward_step": 0.0541820302605629,
      "rewards/format_reward_step": 0.12890625,
      "step": 12
    },
    {
      "calib/answer_extract_rate": 0.2578125,
      "calib/auroc": 0.5850340136054422,
      "calib/avg_num_step_conf": 1.01171875,
      "calib/ece": 0.7215344827586208,
      "calib/final_conf_rate": 0.2265625,
      "calib/format_rate": 0.1796875,
      "calib/frac_conf_gt_0.9": 0.6379310344827587,
      "calib/gap": 0.07754875283446738,
      "calib/mean_conf": 0.8767068965517241,
      "calib/mu_c": 0.9422222222222223,
      "calib/mu_w": 0.8646734693877549,
      "calib/nonempty_final_conf_rate": 0.2265625,
      "calib/nonempty_reasoning_rate": 0.30859375,
      "calib/nonempty_step_conf_rate": 0.2421875,
      "calib/pce": 0.7215344827586208,
      "calib/std_conf": 0.17499911414757954,
      "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": 0.0390625,
      "completions/max_length": 3069.0,
      "completions/max_terminated_length": 3069.0,
      "completions/mean_length": 648.640625,
      "completions/mean_terminated_length": 675.0081176757812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.013866666666666666,
      "grad_norm": 0.01001790538430214,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.1389,
      "num_tokens": 3517815.0,
      "reward": 0.16192597150802612,
      "reward_std": 0.31205716729164124,
      "rewards/accuracy_reward_step": 0.0390625,
      "rewards/final_brier_reward_step": 0.06603945791721344,
      "rewards/format_reward_step": 0.1796875,
      "step": 13
    },
    {
      "calib/answer_extract_rate": 0.32421875,
      "calib/auroc": 0.5225490196078432,
      "calib/avg_num_step_conf": 1.453125,
      "calib/ece": 0.5990408450704225,
      "calib/final_conf_rate": 0.27734375,
      "calib/format_rate": 0.24609375,
      "calib/frac_conf_gt_0.9": 0.647887323943662,
      "calib/gap": 0.03117450980392167,
      "calib/mean_conf": 0.853607042253521,
      "calib/mu_c": 0.876,
      "calib/mu_w": 0.8448254901960783,
      "calib/nonempty_final_conf_rate": 0.27734375,
      "calib/nonempty_reasoning_rate": 0.3671875,
      "calib/nonempty_step_conf_rate": 0.3046875,
      "calib/pce": 0.5854788732394367,
      "calib/std_conf": 0.23616655755663868,
      "calib/step_conf_rate": 0.3046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2982.0,
      "completions/max_terminated_length": 2982.0,
      "completions/mean_length": 554.3046875,
      "completions/mean_terminated_length": 596.2269287109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.014933333333333333,
      "grad_norm": 0.01179414987564087,
      "learning_rate": 3.5e-06,
      "loss": 0.0675,
      "num_tokens": 3765117.0,
      "reward": 0.2536017894744873,
      "reward_std": 0.3610790967941284,
      "rewards/accuracy_reward_step": 0.078125,
      "rewards/final_brier_reward_step": 0.10485979169607162,
      "rewards/format_reward_step": 0.24609375,
      "step": 14
    },
    {
      "calib/answer_extract_rate": 0.40625,
      "calib/auroc": 0.47217741935483876,
      "calib/avg_num_step_conf": 1.87890625,
      "calib/ece": 0.7005487804878048,
      "calib/final_conf_rate": 0.3203125,
      "calib/format_rate": 0.2890625,
      "calib/frac_conf_gt_0.9": 0.8292682926829268,
      "calib/gap": -0.06138709677419352,
      "calib/mean_conf": 0.9159146341463413,
      "calib/mu_c": 0.8694999999999998,
      "calib/mu_w": 0.9308870967741933,
      "calib/nonempty_final_conf_rate": 0.3203125,
      "calib/nonempty_reasoning_rate": 0.47265625,
      "calib/nonempty_step_conf_rate": 0.37109375,
      "calib/pce": 0.6862804878048778,
      "calib/std_conf": 0.1515431357104759,
      "calib/step_conf_rate": 0.37109375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2756.0,
      "completions/max_terminated_length": 2756.0,
      "completions/mean_length": 503.78515625,
      "completions/mean_terminated_length": 517.94775390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 5.0,
      "epoch": 0.016,
      "grad_norm": 0.011700741946697235,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.1398,
      "num_tokens": 4001966.0,
      "reward": 0.2702401876449585,
      "reward_std": 0.41204118728637695,
      "rewards/accuracy_reward_step": 0.078125,
      "rewards/final_brier_reward_step": 0.09516787528991699,
      "rewards/format_reward_step": 0.2890625,
      "step": 15
    },
    {
      "calib/answer_extract_rate": 0.55078125,
      "calib/auroc": 0.4697368421052632,
      "calib/avg_num_step_conf": 2.83203125,
      "calib/ece": 0.5950789062500003,
      "calib/final_conf_rate": 0.5,
      "calib/format_rate": 0.4453125,
      "calib/frac_conf_gt_0.9": 0.7265625,
      "calib/gap": 0.03276497076023377,
      "calib/mean_conf": 0.8911726562500002,
      "calib/mu_c": 0.9142105263157895,
      "calib/mu_w": 0.8814455555555557,
      "calib/nonempty_final_conf_rate": 0.5,
      "calib/nonempty_reasoning_rate": 0.65234375,
      "calib/nonempty_step_conf_rate": 0.5625,
      "calib/pce": 0.5946882812500003,
      "calib/std_conf": 0.18623632488170597,
      "calib/step_conf_rate": 0.5625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2944.0,
      "completions/max_terminated_length": 2944.0,
      "completions/mean_length": 465.9609375,
      "completions/mean_terminated_length": 492.9173278808594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.017066666666666667,
      "grad_norm": 0.012210188433527946,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0795,
      "num_tokens": 4230100.0,
      "reward": 0.4675709009170532,
      "reward_std": 0.5007450580596924,
      "rewards/accuracy_reward_step": 0.1484375,
      "rewards/final_brier_reward_step": 0.19295427203178406,
      "rewards/format_reward_step": 0.4453125,
      "step": 16
    },
    {
      "calib/answer_extract_rate": 0.59765625,
      "calib/auroc": 0.5087833441769682,
      "calib/avg_num_step_conf": 2.65625,
      "calib/ece": 0.683408888888889,
      "calib/final_conf_rate": 0.52734375,
      "calib/format_rate": 0.453125,
      "calib/frac_conf_gt_0.9": 0.7333333333333333,
      "calib/gap": 0.03695647364996746,
      "calib/mean_conf": 0.8982237037037036,
      "calib/mu_c": 0.9272413793103447,
      "calib/mu_w": 0.8902849056603772,
      "calib/nonempty_final_conf_rate": 0.52734375,
      "calib/nonempty_reasoning_rate": 0.66796875,
      "calib/nonempty_step_conf_rate": 0.546875,
      "calib/pce": 0.683408888888889,
      "calib/std_conf": 0.17078073627484708,
      "calib/step_conf_rate": 0.546875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3000.0,
      "completions/max_terminated_length": 3000.0,
      "completions/mean_length": 409.4296875,
      "completions/mean_terminated_length": 415.9285888671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 8.0,
      "epoch": 0.018133333333333335,
      "grad_norm": 0.012316138483583927,
      "learning_rate": 4.25e-06,
      "loss": 0.1109,
      "num_tokens": 4438442.0,
      "reward": 0.42024004459381104,
      "reward_std": 0.49231672286987305,
      "rewards/accuracy_reward_step": 0.11328125,
      "rewards/final_brier_reward_step": 0.16079255938529968,
      "rewards/format_reward_step": 0.453125,
      "step": 17
    },
    {
      "calib/answer_extract_rate": 0.71484375,
      "calib/auroc": 0.5049441494231826,
      "calib/avg_num_step_conf": 3.51953125,
      "calib/ece": 0.6499999999999999,
      "calib/final_conf_rate": 0.6640625,
      "calib/format_rate": 0.59765625,
      "calib/frac_conf_gt_0.9": 0.7058823529411765,
      "calib/gap": -0.003611060245376607,
      "calib/mean_conf": 0.882,
      "calib/mu_c": 0.8793023255813952,
      "calib/mu_w": 0.8829133858267718,
      "calib/nonempty_final_conf_rate": 0.6640625,
      "calib/nonempty_reasoning_rate": 0.7890625,
      "calib/nonempty_step_conf_rate": 0.6953125,
      "calib/pce": 0.6395294117647058,
      "calib/std_conf": 0.18625283258801414,
      "calib/step_conf_rate": 0.6953125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2075.0,
      "completions/max_terminated_length": 2075.0,
      "completions/mean_length": 387.59375,
      "completions/mean_terminated_length": 395.31475830078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0192,
      "grad_norm": 0.010788912884891033,
      "learning_rate": 4.5e-06,
      "loss": 0.091,
      "num_tokens": 4648386.0,
      "reward": 0.5889451503753662,
      "reward_std": 0.5265281796455383,
      "rewards/accuracy_reward_step": 0.171875,
      "rewards/final_brier_reward_step": 0.23648397624492645,
      "rewards/format_reward_step": 0.59765625,
      "step": 18
    },
    {
      "calib/answer_extract_rate": 0.890625,
      "calib/auroc": 0.5793439716312057,
      "calib/avg_num_step_conf": 4.60546875,
      "calib/ece": 0.7155063988095238,
      "calib/final_conf_rate": 0.875,
      "calib/format_rate": 0.81640625,
      "calib/frac_conf_gt_0.9": 0.65625,
      "calib/gap": 0.06672523640661943,
      "calib/mean_conf": 0.8762206845238095,
      "calib/mu_c": 0.9322222222222223,
      "calib/mu_w": 0.8654969858156029,
      "calib/nonempty_final_conf_rate": 0.875,
      "calib/nonempty_reasoning_rate": 0.9609375,
      "calib/nonempty_step_conf_rate": 0.92578125,
      "calib/pce": 0.7155063988095238,
      "calib/std_conf": 0.18285521951342068,
      "calib/step_conf_rate": 0.92578125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3038.0,
      "completions/max_terminated_length": 3038.0,
      "completions/mean_length": 322.078125,
      "completions/mean_terminated_length": 322.078125,
      "completions/min_length": 24.0,
      "completions/min_terminated_length": 24.0,
      "epoch": 0.020266666666666665,
      "grad_norm": 0.010422209277749062,
      "learning_rate": 4.75e-06,
      "loss": 0.1375,
      "num_tokens": 4835598.0,
      "reward": 0.696550190448761,
      "reward_std": 0.4772275686264038,
      "rewards/accuracy_reward_step": 0.1484375,
      "rewards/final_brier_reward_step": 0.2798190712928772,
      "rewards/format_reward_step": 0.81640625,
      "step": 19
    },
    {
      "calib/answer_extract_rate": 0.89453125,
      "calib/auroc": 0.4727515243902438,
      "calib/avg_num_step_conf": 4.66796875,
      "calib/ece": 0.6378179824561403,
      "calib/final_conf_rate": 0.890625,
      "calib/format_rate": 0.84765625,
      "calib/frac_conf_gt_0.9": 0.7368421052631579,
      "calib/gap": -0.024479801829268233,
      "calib/mean_conf": 0.8956864035087719,
      "calib/mu_c": 0.878078125,
      "calib/mu_w": 0.9025579268292683,
      "calib/nonempty_final_conf_rate": 0.890625,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.95703125,
      "calib/pce": 0.6264013157894737,
      "calib/std_conf": 0.16848325998176433,
      "calib/step_conf_rate": 0.95703125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2049.0,
      "completions/max_terminated_length": 2049.0,
      "completions/mean_length": 279.16015625,
      "completions/mean_terminated_length": 280.2549133300781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 27.0,
      "epoch": 0.021333333333333333,
      "grad_norm": 0.011439248919487,
      "learning_rate": 5e-06,
      "loss": 0.0787,
      "num_tokens": 5011935.0,
      "reward": 0.8440353870391846,
      "reward_std": 0.5284815430641174,
      "rewards/accuracy_reward_step": 0.25390625,
      "rewards/final_brier_reward_step": 0.33260199427604675,
      "rewards/format_reward_step": 0.84765625,
      "step": 20
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5728703703703703,
      "calib/avg_num_step_conf": 4.91796875,
      "calib/ece": 0.6571620833333334,
      "calib/final_conf_rate": 0.9375,
      "calib/format_rate": 0.9140625,
      "calib/frac_conf_gt_0.9": 0.7583333333333333,
      "calib/gap": 0.049561666666666615,
      "calib/mean_conf": 0.9028287500000001,
      "calib/mu_c": 0.94,
      "calib/mu_w": 0.8904383333333333,
      "calib/nonempty_final_conf_rate": 0.9375,
      "calib/nonempty_reasoning_rate": 0.9765625,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.6549954166666667,
      "calib/std_conf": 0.16495550172224477,
      "calib/step_conf_rate": 0.96484375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2138.0,
      "completions/max_terminated_length": 2138.0,
      "completions/mean_length": 292.421875,
      "completions/mean_terminated_length": 292.421875,
      "completions/min_length": 27.0,
      "completions/min_terminated_length": 27.0,
      "epoch": 0.0224,
      "grad_norm": 0.009296974167227745,
      "learning_rate": 4.9722222222222224e-06,
      "loss": 0.1052,
      "num_tokens": 5189755.0,
      "reward": 0.8685685396194458,
      "reward_std": 0.5453428030014038,
      "rewards/accuracy_reward_step": 0.23828125,
      "rewards/final_brier_reward_step": 0.3465120196342468,
      "rewards/format_reward_step": 0.9140625,
      "step": 21
    },
    {
      "calib/answer_extract_rate": 0.94140625,
      "calib/auroc": 0.4936801881246326,
      "calib/avg_num_step_conf": 4.83984375,
      "calib/ece": 0.6979629629629629,
      "calib/final_conf_rate": 0.94921875,
      "calib/format_rate": 0.91015625,
      "calib/frac_conf_gt_0.9": 0.7242798353909465,
      "calib/gap": 0.008095238095238044,
      "calib/mean_conf": 0.9201851851851852,
      "calib/mu_c": 0.9264814814814814,
      "calib/mu_w": 0.9183862433862433,
      "calib/nonempty_final_conf_rate": 0.94921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.6979629629629629,
      "calib/std_conf": 0.1163349192765126,
      "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": 2037.0,
      "completions/max_terminated_length": 2037.0,
      "completions/mean_length": 258.171875,
      "completions/mean_terminated_length": 258.171875,
      "completions/min_length": 13.0,
      "completions/min_terminated_length": 13.0,
      "epoch": 0.023466666666666667,
      "grad_norm": 0.00845903530716896,
      "learning_rate": 4.944444444444445e-06,
      "loss": 0.0093,
      "num_tokens": 5357663.0,
      "reward": 0.8230749368667603,
      "reward_std": 0.546536922454834,
      "rewards/accuracy_reward_step": 0.21484375,
      "rewards/final_brier_reward_step": 0.3063061237335205,
      "rewards/format_reward_step": 0.91015625,
      "step": 22
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.504986125563649,
      "calib/avg_num_step_conf": 4.4375,
      "calib/ece": 0.6572983870967741,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.93359375,
      "calib/frac_conf_gt_0.9": 0.717741935483871,
      "calib/gap": 0.02962365591397864,
      "calib/mean_conf": 0.9047177419354839,
      "calib/mu_c": 0.9269354838709678,
      "calib/mu_w": 0.8973118279569892,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.6560080645161289,
      "calib/std_conf": 0.1523363776098416,
      "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": 2508.0,
      "completions/max_terminated_length": 2508.0,
      "completions/mean_length": 261.12890625,
      "completions/mean_terminated_length": 261.12890625,
      "completions/min_length": 77.0,
      "completions/min_terminated_length": 77.0,
      "epoch": 0.024533333333333334,
      "grad_norm": 0.009214174933731556,
      "learning_rate": 4.9166666666666665e-06,
      "loss": 0.0975,
      "num_tokens": 5528448.0,
      "reward": 0.890078067779541,
      "reward_std": 0.47853392362594604,
      "rewards/accuracy_reward_step": 0.24609375,
      "rewards/final_brier_reward_step": 0.3543750047683716,
      "rewards/format_reward_step": 0.93359375,
      "step": 23
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5610976065818997,
      "calib/avg_num_step_conf": 4.90234375,
      "calib/ece": 0.6875303643724695,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.7489878542510121,
      "calib/gap": 0.033760284218399605,
      "calib/mean_conf": 0.9142510121457489,
      "calib/mu_c": 0.9403571428571429,
      "calib/mu_w": 0.9065968586387433,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.6875303643724695,
      "calib/std_conf": 0.13924662007901306,
      "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": 1363.0,
      "completions/max_terminated_length": 1363.0,
      "completions/mean_length": 251.2734375,
      "completions/mean_terminated_length": 251.2734375,
      "completions/min_length": 53.0,
      "completions/min_terminated_length": 53.0,
      "epoch": 0.0256,
      "grad_norm": 0.008825111202895641,
      "learning_rate": 4.888888888888889e-06,
      "loss": 0.0102,
      "num_tokens": 5697286.0,
      "reward": 0.8552491664886475,
      "reward_std": 0.5274129509925842,
      "rewards/accuracy_reward_step": 0.21875,
      "rewards/final_brier_reward_step": 0.3276859521865845,
      "rewards/format_reward_step": 0.9453125,
      "step": 24
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4502214078485265,
      "calib/avg_num_step_conf": 4.87109375,
      "calib/ece": 0.623585657370518,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.6932270916334662,
      "calib/gap": -0.024641930065658957,
      "calib/mean_conf": 0.9000796812749005,
      "calib/mu_c": 0.8827027027027026,
      "calib/mu_w": 0.9073446327683615,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.6144223107569722,
      "calib/std_conf": 0.14409213528743076,
      "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": 721.0,
      "completions/max_terminated_length": 721.0,
      "completions/mean_length": 237.2734375,
      "completions/mean_terminated_length": 238.2039337158203,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 63.0,
      "epoch": 0.02666666666666667,
      "grad_norm": 0.007968959398567677,
      "learning_rate": 4.861111111111111e-06,
      "loss": -0.0052,
      "num_tokens": 5861252.0,
      "reward": 0.9608609080314636,
      "reward_std": 0.5278224945068359,
      "rewards/accuracy_reward_step": 0.2890625,
      "rewards/final_brier_reward_step": 0.382659375667572,
      "rewards/format_reward_step": 0.9609375,
      "step": 25
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.51286333273154,
      "calib/avg_num_step_conf": 5.19140625,
      "calib/ece": 0.6699598393574299,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.6746987951807228,
      "calib/gap": 0.006085936089546884,
      "calib/mean_conf": 0.8942971887550201,
      "calib/mu_c": 0.8989655172413794,
      "calib/mu_w": 0.8928795811518325,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.6656626506024097,
      "calib/std_conf": 0.16410279382109272,
      "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": 2810.0,
      "completions/max_terminated_length": 2810.0,
      "completions/mean_length": 263.484375,
      "completions/mean_terminated_length": 263.484375,
      "completions/min_length": 51.0,
      "completions/min_terminated_length": 51.0,
      "epoch": 0.027733333333333332,
      "grad_norm": 0.00924302265048027,
      "learning_rate": 4.833333333333333e-06,
      "loss": 0.034,
      "num_tokens": 6033944.0,
      "reward": 0.8833884596824646,
      "reward_std": 0.4853111505508423,
      "rewards/accuracy_reward_step": 0.2265625,
      "rewards/final_brier_reward_step": 0.3488081991672516,
      "rewards/format_reward_step": 0.96484375,
      "step": 26
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5088881706528765,
      "calib/avg_num_step_conf": 4.64453125,
      "calib/ece": 0.6110404,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.6,
      "calib/gap": 0.02143771816418849,
      "calib/mean_conf": 0.8770404,
      "calib/mu_c": 0.8926470588235293,
      "calib/mu_w": 0.8712093406593409,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.6080404,
      "calib/std_conf": 0.15886140125228657,
      "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": 1076.0,
      "completions/max_terminated_length": 1076.0,
      "completions/mean_length": 235.78515625,
      "completions/mean_terminated_length": 236.70982360839844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 76.0,
      "epoch": 0.0288,
      "grad_norm": 0.008842146955430508,
      "learning_rate": 4.805555555555556e-06,
      "loss": 0.0214,
      "num_tokens": 6199521.0,
      "reward": 0.9590640068054199,
      "reward_std": 0.4516690969467163,
      "rewards/accuracy_reward_step": 0.26953125,
      "rewards/final_brier_reward_step": 0.406409353017807,
      "rewards/format_reward_step": 0.97265625,
      "step": 27
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4355572998430141,
      "calib/avg_num_step_conf": 4.66796875,
      "calib/ece": 0.6103968253968255,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.5952380952380952,
      "calib/gap": -0.002461538461538626,
      "calib/mean_conf": 0.8863492063492064,
      "calib/mu_c": 0.8845714285714283,
      "calib/mu_w": 0.887032967032967,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.609484126984127,
      "calib/std_conf": 0.13989850728524575,
      "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": 1916.0,
      "completions/max_terminated_length": 1916.0,
      "completions/mean_length": 262.296875,
      "completions/mean_terminated_length": 263.32550048828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 71.0,
      "epoch": 0.029866666666666666,
      "grad_norm": 0.007787063717842102,
      "learning_rate": 4.777777777777778e-06,
      "loss": 0.0222,
      "num_tokens": 6373613.0,
      "reward": 0.9608750343322754,
      "reward_std": 0.49356719851493835,
      "rewards/accuracy_reward_step": 0.27734375,
      "rewards/final_brier_reward_step": 0.3944062292575836,
      "rewards/format_reward_step": 0.97265625,
      "step": 28
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4803363628641808,
      "calib/avg_num_step_conf": 5.140625,
      "calib/ece": 0.6088582677165355,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.4094488188976378,
      "calib/gap": 0.012082731674169622,
      "calib/mean_conf": 0.849015748031496,
      "calib/mu_c": 0.8581967213114754,
      "calib/mu_w": 0.8461139896373058,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.6088582677165355,
      "calib/std_conf": 0.14378535425576625,
      "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": 1972.0,
      "completions/max_terminated_length": 1972.0,
      "completions/mean_length": 284.26953125,
      "completions/mean_terminated_length": 284.26953125,
      "completions/min_length": 59.0,
      "completions/min_terminated_length": 59.0,
      "epoch": 0.030933333333333334,
      "grad_norm": 0.007615217939019203,
      "learning_rate": 4.75e-06,
      "loss": 0.0375,
      "num_tokens": 6553514.0,
      "reward": 0.9362058639526367,
      "reward_std": 0.4864776134490967,
      "rewards/accuracy_reward_step": 0.23828125,
      "rewards/final_brier_reward_step": 0.41928672790527344,
      "rewards/format_reward_step": 0.9765625,
      "step": 29
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.49706959706959714,
      "calib/avg_num_step_conf": 4.69140625,
      "calib/ece": 0.5245275590551182,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.2637795275590551,
      "calib/gap": 0.01864631664631644,
      "calib/mean_conf": 0.7804330708661418,
      "calib/mu_c": 0.7943076923076923,
      "calib/mu_w": 0.7756613756613758,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.5245275590551182,
      "calib/std_conf": 0.18300695269057257,
      "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": 689.0,
      "completions/max_terminated_length": 689.0,
      "completions/mean_length": 253.421875,
      "completions/mean_terminated_length": 254.4156951904297,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 71.0,
      "epoch": 0.032,
      "grad_norm": 0.009619243443012238,
      "learning_rate": 4.722222222222222e-06,
      "loss": -0.0326,
      "num_tokens": 6725374.0,
      "reward": 0.9973111152648926,
      "reward_std": 0.4718785881996155,
      "rewards/accuracy_reward_step": 0.2578125,
      "rewards/final_brier_reward_step": 0.5024347305297852,
      "rewards/format_reward_step": 0.9765625,
      "step": 30
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5776629704301075,
      "calib/avg_num_step_conf": 4.68359375,
      "calib/ece": 0.4586614173228346,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.1377952755905512,
      "calib/gap": 0.06506048387096774,
      "calib/mean_conf": 0.702755905511811,
      "calib/mu_c": 0.7519354838709679,
      "calib/mu_w": 0.6868750000000001,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4586614173228346,
      "calib/std_conf": 0.2185931292742577,
      "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": 1262.0,
      "completions/max_terminated_length": 1262.0,
      "completions/mean_length": 255.47265625,
      "completions/mean_terminated_length": 256.4745178222656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 83.0,
      "epoch": 0.03306666666666667,
      "grad_norm": 0.008545058779418468,
      "learning_rate": 4.694444444444445e-06,
      "loss": 0.0224,
      "num_tokens": 6896687.0,
      "reward": 1.0230578184127808,
      "reward_std": 0.3965718150138855,
      "rewards/accuracy_reward_step": 0.2421875,
      "rewards/final_brier_reward_step": 0.5734593868255615,
      "rewards/format_reward_step": 0.98828125,
      "step": 31
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5439977621483376,
      "calib/avg_num_step_conf": 4.171875,
      "calib/ece": 0.3516666666666667,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.047619047619047616,
      "calib/gap": 0.03894820971867008,
      "calib/mean_conf": 0.5952380952380952,
      "calib/mu_c": 0.6236764705882353,
      "calib/mu_w": 0.5847282608695652,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3385317460317461,
      "calib/std_conf": 0.22601522032507923,
      "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": 2092.0,
      "completions/max_terminated_length": 2092.0,
      "completions/mean_length": 264.09375,
      "completions/mean_terminated_length": 264.09375,
      "completions/min_length": 65.0,
      "completions/min_terminated_length": 65.0,
      "epoch": 0.034133333333333335,
      "grad_norm": 0.009222334250807762,
      "learning_rate": 4.666666666666667e-06,
      "loss": 0.0659,
      "num_tokens": 7070999.0,
      "reward": 1.0783140659332275,
      "reward_std": 0.35461053252220154,
      "rewards/accuracy_reward_step": 0.265625,
      "rewards/final_brier_reward_step": 0.6488156318664551,
      "rewards/format_reward_step": 0.9765625,
      "step": 32
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5759622175141242,
      "calib/avg_num_step_conf": 4.27734375,
      "calib/ece": 0.3166135458167331,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.03187250996015936,
      "calib/gap": 0.06111670197740127,
      "calib/mean_conf": 0.5508764940239043,
      "calib/mu_c": 0.5976271186440679,
      "calib/mu_w": 0.5365104166666667,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3162151394422312,
      "calib/std_conf": 0.22650652938457116,
      "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": 2500.0,
      "completions/max_terminated_length": 2500.0,
      "completions/mean_length": 265.62109375,
      "completions/mean_terminated_length": 266.6627502441406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 52.0,
      "epoch": 0.0352,
      "grad_norm": 0.007931312546133995,
      "learning_rate": 4.638888888888889e-06,
      "loss": 0.0692,
      "num_tokens": 7245870.0,
      "reward": 1.0570218563079834,
      "reward_std": 0.3615739941596985,
      "rewards/accuracy_reward_step": 0.23046875,
      "rewards/final_brier_reward_step": 0.6765437722206116,
      "rewards/format_reward_step": 0.9765625,
      "step": 33
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5165995975855131,
      "calib/avg_num_step_conf": 3.6171875,
      "calib/ece": 0.20731225296442687,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.011857707509881422,
      "calib/gap": 0.008900325027085654,
      "calib/mean_conf": 0.4113438735177865,
      "calib/mu_c": 0.41774647887323946,
      "calib/mu_w": 0.4088461538461538,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1690118577075099,
      "calib/std_conf": 0.21953727013038177,
      "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": 1995.0,
      "completions/max_terminated_length": 1995.0,
      "completions/mean_length": 207.265625,
      "completions/mean_terminated_length": 208.0784454345703,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 51.0,
      "epoch": 0.03626666666666667,
      "grad_norm": 0.012170094065368176,
      "learning_rate": 4.611111111111112e-06,
      "loss": 0.0295,
      "num_tokens": 7404042.0,
      "reward": 1.1278128623962402,
      "reward_std": 0.3320312201976776,
      "rewards/accuracy_reward_step": 0.27734375,
      "rewards/final_brier_reward_step": 0.7204695343971252,
      "rewards/format_reward_step": 0.98046875,
      "step": 34
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5160454621427377,
      "calib/avg_num_step_conf": 4.52734375,
      "calib/ece": 0.15588235294117647,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.00784313725490196,
      "calib/gap": 0.0016839378238341918,
      "calib/mean_conf": 0.3437254901960784,
      "calib/mu_c": 0.34500000000000003,
      "calib/mu_w": 0.34331606217616584,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.12823529411764706,
      "calib/std_conf": 0.19647641962306053,
      "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": 1309.0,
      "completions/max_terminated_length": 1309.0,
      "completions/mean_length": 252.0,
      "completions/mean_terminated_length": 252.98825073242188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 73.0,
      "epoch": 0.037333333333333336,
      "grad_norm": 0.009190870448946953,
      "learning_rate": 4.583333333333333e-06,
      "loss": 0.0373,
      "num_tokens": 7577810.0,
      "reward": 1.1189427375793457,
      "reward_std": 0.3060336709022522,
      "rewards/accuracy_reward_step": 0.2421875,
      "rewards/final_brier_reward_step": 0.7613230347633362,
      "rewards/format_reward_step": 0.9921875,
      "step": 35
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5140435222672064,
      "calib/avg_num_step_conf": 4.34765625,
      "calib/ece": 0.193125,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.007449392712550573,
      "calib/mean_conf": 0.28250000000000003,
      "calib/mu_c": 0.2869230769230769,
      "calib/mu_w": 0.2794736842105263,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.034687499999999996,
      "calib/std_conf": 0.17630761469658648,
      "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": 1185.0,
      "completions/max_terminated_length": 1185.0,
      "completions/mean_length": 245.37890625,
      "completions/mean_terminated_length": 246.3411865234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 47.0,
      "epoch": 0.0384,
      "grad_norm": 0.009832658804953098,
      "learning_rate": 4.555555555555556e-06,
      "loss": -0.0565,
      "num_tokens": 7743339.0,
      "reward": 1.2605117559432983,
      "reward_std": 0.3142458498477936,
      "rewards/accuracy_reward_step": 0.40625,
      "rewards/final_brier_reward_step": 0.7124297022819519,
      "rewards/format_reward_step": 0.99609375,
      "step": 36
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6023396596858639,
      "calib/avg_num_step_conf": 4.08984375,
      "calib/ece": 0.12815686274509808,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.04336142015706804,
      "calib/mean_conf": 0.18549019607843137,
      "calib/mu_c": 0.21796875,
      "calib/mu_w": 0.17460732984293195,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.03133333333333335,
      "calib/std_conf": 0.15180113055315647,
      "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": 2324.0,
      "completions/max_terminated_length": 2324.0,
      "completions/mean_length": 233.12890625,
      "completions/mean_terminated_length": 233.12890625,
      "completions/min_length": 46.0,
      "completions/min_terminated_length": 46.0,
      "epoch": 0.039466666666666664,
      "grad_norm": 0.010432302951812744,
      "learning_rate": 4.527777777777778e-06,
      "loss": -0.0212,
      "num_tokens": 7910116.0,
      "reward": 1.1431448459625244,
      "reward_std": 0.22751232981681824,
      "rewards/accuracy_reward_step": 0.25,
      "rewards/final_brier_reward_step": 0.7941023707389832,
      "rewards/format_reward_step": 0.9921875,
      "step": 37
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5711820231051001,
      "calib/avg_num_step_conf": 4.4921875,
      "calib/ece": 0.20245059288537554,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.020861510284587215,
      "calib/mean_conf": 0.1416600790513834,
      "calib/mu_c": 0.1555952380952381,
      "calib/mu_w": 0.13473372781065088,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.006047430830039525,
      "calib/std_conf": 0.12258582162401513,
      "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": 1343.0,
      "completions/max_terminated_length": 1343.0,
      "completions/mean_length": 250.33984375,
      "completions/mean_terminated_length": 251.3215789794922,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 61.0,
      "epoch": 0.04053333333333333,
      "grad_norm": 0.010481895878911018,
      "learning_rate": 4.5e-06,
      "loss": 0.0221,
      "num_tokens": 8081091.0,
      "reward": 1.1860566139221191,
      "reward_std": 0.24294333159923553,
      "rewards/accuracy_reward_step": 0.328125,
      "rewards/final_brier_reward_step": 0.7275820374488831,
      "rewards/format_reward_step": 0.98828125,
      "step": 38
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5060984624482554,
      "calib/avg_num_step_conf": 4.95703125,
      "calib/ece": 0.22846456692913386,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.010470136014192832,
      "calib/mean_conf": 0.12681102362204724,
      "calib/mu_c": 0.11947368421052629,
      "calib/mu_w": 0.12994382022471912,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.02803149606299212,
      "calib/std_conf": 0.12266587562901954,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2127.0,
      "completions/max_terminated_length": 2127.0,
      "completions/mean_length": 297.2890625,
      "completions/mean_terminated_length": 297.2890625,
      "completions/min_length": 56.0,
      "completions/min_terminated_length": 56.0,
      "epoch": 0.0416,
      "grad_norm": 0.008426911197602749,
      "learning_rate": 4.472222222222223e-06,
      "loss": 0.0851,
      "num_tokens": 8263285.0,
      "reward": 1.1606513261795044,
      "reward_std": 0.21219465136528015,
      "rewards/accuracy_reward_step": 0.296875,
      "rewards/final_brier_reward_step": 0.7353652119636536,
      "rewards/format_reward_step": 0.9921875,
      "step": 39
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5223880597014925,
      "calib/avg_num_step_conf": 5.953125,
      "calib/ece": 0.16732421875,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.008548132354102483,
      "calib/mean_conf": 0.10697265624999999,
      "calib/mu_c": 0.11328358208955223,
      "calib/mu_w": 0.10473544973544975,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.006289062499999998,
      "calib/std_conf": 0.09797318687308959,
      "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": 1023.0,
      "completions/max_terminated_length": 1023.0,
      "completions/mean_length": 325.19921875,
      "completions/mean_terminated_length": 326.4745178222656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 83.0,
      "epoch": 0.042666666666666665,
      "grad_norm": 0.007824345491826534,
      "learning_rate": 4.444444444444444e-06,
      "loss": -0.0146,
      "num_tokens": 8453296.0,
      "reward": 1.144127368927002,
      "reward_std": 0.20972102880477905,
      "rewards/accuracy_reward_step": 0.26171875,
      "rewards/final_brier_reward_step": 0.772629976272583,
      "rewards/format_reward_step": 0.9921875,
      "step": 40
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6262403870007442,
      "calib/avg_num_step_conf": 6.08984375,
      "calib/ece": 0.4732745098039216,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.025394753162986866,
      "calib/mean_conf": 0.08319607843137257,
      "calib/mu_c": 0.09474820143884893,
      "calib/mu_w": 0.06935344827586207,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.005686274509803921,
      "calib/std_conf": 0.09673331433517804,
      "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": 2434.0,
      "completions/max_terminated_length": 2434.0,
      "completions/mean_length": 321.4765625,
      "completions/mean_terminated_length": 321.4765625,
      "completions/min_length": 90.0,
      "completions/min_terminated_length": 90.0,
      "epoch": 0.04373333333333333,
      "grad_norm": 0.008975100703537464,
      "learning_rate": 4.416666666666667e-06,
      "loss": 0.0198,
      "num_tokens": 8642842.0,
      "reward": 1.3000850677490234,
      "reward_std": 0.22270143032073975,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.5298577547073364,
      "rewards/format_reward_step": 0.984375,
      "step": 41
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.58271484375,
      "calib/avg_num_step_conf": 5.83984375,
      "calib/ece": 0.33187500000000003,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009666666666666664,
      "calib/mean_conf": 0.043125,
      "calib/mu_c": 0.049166666666666664,
      "calib/mu_w": 0.0395,
      "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": 0.05222771654016668,
      "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": 736.0,
      "completions/max_terminated_length": 736.0,
      "completions/mean_length": 289.89453125,
      "completions/mean_terminated_length": 291.0314025878906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 67.0,
      "epoch": 0.0448,
      "grad_norm": 0.008974535390734673,
      "learning_rate": 4.388888888888889e-06,
      "loss": 0.0092,
      "num_tokens": 8821423.0,
      "reward": 1.195831298828125,
      "reward_std": 0.17560523748397827,
      "rewards/accuracy_reward_step": 0.375,
      "rewards/final_brier_reward_step": 0.6494749784469604,
      "rewards/format_reward_step": 0.9921875,
      "step": 42
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5491141219385096,
      "calib/avg_num_step_conf": 6.79296875,
      "calib/ece": 0.3434802371541501,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0145904442417926,
      "calib/mean_conf": 0.055729249011857704,
      "calib/mu_c": 0.0644950495049505,
      "calib/mu_w": 0.049904605263157896,
      "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": 0.0754921063213057,
      "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": 2542.0,
      "completions/max_terminated_length": 2542.0,
      "completions/mean_length": 380.34765625,
      "completions/mean_terminated_length": 381.8392333984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 85.0,
      "epoch": 0.04586666666666667,
      "grad_norm": 0.00716799171641469,
      "learning_rate": 4.361111111111112e-06,
      "loss": 0.0194,
      "num_tokens": 9024016.0,
      "reward": 1.2066413164138794,
      "reward_std": 0.23286837339401245,
      "rewards/accuracy_reward_step": 0.39453125,
      "rewards/final_brier_reward_step": 0.6359390020370483,
      "rewards/format_reward_step": 0.98828125,
      "step": 43
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5405405405405406,
      "calib/avg_num_step_conf": 8.015625,
      "calib/ece": 0.2551394422310757,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.007262940907008693,
      "calib/mean_conf": 0.042310756972111556,
      "calib/mu_c": 0.04743243243243243,
      "calib/mu_w": 0.040169491525423734,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0013147410358565737,
      "calib/std_conf": 0.05426288970400109,
      "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": 2002.0,
      "completions/max_terminated_length": 2002.0,
      "completions/mean_length": 422.07421875,
      "completions/mean_terminated_length": 423.72943115234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.046933333333333334,
      "grad_norm": 0.0074686165899038315,
      "learning_rate": 4.333333333333334e-06,
      "loss": 0.0368,
      "num_tokens": 9238387.0,
      "reward": 1.1363898515701294,
      "reward_std": 0.18531301617622375,
      "rewards/accuracy_reward_step": 0.2890625,
      "rewards/final_brier_reward_step": 0.7141859531402588,
      "rewards/format_reward_step": 0.98046875,
      "step": 44
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.541655585106383,
      "calib/avg_num_step_conf": 7.67578125,
      "calib/ece": 0.34565748031496063,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0020385638297872356,
      "calib/mean_conf": 0.02906692913385827,
      "calib/mu_c": 0.030351063829787233,
      "calib/mu_w": 0.028312499999999997,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0023228346456692917,
      "calib/std_conf": 0.045259873512994944,
      "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": 2238.0,
      "completions/max_terminated_length": 2238.0,
      "completions/mean_length": 412.1015625,
      "completions/mean_terminated_length": 412.1015625,
      "completions/min_length": 102.0,
      "completions/min_terminated_length": 102.0,
      "epoch": 0.048,
      "grad_norm": 0.006873466074466705,
      "learning_rate": 4.305555555555556e-06,
      "loss": 0.0854,
      "num_tokens": 9448933.0,
      "reward": 1.185490369796753,
      "reward_std": 0.1912062168121338,
      "rewards/accuracy_reward_step": 0.3671875,
      "rewards/final_brier_reward_step": 0.6444183588027954,
      "rewards/format_reward_step": 0.9921875,
      "step": 45
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5584940677404387,
      "calib/avg_num_step_conf": 9.4140625,
      "calib/ece": 0.39927419354838706,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.006066149665274752,
      "calib/mean_conf": 0.039354838709677424,
      "calib/mu_c": 0.042803738317757016,
      "calib/mu_w": 0.036737588652482264,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003588709677419355,
      "calib/std_conf": 0.07931909764224779,
      "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": 2476.0,
      "completions/max_terminated_length": 2476.0,
      "completions/mean_length": 481.7890625,
      "completions/mean_terminated_length": 481.7890625,
      "completions/min_length": 132.0,
      "completions/min_terminated_length": 132.0,
      "epoch": 0.04906666666666667,
      "grad_norm": 0.0063409628346562386,
      "learning_rate": 4.277777777777778e-06,
      "loss": 0.0921,
      "num_tokens": 9677039.0,
      "reward": 1.1879210472106934,
      "reward_std": 0.23919141292572021,
      "rewards/accuracy_reward_step": 0.41796875,
      "rewards/final_brier_reward_step": 0.5750609040260315,
      "rewards/format_reward_step": 0.96484375,
      "step": 46
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.4665651483050848,
      "calib/avg_num_step_conf": 9.8359375,
      "calib/ece": 0.4530894308943089,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.008414989406779663,
      "calib/mean_conf": 0.036666666666666674,
      "calib/mu_c": 0.032288135593220336,
      "calib/mu_w": 0.040703125,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0050406504065040655,
      "calib/std_conf": 0.05992768903669766,
      "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": 2549.0,
      "completions/max_terminated_length": 2549.0,
      "completions/mean_length": 495.2421875,
      "completions/mean_terminated_length": 499.1417236328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 134.0,
      "epoch": 0.050133333333333335,
      "grad_norm": 0.0063353064469993114,
      "learning_rate": 4.25e-06,
      "loss": 0.1315,
      "num_tokens": 9909797.0,
      "reward": 1.2039176225662231,
      "reward_std": 0.23223577439785004,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.5250226259231567,
      "rewards/format_reward_step": 0.9609375,
      "step": 47
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5106014412416852,
      "calib/avg_num_step_conf": 8.06640625,
      "calib/ece": 0.32361111111111107,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.003968253968253968,
      "calib/gap": 0.011482815964523278,
      "calib/mean_conf": 0.025595238095238095,
      "calib/mu_c": 0.033068181818181816,
      "calib/mu_w": 0.02158536585365854,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.06970489707141789,
      "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": 2494.0,
      "completions/max_terminated_length": 2494.0,
      "completions/mean_length": 440.2578125,
      "completions/mean_terminated_length": 441.9843444824219,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.0512,
      "grad_norm": 0.0075785936787724495,
      "learning_rate": 4.222222222222223e-06,
      "loss": 0.0148,
      "num_tokens": 10126191.0,
      "reward": 1.1649034023284912,
      "reward_std": 0.21381776034832,
      "rewards/accuracy_reward_step": 0.34375,
      "rewards/final_brier_reward_step": 0.6579316854476929,
      "rewards/format_reward_step": 0.984375,
      "step": 48
    },
    {
      "calib/answer_extract_rate": 0.92578125,
      "calib/auroc": 0.5366172316384181,
      "calib/avg_num_step_conf": 9.73046875,
      "calib/ece": 0.481218487394958,
      "calib/final_conf_rate": 0.9296875,
      "calib/format_rate": 0.92578125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.00023022598870057337,
      "calib/mean_conf": 0.02130252100840336,
      "calib/mu_c": 0.021186440677966097,
      "calib/mu_w": 0.02141666666666667,
      "calib/nonempty_final_conf_rate": 0.9296875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003361344537815126,
      "calib/std_conf": 0.038897420497709216,
      "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": 3018.0,
      "completions/max_terminated_length": 3018.0,
      "completions/mean_length": 548.5546875,
      "completions/mean_terminated_length": 550.7059326171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.05226666666666667,
      "grad_norm": 0.0068258256651461124,
      "learning_rate": 4.194444444444445e-06,
      "loss": 0.0895,
      "num_tokens": 10371157.0,
      "reward": 1.165101408958435,
      "reward_std": 0.23141947388648987,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.48254647850990295,
      "rewards/format_reward_step": 0.92578125,
      "step": 49
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5855808486947915,
      "calib/avg_num_step_conf": 10.58203125,
      "calib/ece": 0.4436078431372548,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.00392156862745098,
      "calib/gap": 0.00983607571446244,
      "calib/mean_conf": 0.030901960784313728,
      "calib/mu_c": 0.03618644067796609,
      "calib/mu_w": 0.026350364963503653,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0058823529411764705,
      "calib/std_conf": 0.0949776557627874,
      "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": 2173.0,
      "completions/max_terminated_length": 2173.0,
      "completions/mean_length": 531.93359375,
      "completions/mean_terminated_length": 534.0196533203125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.05333333333333334,
      "grad_norm": 0.005866543855518103,
      "learning_rate": 4.166666666666667e-06,
      "loss": 0.0094,
      "num_tokens": 10612692.0,
      "reward": 1.2332239151000977,
      "reward_std": 0.22173798084259033,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.556291401386261,
      "rewards/format_reward_step": 0.98828125,
      "step": 50
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5589319771137954,
      "calib/avg_num_step_conf": 10.3203125,
      "calib/ece": 0.5042629482071712,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": -0.007488239033693576,
      "calib/mean_conf": 0.029840637450199207,
      "calib/mu_c": 0.02623076923076923,
      "calib/mu_w": 0.03371900826446281,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00808764940239044,
      "calib/std_conf": 0.07733081070862133,
      "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": 2674.0,
      "completions/max_terminated_length": 2674.0,
      "completions/mean_length": 509.7265625,
      "completions/mean_terminated_length": 511.72552490234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.0544,
      "grad_norm": 0.006805317010730505,
      "learning_rate": 4.138888888888889e-06,
      "loss": 0.0505,
      "num_tokens": 10852478.0,
      "reward": 1.244327187538147,
      "reward_std": 0.18637487292289734,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.4925605356693268,
      "rewards/format_reward_step": 0.98046875,
      "step": 51
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5733447488584474,
      "calib/avg_num_step_conf": 8.82421875,
      "calib/ece": 0.5457480314960629,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.012048452562151202,
      "calib/mean_conf": 0.029055118110236217,
      "calib/mu_c": 0.034178082191780826,
      "calib/mu_w": 0.022129629629629624,
      "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": 0.04338861331206069,
      "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": 1351.0,
      "completions/max_terminated_length": 1351.0,
      "completions/mean_length": 467.54296875,
      "completions/mean_terminated_length": 469.3764953613281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 171.0,
      "epoch": 0.055466666666666664,
      "grad_norm": 0.007461534813046455,
      "learning_rate": 4.111111111111111e-06,
      "loss": 0.0133,
      "num_tokens": 11080121.0,
      "reward": 1.2954832315444946,
      "reward_std": 0.18780872225761414,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.4581538736820221,
      "rewards/format_reward_step": 0.9921875,
      "step": 52
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6381730769230769,
      "calib/avg_num_step_conf": 11.16796875,
      "calib/ece": 0.47348,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.024557692307692315,
      "calib/mean_conf": 0.046520000000000006,
      "calib/mu_c": 0.05830769230769231,
      "calib/mu_w": 0.033749999999999995,
      "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": 0.050709857029970024,
      "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": 2969.0,
      "completions/max_terminated_length": 2969.0,
      "completions/mean_length": 561.5625,
      "completions/mean_terminated_length": 565.9842529296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 169.0,
      "epoch": 0.05653333333333333,
      "grad_norm": 0.00657409243285656,
      "learning_rate": 4.083333333333334e-06,
      "loss": -0.0219,
      "num_tokens": 11329705.0,
      "reward": 1.2578158378601074,
      "reward_std": 0.2188633382320404,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.5195378661155701,
      "rewards/format_reward_step": 0.97265625,
      "step": 53
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5264729299363057,
      "calib/avg_num_step_conf": 9.6796875,
      "calib/ece": 0.5624901185770751,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.006783439490445867,
      "calib/mean_conf": 0.07079051383399211,
      "calib/mu_c": 0.06821656050955414,
      "calib/mu_w": 0.07500000000000001,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.006363636363636363,
      "calib/std_conf": 0.07321639099281427,
      "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": 2682.0,
      "completions/max_terminated_length": 2682.0,
      "completions/mean_length": 505.84375,
      "completions/mean_terminated_length": 505.84375,
      "completions/min_length": 131.0,
      "completions/min_terminated_length": 131.0,
      "epoch": 0.0576,
      "grad_norm": 0.006965023465454578,
      "learning_rate": 4.055555555555556e-06,
      "loss": 0.0331,
      "num_tokens": 11565433.0,
      "reward": 1.3283171653747559,
      "reward_std": 0.18125469982624054,
      "rewards/accuracy_reward_step": 0.61328125,
      "rewards/final_brier_reward_step": 0.4456968903541565,
      "rewards/format_reward_step": 0.984375,
      "step": 54
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6126884095634095,
      "calib/avg_num_step_conf": 10.7109375,
      "calib/ece": 0.34039682539682536,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.003968253968253968,
      "calib/gap": 0.011431912681912668,
      "calib/mean_conf": 0.07992063492063492,
      "calib/mu_c": 0.08663461538461538,
      "calib/mu_w": 0.07520270270270271,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003809523809523809,
      "calib/std_conf": 0.06857964934555134,
      "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": 3035.0,
      "completions/max_terminated_length": 3035.0,
      "completions/mean_length": 528.74609375,
      "completions/mean_terminated_length": 532.909423828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.058666666666666666,
      "grad_norm": 0.006629865616559982,
      "learning_rate": 4.027777777777779e-06,
      "loss": 0.047,
      "num_tokens": 11808616.0,
      "reward": 1.2133586406707764,
      "reward_std": 0.22967413067817688,
      "rewards/accuracy_reward_step": 0.40625,
      "rewards/final_brier_reward_step": 0.6337484121322632,
      "rewards/format_reward_step": 0.98046875,
      "step": 55
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.49834107498341074,
      "calib/avg_num_step_conf": 12.3671875,
      "calib/ece": 0.3597165991902834,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0001307232913072276,
      "calib/mean_conf": 0.0908906882591093,
      "calib/mu_c": 0.09081818181818183,
      "calib/mu_w": 0.09094890510948905,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0026315789473684206,
      "calib/std_conf": 0.042110518202119,
      "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": 2970.0,
      "completions/max_terminated_length": 2970.0,
      "completions/mean_length": 607.5078125,
      "completions/mean_terminated_length": 612.2913208007812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 204.0,
      "epoch": 0.05973333333333333,
      "grad_norm": 0.006633771117776632,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0777,
      "num_tokens": 12070978.0,
      "reward": 1.2138700485229492,
      "reward_std": 0.2590479254722595,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.603521466255188,
      "rewards/format_reward_step": 0.96484375,
      "step": 56
    },
    {
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.5082633053221288,
      "calib/avg_num_step_conf": 13.42578125,
      "calib/ece": 0.4905113636363636,
      "calib/final_conf_rate": 0.9453125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.004132231404958678,
      "calib/gap": -0.00479709383753503,
      "calib/mean_conf": 0.09585227272727274,
      "calib/mu_c": 0.09383035714285715,
      "calib/mu_w": 0.09862745098039218,
      "calib/nonempty_final_conf_rate": 0.9453125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003925619834710744,
      "calib/std_conf": 0.06541273226486852,
      "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": 2844.0,
      "completions/max_terminated_length": 2844.0,
      "completions/mean_length": 656.30859375,
      "completions/mean_terminated_length": 661.4763793945312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 186.0,
      "epoch": 0.0608,
      "grad_norm": 0.006253418512642384,
      "learning_rate": 3.972222222222223e-06,
      "loss": 0.1262,
      "num_tokens": 12345785.0,
      "reward": 1.2676377296447754,
      "reward_std": 0.2519448399543762,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.4844941794872284,
      "rewards/format_reward_step": 0.94140625,
      "step": 57
    },
    {
      "calib/answer_extract_rate": 0.94140625,
      "calib/auroc": 0.6332846003898636,
      "calib/avg_num_step_conf": 14.1796875,
      "calib/ece": 0.3548340244813277,
      "calib/final_conf_rate": 0.94140625,
      "calib/format_rate": 0.9375,
      "calib/frac_conf_gt_0.9": 0.004149377593360996,
      "calib/gap": 0.026280283291562248,
      "calib/mean_conf": 0.09910788423236516,
      "calib/mu_c": 0.11361111111111112,
      "calib/mu_w": 0.08733082781954887,
      "calib/nonempty_final_conf_rate": 0.94140625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.002904564315352697,
      "calib/std_conf": 0.07863548954889332,
      "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": 2539.0,
      "completions/max_terminated_length": 2539.0,
      "completions/mean_length": 686.38671875,
      "completions/mean_terminated_length": 694.5256958007812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 188.0,
      "epoch": 0.06186666666666667,
      "grad_norm": 0.0055255042389035225,
      "learning_rate": 3.944444444444445e-06,
      "loss": 0.0364,
      "num_tokens": 12627820.0,
      "reward": 1.1890089511871338,
      "reward_std": 0.30035144090652466,
      "rewards/accuracy_reward_step": 0.421875,
      "rewards/final_brier_reward_step": 0.5967679023742676,
      "rewards/format_reward_step": 0.9375,
      "step": 58
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.537975912975913,
      "calib/avg_num_step_conf": 13.39453125,
      "calib/ece": 0.4355020080321285,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0013247863247863034,
      "calib/mean_conf": 0.10104417670682732,
      "calib/mu_c": 0.10166666666666667,
      "calib/mu_w": 0.10034188034188037,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0032128514056224897,
      "calib/std_conf": 0.05535313844732542,
      "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": 2864.0,
      "completions/max_terminated_length": 2864.0,
      "completions/mean_length": 641.71484375,
      "completions/mean_terminated_length": 646.7677001953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 151.0,
      "epoch": 0.06293333333333333,
      "grad_norm": 0.006042384542524815,
      "learning_rate": 3.916666666666667e-06,
      "loss": 0.0958,
      "num_tokens": 12898347.0,
      "reward": 1.266005516052246,
      "reward_std": 0.27409636974334717,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.5398234128952026,
      "rewards/format_reward_step": 0.9609375,
      "step": 59
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5423945176027949,
      "calib/avg_num_step_conf": 13.0,
      "calib/ece": 0.4001639344262295,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00737704918032786,
      "calib/mean_conf": 0.10147540983606558,
      "calib/mu_c": 0.10516393442622951,
      "calib/mu_w": 0.09778688524590165,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0008196721311475411,
      "calib/std_conf": 0.03207757326855084,
      "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": 2512.0,
      "completions/max_terminated_length": 2512.0,
      "completions/mean_length": 656.50390625,
      "completions/mean_terminated_length": 659.0784912109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 233.0,
      "epoch": 0.064,
      "grad_norm": 0.005965266842395067,
      "learning_rate": 3.88888888888889e-06,
      "loss": 0.1274,
      "num_tokens": 13175268.0,
      "reward": 1.2400319576263428,
      "reward_std": 0.25855231285095215,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.5660015344619751,
      "rewards/format_reward_step": 0.953125,
      "step": 60
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5484985384002126,
      "calib/avg_num_step_conf": 11.09375,
      "calib/ece": 0.46016129032258063,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005378687217645464,
      "calib/mean_conf": 0.1124193548387097,
      "calib/mu_c": 0.11471830985915492,
      "calib/mu_w": 0.10933962264150945,
      "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": 0.0371759323752729,
      "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": 2766.0,
      "completions/max_terminated_length": 2766.0,
      "completions/mean_length": 561.73046875,
      "completions/mean_terminated_length": 563.933349609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 152.0,
      "epoch": 0.06506666666666666,
      "grad_norm": 0.006542930845171213,
      "learning_rate": 3.861111111111112e-06,
      "loss": 0.0477,
      "num_tokens": 13423135.0,
      "reward": 1.2990654706954956,
      "reward_std": 0.21576321125030518,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.5200058817863464,
      "rewards/format_reward_step": 0.9609375,
      "step": 61
    },
    {
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.5646021328958163,
      "calib/avg_num_step_conf": 13.12890625,
      "calib/ece": 0.3220491803278688,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.01139458572600495,
      "calib/mean_conf": 0.11581967213114756,
      "calib/mu_c": 0.12226415094339625,
      "calib/mu_w": 0.1108695652173913,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0017213114754098362,
      "calib/std_conf": 0.053144408338012566,
      "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": 3036.0,
      "completions/max_terminated_length": 3036.0,
      "completions/mean_length": 665.9609375,
      "completions/mean_terminated_length": 671.2047119140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 123.0,
      "epoch": 0.06613333333333334,
      "grad_norm": 0.005913985893130302,
      "learning_rate": 3.833333333333334e-06,
      "loss": 0.094,
      "num_tokens": 13700701.0,
      "reward": 1.1991487741470337,
      "reward_std": 0.30506694316864014,
      "rewards/accuracy_reward_step": 0.4140625,
      "rewards/final_brier_reward_step": 0.6209539175033569,
      "rewards/format_reward_step": 0.94921875,
      "step": 62
    },
    {
      "calib/answer_extract_rate": 0.9453125,
      "calib/auroc": 0.576269422863485,
      "calib/avg_num_step_conf": 13.2421875,
      "calib/ece": 0.4343801652892562,
      "calib/final_conf_rate": 0.9453125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.011179245283018863,
      "calib/mean_conf": 0.12760330578512397,
      "calib/mu_c": 0.1325,
      "calib/mu_w": 0.12132075471698114,
      "calib/nonempty_final_conf_rate": 0.9453125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.047273160721586785,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 2743.0,
      "completions/max_terminated_length": 2743.0,
      "completions/mean_length": 644.63671875,
      "completions/mean_terminated_length": 657.4780883789062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 189.0,
      "epoch": 0.0672,
      "grad_norm": 0.005883402191102505,
      "learning_rate": 3.8055555555555556e-06,
      "loss": 0.0142,
      "num_tokens": 13974368.0,
      "reward": 1.2702515125274658,
      "reward_std": 0.32960256934165955,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.5365968942642212,
      "rewards/format_reward_step": 0.94140625,
      "step": 63
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.524519833674345,
      "calib/avg_num_step_conf": 12.22265625,
      "calib/ece": 0.43112096774193537,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.00032618309022508507,
      "calib/mean_conf": 0.1333951612903226,
      "calib/mu_c": 0.13325179856115107,
      "calib/mu_w": 0.13357798165137616,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0020161290322580645,
      "calib/std_conf": 0.048966136842026345,
      "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": 2339.0,
      "completions/max_terminated_length": 2339.0,
      "completions/mean_length": 603.81640625,
      "completions/mean_terminated_length": 608.5708618164062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 213.0,
      "epoch": 0.06826666666666667,
      "grad_norm": 0.006580499932169914,
      "learning_rate": 3.777777777777778e-06,
      "loss": 0.0234,
      "num_tokens": 14232721.0,
      "reward": 1.3028054237365723,
      "reward_std": 0.22784338891506195,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.5509234666824341,
      "rewards/format_reward_step": 0.96875,
      "step": 64
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5524828549657099,
      "calib/avg_num_step_conf": 10.08984375,
      "calib/ece": 0.36749003984063744,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009918084836169683,
      "calib/mean_conf": 0.13848605577689244,
      "calib/mu_c": 0.14338582677165354,
      "calib/mu_w": 0.13346774193548386,
      "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": 0.04704091214167084,
      "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": 2925.0,
      "completions/max_terminated_length": 2925.0,
      "completions/mean_length": 512.69140625,
      "completions/mean_terminated_length": 516.7283325195312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 132.0,
      "epoch": 0.06933333333333333,
      "grad_norm": 0.007941358722746372,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.0231,
      "num_tokens": 14468994.0,
      "reward": 1.2930680513381958,
      "reward_std": 0.17943525314331055,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.6056671738624573,
      "rewards/format_reward_step": 0.98046875,
      "step": 65
    },
    {
      "calib/answer_extract_rate": 0.9375,
      "calib/auroc": 0.6290690220101147,
      "calib/avg_num_step_conf": 13.484375,
      "calib/ece": 0.2954583333333334,
      "calib/final_conf_rate": 0.9375,
      "calib/format_rate": 0.9296875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.015108625970510714,
      "calib/mean_conf": 0.13204166666666667,
      "calib/mu_c": 0.14079207920792078,
      "calib/mu_w": 0.12568345323741006,
      "calib/nonempty_final_conf_rate": 0.9375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0033333333333333335,
      "calib/std_conf": 0.063039325799236,
      "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": 2800.0,
      "completions/max_terminated_length": 2800.0,
      "completions/mean_length": 675.0078125,
      "completions/mean_terminated_length": 683.0119018554688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 65.0,
      "epoch": 0.0704,
      "grad_norm": 0.006969817914068699,
      "learning_rate": 3.7222222222222225e-06,
      "loss": 0.0674,
      "num_tokens": 14748148.0,
      "reward": 1.1725585460662842,
      "reward_std": 0.265777587890625,
      "rewards/accuracy_reward_step": 0.39453125,
      "rewards/final_brier_reward_step": 0.6263669729232788,
      "rewards/format_reward_step": 0.9296875,
      "step": 66
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6252823048281755,
      "calib/avg_num_step_conf": 11.24609375,
      "calib/ece": 0.3509374999999999,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.024613318684001656,
      "calib/mean_conf": 0.14515625,
      "calib/mu_c": 0.1575590551181102,
      "calib/mu_w": 0.13294573643410854,
      "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": 0.05509203060277866,
      "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": 1709.0,
      "completions/max_terminated_length": 1709.0,
      "completions/mean_length": 601.875,
      "completions/mean_terminated_length": 604.2353515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 227.0,
      "epoch": 0.07146666666666666,
      "grad_norm": 0.006899421103298664,
      "learning_rate": 3.694444444444445e-06,
      "loss": -0.0108,
      "num_tokens": 15007236.0,
      "reward": 1.3141582012176514,
      "reward_std": 0.14711973071098328,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.6361289024353027,
      "rewards/format_reward_step": 1.0,
      "step": 67
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6069020356234096,
      "calib/avg_num_step_conf": 10.9296875,
      "calib/ece": 0.32878486055776895,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.023849236641221405,
      "calib/mean_conf": 0.1493027888446215,
      "calib/mu_c": 0.16175000000000003,
      "calib/mu_w": 0.13790076335877863,
      "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": 0.050258415403945766,
      "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": 2691.0,
      "completions/max_terminated_length": 2691.0,
      "completions/mean_length": 571.80859375,
      "completions/mean_terminated_length": 571.80859375,
      "completions/min_length": 152.0,
      "completions/min_terminated_length": 152.0,
      "epoch": 0.07253333333333334,
      "grad_norm": 0.007070120424032211,
      "learning_rate": 3.6666666666666666e-06,
      "loss": 0.0571,
      "num_tokens": 15257707.0,
      "reward": 1.274619698524475,
      "reward_std": 0.17965587973594666,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6351768374443054,
      "rewards/format_reward_step": 0.9765625,
      "step": 68
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.48980032047331445,
      "calib/avg_num_step_conf": 10.9765625,
      "calib/ece": 0.3248235294117647,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0004690003697769285,
      "calib/mean_conf": 0.1536078431372549,
      "calib/mu_c": 0.15385245901639347,
      "calib/mu_w": 0.15338345864661654,
      "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": 0.057846340888352415,
      "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": 2086.0,
      "completions/max_terminated_length": 2086.0,
      "completions/mean_length": 596.078125,
      "completions/mean_terminated_length": 598.4157104492188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 221.0,
      "epoch": 0.0736,
      "grad_norm": 0.007091444917023182,
      "learning_rate": 3.638888888888889e-06,
      "loss": 0.0157,
      "num_tokens": 15514799.0,
      "reward": 1.2942771911621094,
      "reward_std": 0.23002542555332184,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.6393355131149292,
      "rewards/format_reward_step": 0.99609375,
      "step": 69
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.7273175542406312,
      "calib/avg_num_step_conf": 10.15625,
      "calib/ece": 0.3274898785425101,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.012145748987854251,
      "calib/gap": 0.029769230769230798,
      "calib/mean_conf": 0.17048582995951417,
      "calib/mu_c": 0.18615384615384617,
      "calib/mu_w": 0.15638461538461537,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.012145748987854251,
      "calib/std_conf": 0.10982764746359674,
      "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": 2881.0,
      "completions/max_terminated_length": 2881.0,
      "completions/mean_length": 601.23046875,
      "completions/mean_terminated_length": 603.5882568359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.07466666666666667,
      "grad_norm": 0.006808578036725521,
      "learning_rate": 3.6111111111111115e-06,
      "loss": 0.0656,
      "num_tokens": 15775706.0,
      "reward": 1.258596658706665,
      "reward_std": 0.22801057994365692,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.6382870674133301,
      "rewards/format_reward_step": 0.96484375,
      "step": 70
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6433621366849961,
      "calib/avg_num_step_conf": 10.703125,
      "calib/ece": 0.280875,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0325803875360042,
      "calib/mean_conf": 0.18283467741935486,
      "calib/mu_c": 0.20043859649122808,
      "calib/mu_w": 0.16785820895522388,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0020161290322580645,
      "calib/std_conf": 0.07578663336945762,
      "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": 2464.0,
      "completions/max_terminated_length": 2464.0,
      "completions/mean_length": 605.359375,
      "completions/mean_terminated_length": 614.96826171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 208.0,
      "epoch": 0.07573333333333333,
      "grad_norm": 0.006682004313915968,
      "learning_rate": 3.5833333333333335e-06,
      "loss": 0.0171,
      "num_tokens": 16035086.0,
      "reward": 1.2577886581420898,
      "reward_std": 0.2718082070350647,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.6601086258888245,
      "rewards/format_reward_step": 0.96484375,
      "step": 71
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6123209540475727,
      "calib/avg_num_step_conf": 11.37890625,
      "calib/ece": 0.2582,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.02505347073692399,
      "calib/mean_conf": 0.18580000000000002,
      "calib/mu_c": 0.19972972972972972,
      "calib/mu_w": 0.17467625899280573,
      "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": 0.06568074299214344,
      "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": 2572.0,
      "completions/max_terminated_length": 2572.0,
      "completions/mean_length": 578.66015625,
      "completions/mean_terminated_length": 578.66015625,
      "completions/min_length": 219.0,
      "completions/min_terminated_length": 219.0,
      "epoch": 0.0768,
      "grad_norm": 0.007656267378479242,
      "learning_rate": 3.555555555555556e-06,
      "loss": 0.0642,
      "num_tokens": 16287631.0,
      "reward": 1.260998249053955,
      "reward_std": 0.20352822542190552,
      "rewards/accuracy_reward_step": 0.43359375,
      "rewards/final_brier_reward_step": 0.6782464981079102,
      "rewards/format_reward_step": 0.9765625,
      "step": 72
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6154974659909309,
      "calib/avg_num_step_conf": 10.42578125,
      "calib/ece": 0.44007843137254904,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.025857562016537772,
      "calib/mean_conf": 0.1991372549019608,
      "calib/mu_c": 0.20846625766871169,
      "calib/mu_w": 0.1826086956521739,
      "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": 0.06644502950112682,
      "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": 2272.0,
      "completions/max_terminated_length": 2272.0,
      "completions/mean_length": 562.6796875,
      "completions/mean_terminated_length": 562.6796875,
      "completions/min_length": 176.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.07786666666666667,
      "grad_norm": 0.00714003574103117,
      "learning_rate": 3.5277777777777784e-06,
      "loss": 0.024,
      "num_tokens": 16538709.0,
      "reward": 1.4214353561401367,
      "reward_std": 0.19357150793075562,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.5772457122802734,
      "rewards/format_reward_step": 0.9921875,
      "step": 73
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.48861584107327144,
      "calib/avg_num_step_conf": 11.1015625,
      "calib/ece": 0.34632000000000007,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0007172342621258876,
      "calib/mean_conf": 0.19768,
      "calib/mu_c": 0.1973529411764706,
      "calib/mu_w": 0.19807017543859648,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0,
      "calib/std_conf": 0.06415152063669263,
      "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": 2306.0,
      "completions/max_terminated_length": 2306.0,
      "completions/mean_length": 582.40234375,
      "completions/mean_terminated_length": 584.686279296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 148.0,
      "epoch": 0.07893333333333333,
      "grad_norm": 0.007149734999984503,
      "learning_rate": 3.5e-06,
      "loss": 0.0427,
      "num_tokens": 16791732.0,
      "reward": 1.3170249462127686,
      "reward_std": 0.25684505701065063,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6067058444023132,
      "rewards/format_reward_step": 0.96484375,
      "step": 74
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5926470588235293,
      "calib/avg_num_step_conf": 10.203125,
      "calib/ece": 0.43873015873015864,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.023167862266857964,
      "calib/mean_conf": 0.23587301587301587,
      "calib/mu_c": 0.24341176470588233,
      "calib/mu_w": 0.22024390243902436,
      "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": 0.07881332013427077,
      "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": 529.44140625,
      "completions/mean_terminated_length": 533.6102294921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 211.0,
      "epoch": 0.08,
      "grad_norm": 0.007752260658890009,
      "learning_rate": 3.4722222222222224e-06,
      "loss": 0.0114,
      "num_tokens": 17032021.0,
      "reward": 1.4476063251495361,
      "reward_std": 0.2468341886997223,
      "rewards/accuracy_reward_step": 0.6640625,
      "rewards/final_brier_reward_step": 0.5827124714851379,
      "rewards/format_reward_step": 0.984375,
      "step": 75
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.7110824059353472,
      "calib/avg_num_step_conf": 11.94921875,
      "calib/ece": 0.33240000000000003,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.059536301006889225,
      "calib/mean_conf": 0.26152,
      "calib/mu_c": 0.2858108108108108,
      "calib/mu_w": 0.22627450980392155,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00096,
      "calib/std_conf": 0.08311491803521195,
      "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": 2689.0,
      "completions/max_terminated_length": 2689.0,
      "completions/mean_length": 626.265625,
      "completions/mean_terminated_length": 628.7216186523438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 208.0,
      "epoch": 0.08106666666666666,
      "grad_norm": 0.006362370681017637,
      "learning_rate": 3.444444444444445e-06,
      "loss": 0.0334,
      "num_tokens": 17295401.0,
      "reward": 1.3979976177215576,
      "reward_std": 0.26154667139053345,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.6553703546524048,
      "rewards/format_reward_step": 0.9765625,
      "step": 76
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5586135895676047,
      "calib/avg_num_step_conf": 11.078125,
      "calib/ece": 0.342570281124498,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.010699382292381543,
      "calib/mean_conf": 0.2857028112449799,
      "calib/mu_c": 0.28974193548387095,
      "calib/mu_w": 0.2790425531914894,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.002891566265060241,
      "calib/std_conf": 0.0844709975663121,
      "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": 2685.0,
      "completions/max_terminated_length": 2685.0,
      "completions/mean_length": 566.75,
      "completions/mean_terminated_length": 568.9725952148438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 177.0,
      "epoch": 0.08213333333333334,
      "grad_norm": 0.007629550527781248,
      "learning_rate": 3.416666666666667e-06,
      "loss": 0.0344,
      "num_tokens": 17545153.0,
      "reward": 1.4076530933380127,
      "reward_std": 0.2604719400405884,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.6317124962806702,
      "rewards/format_reward_step": 0.97265625,
      "step": 77
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6027968909276248,
      "calib/avg_num_step_conf": 11.125,
      "calib/ece": 0.2543083003952569,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.020599515800203805,
      "calib/mean_conf": 0.3189723320158102,
      "calib/mu_c": 0.32784722222222223,
      "calib/mu_w": 0.30724770642201843,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0020553359683794467,
      "calib/std_conf": 0.08237417777032076,
      "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": 1963.0,
      "completions/max_terminated_length": 1963.0,
      "completions/mean_length": 627.703125,
      "completions/mean_terminated_length": 630.1647338867188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 250.0,
      "epoch": 0.0832,
      "grad_norm": 0.007345384452491999,
      "learning_rate": 3.3888888888888893e-06,
      "loss": 0.0249,
      "num_tokens": 17813869.0,
      "reward": 1.4042229652404785,
      "reward_std": 0.27890828251838684,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.6873522996902466,
      "rewards/format_reward_step": 0.98828125,
      "step": 78
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6281788079470199,
      "calib/avg_num_step_conf": 11.22265625,
      "calib/ece": 0.24366533864541837,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.032754304635761566,
      "calib/mean_conf": 0.35880478087649403,
      "calib/mu_c": 0.3718543046357615,
      "calib/mu_w": 0.33909999999999996,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00043824701195219337,
      "calib/std_conf": 0.08453862207631224,
      "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": 2549.0,
      "completions/max_terminated_length": 2549.0,
      "completions/mean_length": 620.44140625,
      "completions/mean_terminated_length": 622.8745727539062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 209.0,
      "epoch": 0.08426666666666667,
      "grad_norm": 0.006675400771200657,
      "learning_rate": 3.3611111111111117e-06,
      "loss": 0.0373,
      "num_tokens": 18079078.0,
      "reward": 1.428109884262085,
      "reward_std": 0.26775938272476196,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.6960632801055908,
      "rewards/format_reward_step": 0.98046875,
      "step": 79
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5726342371079214,
      "calib/avg_num_step_conf": 11.06640625,
      "calib/ece": 0.2271713147410359,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.013997873471557731,
      "calib/mean_conf": 0.38342629482071716,
      "calib/mu_c": 0.38894736842105265,
      "calib/mu_w": 0.3749494949494949,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00250996015936255,
      "calib/std_conf": 0.08255514603746796,
      "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": 3008.0,
      "completions/max_terminated_length": 3008.0,
      "completions/mean_length": 583.53515625,
      "completions/mean_terminated_length": 583.53515625,
      "completions/min_length": 233.0,
      "completions/min_terminated_length": 233.0,
      "epoch": 0.08533333333333333,
      "grad_norm": 0.00843469426035881,
      "learning_rate": 3.3333333333333333e-06,
      "loss": 0.0404,
      "num_tokens": 18330623.0,
      "reward": 1.4328678846359253,
      "reward_std": 0.2943243980407715,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.6977671980857849,
      "rewards/format_reward_step": 0.98046875,
      "step": 80
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5654385964912281,
      "calib/avg_num_step_conf": 11.47265625,
      "calib/ece": 0.21689795918367347,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.004081632653061225,
      "calib/gap": 0.02476140350877193,
      "calib/mean_conf": 0.407265306122449,
      "calib/mu_c": 0.41686666666666666,
      "calib/mu_w": 0.39210526315789473,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.005959183673469389,
      "calib/std_conf": 0.08770236668907784,
      "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": 3013.0,
      "completions/max_terminated_length": 3013.0,
      "completions/mean_length": 641.203125,
      "completions/mean_terminated_length": 648.8063354492188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 233.0,
      "epoch": 0.0864,
      "grad_norm": 0.007108001969754696,
      "learning_rate": 3.3055555555555558e-06,
      "loss": 0.0315,
      "num_tokens": 18601019.0,
      "reward": 1.411208152770996,
      "reward_std": 0.2808418571949005,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.6935101747512817,
      "rewards/format_reward_step": 0.95703125,
      "step": 81
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5786701777485188,
      "calib/avg_num_step_conf": 10.796875,
      "calib/ece": 0.18498023715415013,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.030245556287030906,
      "calib/mean_conf": 0.44189723320158103,
      "calib/mu_c": 0.45361290322580644,
      "calib/mu_w": 0.42336734693877554,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.007114624505928853,
      "calib/std_conf": 0.09293819035007303,
      "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": 2519.0,
      "completions/max_terminated_length": 2519.0,
      "completions/mean_length": 587.90234375,
      "completions/mean_terminated_length": 590.2078857421875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 133.0,
      "epoch": 0.08746666666666666,
      "grad_norm": 0.007196605671197176,
      "learning_rate": 3.277777777777778e-06,
      "loss": -0.0045,
      "num_tokens": 18857074.0,
      "reward": 1.4649035930633545,
      "reward_std": 0.2847653031349182,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.7305883169174194,
      "rewards/format_reward_step": 0.98828125,
      "step": 82
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5513020833333333,
      "calib/avg_num_step_conf": 12.9765625,
      "calib/ece": 0.07092741935483868,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.004032258064516129,
      "calib/gap": 0.017000000000000015,
      "calib/mean_conf": 0.5105241935483872,
      "calib/mu_c": 0.51875,
      "calib/mu_w": 0.50175,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.032661290322580645,
      "calib/std_conf": 0.10662021563668943,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 2724.0,
      "completions/max_terminated_length": 2724.0,
      "completions/mean_length": 678.69140625,
      "completions/mean_terminated_length": 692.211181640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 223.0,
      "epoch": 0.08853333333333334,
      "grad_norm": 0.006281640846282244,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.0029,
      "num_tokens": 19138083.0,
      "reward": 1.3463736772537231,
      "reward_std": 0.3007776141166687,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.7239972352981567,
      "rewards/format_reward_step": 0.96875,
      "step": 83
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.635747865891377,
      "calib/avg_num_step_conf": 9.83984375,
      "calib/ece": 0.05466666666666668,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.04765000618582216,
      "calib/mean_conf": 0.5251764705882354,
      "calib/mu_c": 0.5472262773722628,
      "calib/mu_w": 0.49957627118644066,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.021294117647058824,
      "calib/std_conf": 0.084397915103501,
      "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": 2376.0,
      "completions/max_terminated_length": 2376.0,
      "completions/mean_length": 573.56640625,
      "completions/mean_terminated_length": 573.56640625,
      "completions/min_length": 192.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.0896,
      "grad_norm": 0.006773416418582201,
      "learning_rate": 3.2222222222222227e-06,
      "loss": 0.0459,
      "num_tokens": 19390836.0,
      "reward": 1.415609359741211,
      "reward_std": 0.2515171766281128,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.7648124694824219,
      "rewards/format_reward_step": 0.99609375,
      "step": 84
    },
    {
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.4925797503467406,
      "calib/avg_num_step_conf": 10.125,
      "calib/ece": 0.10993827160493824,
      "calib/final_conf_rate": 0.94921875,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.00823045267489712,
      "calib/gap": -0.009399445214979263,
      "calib/mean_conf": 0.5405555555555557,
      "calib/mu_c": 0.5365714285714286,
      "calib/mu_w": 0.5459708737864079,
      "calib/nonempty_final_conf_rate": 0.94921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.03718106995884774,
      "calib/std_conf": 0.08467682450974114,
      "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": 2831.0,
      "completions/max_terminated_length": 2831.0,
      "completions/mean_length": 619.8359375,
      "completions/mean_terminated_length": 629.6746215820312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 241.0,
      "epoch": 0.09066666666666667,
      "grad_norm": 0.00668597687035799,
      "learning_rate": 3.1944444444444443e-06,
      "loss": 0.0884,
      "num_tokens": 19657338.0,
      "reward": 1.3684334754943848,
      "reward_std": 0.34873247146606445,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.7017108201980591,
      "rewards/format_reward_step": 0.94140625,
      "step": 85
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5235327743902439,
      "calib/avg_num_step_conf": 9.296875,
      "calib/ece": 0.04788844621513937,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": 0.00867695630081311,
      "calib/mean_conf": 0.5462948207171315,
      "calib/mu_c": 0.550546875,
      "calib/mu_w": 0.5418699186991869,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.04211155378486048,
      "calib/std_conf": 0.0855694350816165,
      "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": 2709.0,
      "completions/max_terminated_length": 2709.0,
      "completions/mean_length": 613.1640625,
      "completions/mean_terminated_length": 615.5686645507812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.09173333333333333,
      "grad_norm": 0.006188141647726297,
      "learning_rate": 3.1666666666666667e-06,
      "loss": 0.006,
      "num_tokens": 19919820.0,
      "reward": 1.3382601737976074,
      "reward_std": 0.31673046946525574,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.7155828475952148,
      "rewards/format_reward_step": 0.9609375,
      "step": 86
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4794137695978186,
      "calib/avg_num_step_conf": 8.921875,
      "calib/ece": 0.09019762845849802,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.003952569169960474,
      "calib/gap": 0.0007171097477847166,
      "calib/mean_conf": 0.5555731225296443,
      "calib/mu_c": 0.5558282208588957,
      "calib/mu_w": 0.555111111111111,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0007509881422924868,
      "calib/std_conf": 0.0759977609726662,
      "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": 2678.0,
      "completions/max_terminated_length": 2678.0,
      "completions/mean_length": 578.31640625,
      "completions/mean_terminated_length": 580.5843505859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 181.0,
      "epoch": 0.0928,
      "grad_norm": 0.006629766430705786,
      "learning_rate": 3.138888888888889e-06,
      "loss": 0.0131,
      "num_tokens": 20173365.0,
      "reward": 1.5051707029342651,
      "reward_std": 0.2687286138534546,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.7486226558685303,
      "rewards/format_reward_step": 0.98828125,
      "step": 87
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5916537166537167,
      "calib/avg_num_step_conf": 9.8984375,
      "calib/ece": 0.0627888446215139,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": 0.02293900543900529,
      "calib/mean_conf": 0.5833466135458166,
      "calib/mu_c": 0.5932167832167831,
      "calib/mu_w": 0.5702777777777778,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.03820717131474104,
      "calib/std_conf": 0.08415951569583457,
      "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": 2345.0,
      "completions/max_terminated_length": 2345.0,
      "completions/mean_length": 683.08984375,
      "completions/mean_terminated_length": 685.7686767578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 247.0,
      "epoch": 0.09386666666666667,
      "grad_norm": 0.005916224326938391,
      "learning_rate": 3.1111111111111116e-06,
      "loss": 0.0287,
      "num_tokens": 20458084.0,
      "reward": 1.417344093322754,
      "reward_std": 0.3309006094932556,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.7409383058547974,
      "rewards/format_reward_step": 0.9765625,
      "step": 88
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5273050546101093,
      "calib/avg_num_step_conf": 9.9921875,
      "calib/ece": 0.13605577689243026,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.00796812749003984,
      "calib/gap": 0.006394462788925415,
      "calib/mean_conf": 0.610796812749004,
      "calib/mu_c": 0.6140322580645161,
      "calib/mu_w": 0.6076377952755907,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.12641434262948206,
      "calib/std_conf": 0.088092685648366,
      "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": 2773.0,
      "completions/max_terminated_length": 2773.0,
      "completions/mean_length": 666.9140625,
      "completions/mean_terminated_length": 669.5294799804688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 187.0,
      "epoch": 0.09493333333333333,
      "grad_norm": 0.00593179976567626,
      "learning_rate": 3.0833333333333336e-06,
      "loss": 0.0392,
      "num_tokens": 20737702.0,
      "reward": 1.3242459297180176,
      "reward_std": 0.31110048294067383,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.7109917998313904,
      "rewards/format_reward_step": 0.96875,
      "step": 89
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4754514100223169,
      "calib/avg_num_step_conf": 10.07421875,
      "calib/ece": 0.0828174603174604,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.007936507936507936,
      "calib/gap": -0.007033881111787377,
      "calib/mean_conf": 0.6326587301587302,
      "calib/mu_c": 0.630062893081761,
      "calib/mu_w": 0.6370967741935484,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.04226190476190482,
      "calib/std_conf": 0.09127228257604533,
      "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": 2457.0,
      "completions/max_terminated_length": 2457.0,
      "completions/mean_length": 639.89453125,
      "completions/mean_terminated_length": 644.9330444335938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 206.0,
      "epoch": 0.096,
      "grad_norm": 0.006317852530628443,
      "learning_rate": 3.055555555555556e-06,
      "loss": 0.0223,
      "num_tokens": 21004835.0,
      "reward": 1.4782195091247559,
      "reward_std": 0.3107277750968933,
      "rewards/accuracy_reward_step": 0.62109375,
      "rewards/final_brier_reward_step": 0.7376890778541565,
      "rewards/format_reward_step": 0.9765625,
      "step": 90
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4880708929788684,
      "calib/avg_num_step_conf": 9.8046875,
      "calib/ece": 0.09620553359683795,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.02766798418972332,
      "calib/gap": -0.003746421267893596,
      "calib/mean_conf": 0.6722529644268775,
      "calib/mu_c": 0.670920245398773,
      "calib/mu_w": 0.6746666666666666,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.06209486166007907,
      "calib/std_conf": 0.1028029276716013,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 2565.0,
      "completions/max_terminated_length": 2565.0,
      "completions/mean_length": 663.2265625,
      "completions/mean_terminated_length": 668.4487915039062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 343.0,
      "epoch": 0.09706666666666666,
      "grad_norm": 0.005937293637543917,
      "learning_rate": 3.0277777777777776e-06,
      "loss": -0.0122,
      "num_tokens": 21282333.0,
      "reward": 1.4947071075439453,
      "reward_std": 0.3460769057273865,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.7394140362739563,
      "rewards/format_reward_step": 0.9765625,
      "step": 91
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5356643356643357,
      "calib/avg_num_step_conf": 9.9921875,
      "calib/ece": 0.13889328063241108,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.023715415019762844,
      "calib/gap": 0.01579720279720276,
      "calib/mean_conf": 0.7041106719367588,
      "calib/mu_c": 0.7109790209790209,
      "calib/mu_w": 0.6951818181818181,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.13889328063241108,
      "calib/std_conf": 0.10683600711255561,
      "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": 2204.0,
      "completions/max_terminated_length": 2204.0,
      "completions/mean_length": 646.125,
      "completions/mean_terminated_length": 648.6588745117188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 245.0,
      "epoch": 0.09813333333333334,
      "grad_norm": 0.006593839265406132,
      "learning_rate": 3e-06,
      "loss": 0.0095,
      "num_tokens": 21554461.0,
      "reward": 1.4110760688781738,
      "reward_std": 0.37093257904052734,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.7205894589424133,
      "rewards/format_reward_step": 0.984375,
      "step": 92
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.376097057305111,
      "calib/avg_num_step_conf": 9.8671875,
      "calib/ece": 0.19332015810276676,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.043478260869565216,
      "calib/gap": -0.04713538977800724,
      "calib/mean_conf": 0.7196442687747036,
      "calib/mu_c": 0.7002684563758389,
      "calib/mu_w": 0.7474038461538461,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.16201581027667983,
      "calib/std_conf": 0.1155852607674471,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2327.0,
      "completions/max_terminated_length": 2327.0,
      "completions/mean_length": 680.26171875,
      "completions/mean_terminated_length": 680.26171875,
      "completions/min_length": 272.0,
      "completions/min_terminated_length": 272.0,
      "epoch": 0.0992,
      "grad_norm": 0.012477902695536613,
      "learning_rate": 2.9722222222222225e-06,
      "loss": 0.0395,
      "num_tokens": 21834384.0,
      "reward": 1.4113017320632935,
      "reward_std": 0.3995455503463745,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.6858847141265869,
      "rewards/format_reward_step": 0.97265625,
      "step": 93
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.6025943396226414,
      "calib/avg_num_step_conf": 9.1875,
      "calib/ece": 0.15650406504065043,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.024390243902439025,
      "calib/gap": 0.035385444743935346,
      "calib/mean_conf": 0.725609756097561,
      "calib/mu_c": 0.7408571428571429,
      "calib/mu_w": 0.7054716981132075,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.15650406504065043,
      "calib/std_conf": 0.08714203284614376,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 2897.0,
      "completions/max_terminated_length": 2897.0,
      "completions/mean_length": 626.2734375,
      "completions/mean_terminated_length": 636.2142944335938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.10026666666666667,
      "grad_norm": 0.006550987716764212,
      "learning_rate": 2.944444444444445e-06,
      "loss": 0.0304,
      "num_tokens": 22103390.0,
      "reward": 1.3804006576538086,
      "reward_std": 0.38110488653182983,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.7061136960983276,
      "rewards/format_reward_step": 0.953125,
      "step": 94
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5633333333333334,
      "calib/avg_num_step_conf": 10.34375,
      "calib/ece": 0.12207843137254909,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.09803921568627451,
      "calib/gap": 0.023333333333333428,
      "calib/mean_conf": 0.7684313725490197,
      "calib/mu_c": 0.7766666666666667,
      "calib/mu_w": 0.7533333333333333,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.1217254901960785,
      "calib/std_conf": 0.09572371495088593,
      "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": 2429.0,
      "completions/max_terminated_length": 2429.0,
      "completions/mean_length": 664.5703125,
      "completions/mean_terminated_length": 664.5703125,
      "completions/min_length": 159.0,
      "completions/min_terminated_length": 159.0,
      "epoch": 0.10133333333333333,
      "grad_norm": 0.005554381292313337,
      "learning_rate": 2.916666666666667e-06,
      "loss": 0.0502,
      "num_tokens": 22379648.0,
      "reward": 1.5141574144363403,
      "reward_std": 0.3407779932022095,
      "rewards/accuracy_reward_step": 0.64453125,
      "rewards/final_brier_reward_step": 0.7509710788726807,
      "rewards/format_reward_step": 0.98828125,
      "step": 95
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5033395176252319,
      "calib/avg_num_step_conf": 9.828125,
      "calib/ece": 0.0899424603174603,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.09126984126984126,
      "calib/gap": 0.010525974025974105,
      "calib/mean_conf": 0.7712123015873016,
      "calib/mu_c": 0.7744285714285715,
      "calib/mu_w": 0.7639025974025974,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.08335515873015874,
      "calib/std_conf": 0.10216301344872772,
      "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": 2655.0,
      "completions/max_terminated_length": 2655.0,
      "completions/mean_length": 610.5390625,
      "completions/mean_terminated_length": 612.933349609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 40.0,
      "epoch": 0.1024,
      "grad_norm": 0.007186867296695709,
      "learning_rate": 2.888888888888889e-06,
      "loss": -0.0203,
      "num_tokens": 22641762.0,
      "reward": 1.5587904453277588,
      "reward_std": 0.2784029245376587,
      "rewards/accuracy_reward_step": 0.6875,
      "rewards/final_brier_reward_step": 0.7621119618415833,
      "rewards/format_reward_step": 0.98046875,
      "step": 96
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.45097250167672703,
      "calib/avg_num_step_conf": 10.54296875,
      "calib/ece": 0.24398785425101216,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.13765182186234817,
      "calib/gap": -0.023768276324614313,
      "calib/mean_conf": 0.8047165991902834,
      "calib/mu_c": 0.7946126760563381,
      "calib/mu_w": 0.8183809523809524,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2369028340080972,
      "calib/std_conf": 0.08963330971406133,
      "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": 2906.0,
      "completions/max_terminated_length": 2906.0,
      "completions/mean_length": 684.12109375,
      "completions/mean_terminated_length": 684.12109375,
      "completions/min_length": 213.0,
      "completions/min_terminated_length": 213.0,
      "epoch": 0.10346666666666667,
      "grad_norm": 0.005954548250883818,
      "learning_rate": 2.861111111111111e-06,
      "loss": 0.077,
      "num_tokens": 22921969.0,
      "reward": 1.3565677404403687,
      "reward_std": 0.4397056996822357,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6506354808807373,
      "rewards/format_reward_step": 0.953125,
      "step": 97
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4760621229623925,
      "calib/avg_num_step_conf": 9.5390625,
      "calib/ece": 0.22209486166007905,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.09881422924901186,
      "calib/gap": -0.0055044281863689815,
      "calib/mean_conf": 0.8031225296442688,
      "calib/mu_c": 0.8008163265306122,
      "calib/mu_w": 0.8063207547169812,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.22209486166007905,
      "calib/std_conf": 0.07154514872892766,
      "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": 2874.0,
      "completions/max_terminated_length": 2874.0,
      "completions/mean_length": 654.6484375,
      "completions/mean_terminated_length": 657.2156982421875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 169.0,
      "epoch": 0.10453333333333334,
      "grad_norm": 0.005915526766330004,
      "learning_rate": 2.8333333333333335e-06,
      "loss": -0.0004,
      "num_tokens": 23195743.0,
      "reward": 1.4139814376831055,
      "reward_std": 0.31640613079071045,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.6912441253662109,
      "rewards/format_reward_step": 0.98828125,
      "step": 98
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5610806047047657,
      "calib/avg_num_step_conf": 10.80859375,
      "calib/ece": 0.43210483870967736,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.21774193548387097,
      "calib/gap": 0.017038709240051464,
      "calib/mean_conf": 0.8312983870967742,
      "calib/mu_c": 0.8415353535353535,
      "calib/mu_w": 0.824496644295302,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.43210483870967736,
      "calib/std_conf": 0.08582958357555616,
      "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": 2915.0,
      "completions/max_terminated_length": 2915.0,
      "completions/mean_length": 757.18359375,
      "completions/mean_terminated_length": 766.162109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 173.0,
      "epoch": 0.1056,
      "grad_norm": 0.005351560655981302,
      "learning_rate": 2.805555555555556e-06,
      "loss": 0.0249,
      "num_tokens": 23495382.0,
      "reward": 1.1461924314498901,
      "reward_std": 0.4089640974998474,
      "rewards/accuracy_reward_step": 0.38671875,
      "rewards/final_brier_reward_step": 0.5541036128997803,
      "rewards/format_reward_step": 0.96484375,
      "step": 99
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5243972445464983,
      "calib/avg_num_step_conf": 10.3515625,
      "calib/ece": 0.3070498007968127,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.23904382470119523,
      "calib/gap": 0.006796051792320235,
      "calib/mean_conf": 0.8330298804780877,
      "calib/mu_c": 0.8361977611940299,
      "calib/mu_w": 0.8294017094017097,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3031075697211155,
      "calib/std_conf": 0.10237179380531525,
      "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": 2322.0,
      "completions/max_terminated_length": 2322.0,
      "completions/mean_length": 713.4375,
      "completions/mean_terminated_length": 716.2353515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 290.0,
      "epoch": 0.10666666666666667,
      "grad_norm": 0.005961840972304344,
      "learning_rate": 2.7777777777777783e-06,
      "loss": 0.0355,
      "num_tokens": 23785430.0,
      "reward": 1.3258066177368164,
      "reward_std": 0.4145909547805786,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.6320819854736328,
      "rewards/format_reward_step": 0.96484375,
      "step": 100
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5445255942084695,
      "calib/avg_num_step_conf": 12.74609375,
      "calib/ece": 0.354,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.372,
      "calib/gap": 0.013453776667307182,
      "calib/mean_conf": 0.87,
      "calib/mu_c": 0.8765116279069767,
      "calib/mu_w": 0.8630578512396695,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.354,
      "calib/std_conf": 0.08025459488403139,
      "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": 2557.0,
      "completions/max_terminated_length": 2557.0,
      "completions/mean_length": 791.40234375,
      "completions/mean_terminated_length": 794.5059204101562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 211.0,
      "epoch": 0.10773333333333333,
      "grad_norm": 0.005466044880449772,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.031,
      "num_tokens": 24095021.0,
      "reward": 1.2949095964431763,
      "reward_std": 0.42435914278030396,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.6093503832817078,
      "rewards/format_reward_step": 0.97265625,
      "step": 101
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5355297157622739,
      "calib/avg_num_step_conf": 11.546875,
      "calib/ece": 0.19378418972332023,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.34782608695652173,
      "calib/gap": 0.010083304622451772,
      "calib/mean_conf": 0.8606822134387352,
      "calib/mu_c": 0.8639104651162791,
      "calib/mu_w": 0.8538271604938273,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.18731225296442694,
      "calib/std_conf": 0.10375497530735353,
      "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": 2910.0,
      "completions/max_terminated_length": 2910.0,
      "completions/mean_length": 678.75,
      "completions/mean_terminated_length": 678.75,
      "completions/min_length": 215.0,
      "completions/min_terminated_length": 215.0,
      "epoch": 0.1088,
      "grad_norm": 0.00616810005158186,
      "learning_rate": 2.7222222222222224e-06,
      "loss": 0.0061,
      "num_tokens": 24375477.0,
      "reward": 1.5273957252502441,
      "reward_std": 0.347787082195282,
      "rewards/accuracy_reward_step": 0.671875,
      "rewards/final_brier_reward_step": 0.7305728197097778,
      "rewards/format_reward_step": 0.98046875,
      "step": 102
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5376986690786922,
      "calib/avg_num_step_conf": 11.58984375,
      "calib/ece": 0.3079418326693227,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.3784860557768924,
      "calib/gap": 0.015482484817159414,
      "calib/mean_conf": 0.8658828685258965,
      "calib/mu_c": 0.8726063380281688,
      "calib/mu_w": 0.8571238532110094,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.3040438247011952,
      "calib/std_conf": 0.11550548211985302,
      "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": 2639.0,
      "completions/max_terminated_length": 2639.0,
      "completions/mean_length": 799.2421875,
      "completions/mean_terminated_length": 805.535400390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 190.0,
      "epoch": 0.10986666666666667,
      "grad_norm": 0.005662613082677126,
      "learning_rate": 2.6944444444444444e-06,
      "loss": -0.0035,
      "num_tokens": 24684635.0,
      "reward": 1.3617302179336548,
      "reward_std": 0.30840951204299927,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.64142906665802,
      "rewards/format_reward_step": 0.97265625,
      "step": 103
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5894255050505051,
      "calib/avg_num_step_conf": 11.5703125,
      "calib/ece": 0.4127753968253968,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.5198412698412699,
      "calib/gap": 0.02779439393939398,
      "calib/mean_conf": 0.8811293650793651,
      "calib/mu_c": 0.8956883333333334,
      "calib/mu_w": 0.8678939393939394,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.40885714285714286,
      "calib/std_conf": 0.12274708386764051,
      "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": 2462.0,
      "completions/max_terminated_length": 2462.0,
      "completions/mean_length": 698.78515625,
      "completions/mean_terminated_length": 704.2874145507812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 189.0,
      "epoch": 0.11093333333333333,
      "grad_norm": 0.005730877164751291,
      "learning_rate": 2.666666666666667e-06,
      "loss": -0.0146,
      "num_tokens": 24970204.0,
      "reward": 1.2464662790298462,
      "reward_std": 0.3567374646663666,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5749638080596924,
      "rewards/format_reward_step": 0.98046875,
      "step": 104
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.43374999999999997,
      "calib/avg_num_step_conf": 12.66796875,
      "calib/ece": 0.41568119999999986,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.608,
      "calib/gap": 0.01303294871794869,
      "calib/mean_conf": 0.8956812,
      "calib/mu_c": 0.9024583333333334,
      "calib/mu_w": 0.8894253846153847,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.41568119999999986,
      "calib/std_conf": 0.13889352787858764,
      "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": 2826.0,
      "completions/max_terminated_length": 2826.0,
      "completions/mean_length": 752.65234375,
      "completions/mean_terminated_length": 758.5787353515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 262.0,
      "epoch": 0.112,
      "grad_norm": 0.005838092416524887,
      "learning_rate": 2.6388888888888893e-06,
      "loss": 0.0502,
      "num_tokens": 25268643.0,
      "reward": 1.2328240871429443,
      "reward_std": 0.4923657774925232,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5515856742858887,
      "rewards/format_reward_step": 0.9765625,
      "step": 105
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5479192938209332,
      "calib/avg_num_step_conf": 12.51953125,
      "calib/ece": 0.4075412698412699,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.7063492063492064,
      "calib/gap": 0.01662219419924338,
      "calib/mean_conf": 0.9234142857142859,
      "calib/mu_c": 0.9314615384615385,
      "calib/mu_w": 0.9148393442622951,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4075412698412699,
      "calib/std_conf": 0.08077340705938965,
      "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": 2771.0,
      "completions/max_terminated_length": 2771.0,
      "completions/mean_length": 714.91796875,
      "completions/mean_terminated_length": 717.7216186523438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 294.0,
      "epoch": 0.11306666666666666,
      "grad_norm": 0.0060480451211333275,
      "learning_rate": 2.6111111111111113e-06,
      "loss": 0.0325,
      "num_tokens": 25556246.0,
      "reward": 1.2863620519638062,
      "reward_std": 0.2843177318572998,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5766304135322571,
      "rewards/format_reward_step": 0.98046875,
      "step": 106
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.39340350877192987,
      "calib/avg_num_step_conf": 13.11328125,
      "calib/ece": 0.33019795918367345,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.7306122448979592,
      "calib/gap": -0.02555140350877183,
      "calib/mean_conf": 0.9202510204081633,
      "calib/mu_c": 0.9103433333333334,
      "calib/mu_w": 0.9358947368421052,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3191020408163265,
      "calib/std_conf": 0.12222855293269241,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 2945.0,
      "completions/max_terminated_length": 2945.0,
      "completions/mean_length": 729.59375,
      "completions/mean_terminated_length": 741.1746215820312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 270.0,
      "epoch": 0.11413333333333334,
      "grad_norm": 0.005714102182537317,
      "learning_rate": 2.5833333333333337e-06,
      "loss": 0.0317,
      "num_tokens": 25847638.0,
      "reward": 1.3682703971862793,
      "reward_std": 0.5019102096557617,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.6115409135818481,
      "rewards/format_reward_step": 0.953125,
      "step": 107
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.519327731092437,
      "calib/avg_num_step_conf": 15.3046875,
      "calib/ece": 0.2787963562753036,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.8704453441295547,
      "calib/gap": -0.010827341482047226,
      "calib/mean_conf": 0.9477947368421051,
      "calib/mu_c": 0.9444194117647059,
      "calib/mu_w": 0.9552467532467531,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.26916599190283397,
      "calib/std_conf": 0.09214234901727124,
      "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": 2955.0,
      "completions/max_terminated_length": 2955.0,
      "completions/mean_length": 808.41015625,
      "completions/mean_terminated_length": 817.99609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 170.0,
      "epoch": 0.1152,
      "grad_norm": 0.005012467037886381,
      "learning_rate": 2.5555555555555557e-06,
      "loss": 0.0467,
      "num_tokens": 26157823.0,
      "reward": 1.4845738410949707,
      "reward_std": 0.42959311604499817,
      "rewards/accuracy_reward_step": 0.6640625,
      "rewards/final_brier_reward_step": 0.6800853610038757,
      "rewards/format_reward_step": 0.9609375,
      "step": 108
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5711924907456373,
      "calib/avg_num_step_conf": 14.2109375,
      "calib/ece": 0.45188130081300815,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.8699186991869918,
      "calib/gap": 0.023055962453728207,
      "calib/mean_conf": 0.9399479674796747,
      "calib/mu_c": 0.9515696721311475,
      "calib/mu_w": 0.9285137096774193,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.44794715447154476,
      "calib/std_conf": 0.11181130580413126,
      "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": 2734.0,
      "completions/max_terminated_length": 2734.0,
      "completions/mean_length": 782.02734375,
      "completions/mean_terminated_length": 788.18505859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 291.0,
      "epoch": 0.11626666666666667,
      "grad_norm": 0.005501654930412769,
      "learning_rate": 2.5277777777777778e-06,
      "loss": 0.0401,
      "num_tokens": 26462622.0,
      "reward": 1.2221993207931519,
      "reward_std": 0.30708974599838257,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.5303360223770142,
      "rewards/format_reward_step": 0.9609375,
      "step": 109
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.44938884644767,
      "calib/avg_num_step_conf": 11.86328125,
      "calib/ece": 0.42629402390438237,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.8725099601593626,
      "calib/gap": -0.009233906289788707,
      "calib/mean_conf": 0.9443035856573706,
      "calib/mu_c": 0.9399257575757576,
      "calib/mu_w": 0.9491596638655463,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4223505976095617,
      "calib/std_conf": 0.07210492870889737,
      "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": 2666.0,
      "completions/max_terminated_length": 2666.0,
      "completions/mean_length": 708.6640625,
      "completions/mean_terminated_length": 708.6640625,
      "completions/min_length": 230.0,
      "completions/min_terminated_length": 230.0,
      "epoch": 0.11733333333333333,
      "grad_norm": 0.0059098368510603905,
      "learning_rate": 2.5e-06,
      "loss": 0.055,
      "num_tokens": 26748960.0,
      "reward": 1.2810180187225342,
      "reward_std": 0.4634544253349304,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.5542237758636475,
      "rewards/format_reward_step": 0.9765625,
      "step": 110
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.47879353233830846,
      "calib/avg_num_step_conf": 12.7890625,
      "calib/ece": 0.43950433070866135,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.8976377952755905,
      "calib/gap": -0.015294900497512587,
      "calib/mean_conf": 0.9358893700787401,
      "calib/mu_c": 0.9286634328358208,
      "calib/mu_w": 0.9439583333333333,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4239173228346456,
      "calib/std_conf": 0.1372516999493022,
      "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": 2435.0,
      "completions/max_terminated_length": 2435.0,
      "completions/mean_length": 712.34375,
      "completions/mean_terminated_length": 715.1372680664062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 198.0,
      "epoch": 0.1184,
      "grad_norm": 0.005665524862706661,
      "learning_rate": 2.4722222222222226e-06,
      "loss": 0.027,
      "num_tokens": 27038728.0,
      "reward": 1.2901296615600586,
      "reward_std": 0.3616183400154114,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.5490092039108276,
      "rewards/format_reward_step": 0.984375,
      "step": 111
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5591823491239455,
      "calib/avg_num_step_conf": 13.7421875,
      "calib/ece": 0.4081048192771083,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.9076305220883534,
      "calib/gap": 0.035507482154445213,
      "calib/mean_conf": 0.9391240963855422,
      "calib/mu_c": 0.9555231343283582,
      "calib/mu_w": 0.920015652173913,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.4045381526104416,
      "calib/std_conf": 0.13779499754009591,
      "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": 3032.0,
      "completions/max_terminated_length": 3032.0,
      "completions/mean_length": 755.484375,
      "completions/mean_terminated_length": 764.4426879882812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 277.0,
      "epoch": 0.11946666666666667,
      "grad_norm": 0.005625101737678051,
      "learning_rate": 2.4444444444444447e-06,
      "loss": 0.0066,
      "num_tokens": 27340052.0,
      "reward": 1.294386386871338,
      "reward_std": 0.47233307361602783,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.5731478929519653,
      "rewards/format_reward_step": 0.96875,
      "step": 112
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5480958738635641,
      "calib/avg_num_step_conf": 12.6796875,
      "calib/ece": 0.3921555118110236,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.9448818897637795,
      "calib/gap": -0.009733772013478359,
      "calib/mean_conf": 0.9474507874015748,
      "calib/mu_c": 0.9433503401360546,
      "calib/mu_w": 0.9530841121495329,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3804330708661417,
      "calib/std_conf": 0.10971966272798954,
      "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": 2566.0,
      "completions/max_terminated_length": 2566.0,
      "completions/mean_length": 671.3515625,
      "completions/mean_terminated_length": 671.3515625,
      "completions/min_length": 278.0,
      "completions/min_terminated_length": 278.0,
      "epoch": 0.12053333333333334,
      "grad_norm": 0.0059531074948608875,
      "learning_rate": 2.4166666666666667e-06,
      "loss": -0.007,
      "num_tokens": 27617118.0,
      "reward": 1.3696892261505127,
      "reward_std": 0.4173819124698639,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.5987532734870911,
      "rewards/format_reward_step": 0.9921875,
      "step": 113
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5548615676820805,
      "calib/avg_num_step_conf": 12.5859375,
      "calib/ece": 0.2909612000000001,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.948,
      "calib/gap": 0.013789853166776278,
      "calib/mean_conf": 0.9548812,
      "calib/mu_c": 0.9593491124260355,
      "calib/mu_w": 0.9455592592592592,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.28492120000000004,
      "calib/std_conf": 0.09330122317826278,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 2830.0,
      "completions/max_terminated_length": 2830.0,
      "completions/mean_length": 644.24609375,
      "completions/mean_terminated_length": 654.4722900390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 251.0,
      "epoch": 0.1216,
      "grad_norm": 0.005589292850345373,
      "learning_rate": 2.388888888888889e-06,
      "loss": 0.003,
      "num_tokens": 27887069.0,
      "reward": 1.488504409790039,
      "reward_std": 0.39160531759262085,
      "rewards/accuracy_reward_step": 0.66015625,
      "rewards/final_brier_reward_step": 0.6840401887893677,
      "rewards/format_reward_step": 0.97265625,
      "step": 114
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5957170668397144,
      "calib/avg_num_step_conf": 12.88671875,
      "calib/ece": 0.4991056224899598,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.9156626506024096,
      "calib/gap": 0.007179935107073376,
      "calib/mean_conf": 0.9393682730923694,
      "calib/mu_c": 0.9432321739130435,
      "calib/mu_w": 0.9360522388059701,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4883132530120482,
      "calib/std_conf": 0.1402333299352498,
      "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": 2767.0,
      "completions/max_terminated_length": 2767.0,
      "completions/mean_length": 661.30859375,
      "completions/mean_terminated_length": 669.1502075195312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 268.0,
      "epoch": 0.12266666666666666,
      "grad_norm": 0.006005452014505863,
      "learning_rate": 2.361111111111111e-06,
      "loss": 0.013,
      "num_tokens": 28161628.0,
      "reward": 1.1822772026062012,
      "reward_std": 0.46157434582710266,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.4934607148170471,
      "rewards/format_reward_step": 0.97265625,
      "step": 115
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5137931034482758,
      "calib/avg_num_step_conf": 13.109375,
      "calib/ece": 0.4208764940239045,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.9362549800796812,
      "calib/gap": 0.006124521072796929,
      "calib/mean_conf": 0.9587250996015936,
      "calib/mu_c": 0.9615555555555554,
      "calib/mu_w": 0.9554310344827585,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4208764940239045,
      "calib/std_conf": 0.06891278719175344,
      "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": 3012.0,
      "completions/max_terminated_length": 3012.0,
      "completions/mean_length": 712.015625,
      "completions/mean_terminated_length": 720.4585571289062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 184.0,
      "epoch": 0.12373333333333333,
      "grad_norm": 0.005244911182671785,
      "learning_rate": 2.3333333333333336e-06,
      "loss": 0.0089,
      "num_tokens": 28448424.0,
      "reward": 1.2982820272445679,
      "reward_std": 0.3899441361427307,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.5614078044891357,
      "rewards/format_reward_step": 0.98046875,
      "step": 116
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.49435936219827314,
      "calib/avg_num_step_conf": 13.55078125,
      "calib/ece": 0.4792376984126985,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.9603174603174603,
      "calib/gap": -0.00806044620911317,
      "calib/mean_conf": 0.9594765873015872,
      "calib/mu_c": 0.9553504065040651,
      "calib/mu_w": 0.9634108527131783,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4753095238095239,
      "calib/std_conf": 0.06617931904534732,
      "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": 2936.0,
      "completions/max_terminated_length": 2936.0,
      "completions/mean_length": 691.11328125,
      "completions/mean_terminated_length": 691.11328125,
      "completions/min_length": 280.0,
      "completions/min_terminated_length": 280.0,
      "epoch": 0.1248,
      "grad_norm": 0.005855880212038755,
      "learning_rate": 2.305555555555556e-06,
      "loss": 0.0302,
      "num_tokens": 28731949.0,
      "reward": 1.2283642292022705,
      "reward_std": 0.4078208804130554,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.511415958404541,
      "rewards/format_reward_step": 0.984375,
      "step": 117
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.47035748180955395,
      "calib/avg_num_step_conf": 13.6015625,
      "calib/ece": 0.3901968503937007,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.8937007874015748,
      "calib/gap": -0.0076817462828217,
      "calib/mean_conf": 0.9540551181102362,
      "calib/mu_c": 0.9507586206896552,
      "calib/mu_w": 0.9584403669724769,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3866929133858267,
      "calib/std_conf": 0.06832423937669099,
      "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": 2630.0,
      "completions/max_terminated_length": 2630.0,
      "completions/mean_length": 721.875,
      "completions/mean_terminated_length": 721.875,
      "completions/min_length": 282.0,
      "completions/min_terminated_length": 282.0,
      "epoch": 0.12586666666666665,
      "grad_norm": 0.005256311967968941,
      "learning_rate": 2.277777777777778e-06,
      "loss": 0.0107,
      "num_tokens": 29020757.0,
      "reward": 1.3600351810455322,
      "reward_std": 0.31913137435913086,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.5950707197189331,
      "rewards/format_reward_step": 0.9921875,
      "step": 118
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.4731182795698925,
      "calib/avg_num_step_conf": 13.8359375,
      "calib/ece": 0.4785492,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.892,
      "calib/gap": -0.023923758320532618,
      "calib/mean_conf": 0.9348108,
      "calib/mu_c": 0.9227532258064516,
      "calib/mu_w": 0.9466769841269842,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.45868,
      "calib/std_conf": 0.15223729688666968,
      "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": 2573.0,
      "completions/max_terminated_length": 2573.0,
      "completions/mean_length": 711.60546875,
      "completions/mean_terminated_length": 717.2086791992188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 231.0,
      "epoch": 0.12693333333333334,
      "grad_norm": 0.006077486090362072,
      "learning_rate": 2.25e-06,
      "loss": 0.0325,
      "num_tokens": 29307992.0,
      "reward": 1.2276971340179443,
      "reward_std": 0.46560826897621155,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.5100818276405334,
      "rewards/format_reward_step": 0.9765625,
      "step": 119
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5806209150326798,
      "calib/avg_num_step_conf": 12.56640625,
      "calib/ece": 0.34963730158730155,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.8849206349206349,
      "calib/gap": 0.040613098039215756,
      "calib/mean_conf": 0.9286960317460315,
      "calib/mu_c": 0.9451346666666668,
      "calib/mu_w": 0.904521568627451,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.34154761904761904,
      "calib/std_conf": 0.14665715061163162,
      "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": 2417.0,
      "completions/max_terminated_length": 2417.0,
      "completions/mean_length": 671.71484375,
      "completions/mean_terminated_length": 674.3490600585938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 300.0,
      "epoch": 0.128,
      "grad_norm": 0.005706976167857647,
      "learning_rate": 2.222222222222222e-06,
      "loss": 0.0572,
      "num_tokens": 29586639.0,
      "reward": 1.3960473537445068,
      "reward_std": 0.37202006578445435,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.6358448266983032,
      "rewards/format_reward_step": 0.984375,
      "step": 120
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5235796387520525,
      "calib/avg_num_step_conf": 13.55859375,
      "calib/ece": 0.3735147999999999,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.892,
      "calib/gap": 0.012961346469622548,
      "calib/mean_conf": 0.9445652,
      "calib/mu_c": 0.9500089655172415,
      "calib/mu_w": 0.9370476190476189,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3690399999999999,
      "calib/std_conf": 0.0975125310355546,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 2807.0,
      "completions/max_terminated_length": 2807.0,
      "completions/mean_length": 700.7578125,
      "completions/mean_terminated_length": 714.7171630859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 212.0,
      "epoch": 0.12906666666666666,
      "grad_norm": 0.00541492085903883,
      "learning_rate": 2.1944444444444445e-06,
      "loss": -0.0367,
      "num_tokens": 29871089.0,
      "reward": 1.355072259902954,
      "reward_std": 0.44197413325309753,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.6046760678291321,
      "rewards/format_reward_step": 0.97265625,
      "step": 121
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5131178958498546,
      "calib/avg_num_step_conf": 11.46875,
      "calib/ece": 0.33924387351778645,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.8181818181818182,
      "calib/gap": 0.004811994448850387,
      "calib/mean_conf": 0.934503162055336,
      "calib/mu_c": 0.9363480769230771,
      "calib/mu_w": 0.9315360824742267,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.32857312252964416,
      "calib/std_conf": 0.12044349268728828,
      "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": 2472.0,
      "completions/max_terminated_length": 2472.0,
      "completions/mean_length": 666.66015625,
      "completions/mean_terminated_length": 666.66015625,
      "completions/min_length": 282.0,
      "completions/min_terminated_length": 282.0,
      "epoch": 0.13013333333333332,
      "grad_norm": 0.005828322377055883,
      "learning_rate": 2.166666666666667e-06,
      "loss": 0.0381,
      "num_tokens": 30149098.0,
      "reward": 1.4132450819015503,
      "reward_std": 0.28737032413482666,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.6311777234077454,
      "rewards/format_reward_step": 0.9765625,
      "step": 122
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5482349942653243,
      "calib/avg_num_step_conf": 12.73046875,
      "calib/ece": 0.4184075697211155,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.8764940239043825,
      "calib/gap": 0.020026131005479897,
      "calib/mean_conf": 0.9303597609561753,
      "calib/mu_c": 0.9397744360902255,
      "calib/mu_w": 0.9197483050847456,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.40944342629482067,
      "calib/std_conf": 0.13480568224847467,
      "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": 2813.0,
      "completions/max_terminated_length": 2813.0,
      "completions/mean_length": 747.79296875,
      "completions/mean_terminated_length": 753.6810913085938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 333.0,
      "epoch": 0.1312,
      "grad_norm": 0.0056311506778001785,
      "learning_rate": 2.138888888888889e-06,
      "loss": 0.013,
      "num_tokens": 30445821.0,
      "reward": 1.2952358722686768,
      "reward_std": 0.5198529958724976,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.5709406137466431,
      "rewards/format_reward_step": 0.98046875,
      "step": 123
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5005504334663549,
      "calib/avg_num_step_conf": 11.125,
      "calib/ece": 0.2975486274509806,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.9215686274509803,
      "calib/gap": 0.0043435117655155064,
      "calib/mean_conf": 0.9525298039215686,
      "calib/mu_c": 0.953994674556213,
      "calib/mu_w": 0.9496511627906975,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.29366666666666685,
      "calib/std_conf": 0.09058859889070567,
      "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": 2449.0,
      "completions/max_terminated_length": 2449.0,
      "completions/mean_length": 649.234375,
      "completions/mean_terminated_length": 649.234375,
      "completions/min_length": 267.0,
      "completions/min_terminated_length": 267.0,
      "epoch": 0.13226666666666667,
      "grad_norm": 0.0056220246478915215,
      "learning_rate": 2.1111111111111114e-06,
      "loss": 0.007,
      "num_tokens": 30718841.0,
      "reward": 1.4999858140945435,
      "reward_std": 0.350582480430603,
      "rewards/accuracy_reward_step": 0.66015625,
      "rewards/final_brier_reward_step": 0.6835654973983765,
      "rewards/format_reward_step": 0.99609375,
      "step": 124
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4194749581083597,
      "calib/avg_num_step_conf": 11.40234375,
      "calib/ece": 0.4318712598425197,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.8858267716535433,
      "calib/gap": 0.004049258362812558,
      "calib/mean_conf": 0.9476192913385828,
      "calib/mu_c": 0.9495801526717558,
      "calib/mu_w": 0.9455308943089432,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.4318712598425197,
      "calib/std_conf": 0.09168418653680209,
      "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": 671.03515625,
      "completions/mean_terminated_length": 673.6666870117188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 178.0,
      "epoch": 0.13333333333333333,
      "grad_norm": 0.005516288802027702,
      "learning_rate": 2.0833333333333334e-06,
      "loss": -0.0022,
      "num_tokens": 30995434.0,
      "reward": 1.2792880535125732,
      "reward_std": 0.4651259183883667,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.5507635474205017,
      "rewards/format_reward_step": 0.984375,
      "step": 125
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5485616010006255,
      "calib/avg_num_step_conf": 11.75390625,
      "calib/ece": 0.48078458498023713,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.8972332015810277,
      "calib/gap": -0.005174796747967503,
      "calib/mean_conf": 0.9434841897233202,
      "calib/mu_c": 0.9408252032520326,
      "calib/mu_w": 0.9460000000000001,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.4690513833992095,
      "calib/std_conf": 0.12410852698103679,
      "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": 2830.0,
      "completions/max_terminated_length": 2830.0,
      "completions/mean_length": 694.0390625,
      "completions/mean_terminated_length": 694.0390625,
      "completions/min_length": 286.0,
      "completions/min_terminated_length": 286.0,
      "epoch": 0.1344,
      "grad_norm": 0.0055323997512459755,
      "learning_rate": 2.0555555555555555e-06,
      "loss": 0.0193,
      "num_tokens": 31278572.0,
      "reward": 1.2310841083526611,
      "reward_std": 0.33184194564819336,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.5168557167053223,
      "rewards/format_reward_step": 0.984375,
      "step": 126
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.49639408002006774,
      "calib/avg_num_step_conf": 11.26171875,
      "calib/ece": 0.4167588932806323,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.849802371541502,
      "calib/gap": 0.010983945817133023,
      "calib/mean_conf": 0.9410276679841898,
      "calib/mu_c": 0.9461940298507463,
      "calib/mu_w": 0.9352100840336133,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4140711462450592,
      "calib/std_conf": 0.10207981079543622,
      "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": 2512.0,
      "completions/max_terminated_length": 2512.0,
      "completions/mean_length": 605.3984375,
      "completions/mean_terminated_length": 610.1653442382812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.13546666666666668,
      "grad_norm": 0.006298969965428114,
      "learning_rate": 2.027777777777778e-06,
      "loss": -0.0012,
      "num_tokens": 31537226.0,
      "reward": 1.302546501159668,
      "reward_std": 0.4561861753463745,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.5699367523193359,
      "rewards/format_reward_step": 0.98828125,
      "step": 127
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.49363779527559054,
      "calib/avg_num_step_conf": 9.62109375,
      "calib/ece": 0.440357142857143,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.8650793650793651,
      "calib/gap": 0.01874204724409445,
      "calib/mean_conf": 0.9443253968253968,
      "calib/mu_c": 0.9536220472440944,
      "calib/mu_w": 0.9348799999999999,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.440357142857143,
      "calib/std_conf": 0.08828132605259412,
      "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": 2822.0,
      "completions/max_terminated_length": 2822.0,
      "completions/mean_length": 650.828125,
      "completions/mean_terminated_length": 655.9527587890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 204.0,
      "epoch": 0.13653333333333334,
      "grad_norm": 0.0057527353055775166,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0029,
      "num_tokens": 31810502.0,
      "reward": 1.2602683305740356,
      "reward_std": 0.4331348240375519,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5478804707527161,
      "rewards/format_reward_step": 0.98046875,
      "step": 128
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6097074468085107,
      "calib/avg_num_step_conf": 10.48828125,
      "calib/ece": 0.3249862204724409,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.9015748031496063,
      "calib/gap": 0.03441356382978744,
      "calib/mean_conf": 0.9410767716535433,
      "calib/mu_c": 0.9538125000000001,
      "calib/mu_w": 0.9193989361702126,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.31807086614173224,
      "calib/std_conf": 0.134027670091693,
      "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": 2093.0,
      "completions/max_terminated_length": 2093.0,
      "completions/mean_length": 579.73046875,
      "completions/mean_terminated_length": 582.0039672851562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 248.0,
      "epoch": 0.1376,
      "grad_norm": 0.005614932626485825,
      "learning_rate": 1.9722222222222224e-06,
      "loss": -0.0057,
      "num_tokens": 32061297.0,
      "reward": 1.4486931562423706,
      "reward_std": 0.31342941522598267,
      "rewards/accuracy_reward_step": 0.625,
      "rewards/final_brier_reward_step": 0.659105122089386,
      "rewards/format_reward_step": 0.98828125,
      "step": 129
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5679012345679012,
      "calib/avg_num_step_conf": 9.734375,
      "calib/ece": 0.32929296875000014,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.91015625,
      "calib/gap": 0.013321907013396306,
      "calib/mean_conf": 0.95108984375,
      "calib/mu_c": 0.9559814814814814,
      "calib/mu_w": 0.9426595744680851,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.3237851562500001,
      "calib/std_conf": 0.09305150064265802,
      "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": 1411.0,
      "completions/max_terminated_length": 1411.0,
      "completions/mean_length": 583.7109375,
      "completions/mean_terminated_length": 586.0000610351562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 232.0,
      "epoch": 0.13866666666666666,
      "grad_norm": 0.005677498877048492,
      "learning_rate": 1.944444444444445e-06,
      "loss": -0.0185,
      "num_tokens": 32316015.0,
      "reward": 1.4557148218154907,
      "reward_std": 0.33224961161613464,
      "rewards/accuracy_reward_step": 0.6328125,
      "rewards/final_brier_reward_step": 0.6575233936309814,
      "rewards/format_reward_step": 0.98828125,
      "step": 130
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5884824518042511,
      "calib/avg_num_step_conf": 8.578125,
      "calib/ece": 0.47670509803921574,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.8745098039215686,
      "calib/gap": 0.02879873949579803,
      "calib/mean_conf": 0.9433717647058824,
      "calib/mu_c": 0.9587310924369746,
      "calib/mu_w": 0.9299323529411766,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.47670509803921574,
      "calib/std_conf": 0.09973483071460103,
      "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": 2502.0,
      "completions/max_terminated_length": 2502.0,
      "completions/mean_length": 558.2265625,
      "completions/mean_terminated_length": 558.2265625,
      "completions/min_length": 235.0,
      "completions/min_terminated_length": 235.0,
      "epoch": 0.13973333333333332,
      "grad_norm": 0.006135825999081135,
      "learning_rate": 1.916666666666667e-06,
      "loss": 0.0184,
      "num_tokens": 32565129.0,
      "reward": 1.2259846925735474,
      "reward_std": 0.2734970450401306,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5261881947517395,
      "rewards/format_reward_step": 0.99609375,
      "step": 131
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.44361143150070487,
      "calib/avg_num_step_conf": 11.02734375,
      "calib/ece": 0.36406784313725493,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.9058823529411765,
      "calib/gap": -0.0017071895424835004,
      "calib/mean_conf": 0.9563070588235294,
      "calib/mu_c": 0.955624183006536,
      "calib/mu_w": 0.9573313725490195,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3601874509803922,
      "calib/std_conf": 0.08176965848169004,
      "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": 2102.0,
      "completions/max_terminated_length": 2102.0,
      "completions/mean_length": 652.91015625,
      "completions/mean_terminated_length": 652.91015625,
      "completions/min_length": 226.0,
      "completions/min_terminated_length": 226.0,
      "epoch": 0.1408,
      "grad_norm": 0.005297274794429541,
      "learning_rate": 1.888888888888889e-06,
      "loss": 0.0382,
      "num_tokens": 32837866.0,
      "reward": 1.399484395980835,
      "reward_std": 0.4310172200202942,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.6153749227523804,
      "rewards/format_reward_step": 0.98828125,
      "step": 132
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5575564971751412,
      "calib/avg_num_step_conf": 10.48046875,
      "calib/ece": 0.48669360000000006,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.924,
      "calib/gap": 0.010572688751926074,
      "calib/mean_conf": 0.9507744,
      "calib/mu_c": 0.956356779661017,
      "calib/mu_w": 0.9457840909090909,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.48273400000000005,
      "calib/std_conf": 0.10007373953560444,
      "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": 2982.0,
      "completions/max_terminated_length": 2982.0,
      "completions/mean_length": 680.65625,
      "completions/mean_terminated_length": 686.0157470703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 272.0,
      "epoch": 0.14186666666666667,
      "grad_norm": 0.005452999845147133,
      "learning_rate": 1.8611111111111113e-06,
      "loss": 0.0281,
      "num_tokens": 33118458.0,
      "reward": 1.2015693187713623,
      "reward_std": 0.4506909251213074,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.5047012567520142,
      "rewards/format_reward_step": 0.9765625,
      "step": 133
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5606641123882503,
      "calib/avg_num_step_conf": 10.28125,
      "calib/ece": 0.40577091633466145,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.8685258964143426,
      "calib/gap": 0.029709316730523683,
      "calib/mean_conf": 0.935902390438247,
      "calib/mu_c": 0.9496325925925926,
      "calib/mu_w": 0.9199232758620689,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.40191235059760966,
      "calib/std_conf": 0.12729638062483245,
      "calib/step_conf_rate": 0.98046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 2269.0,
      "completions/max_terminated_length": 2269.0,
      "completions/mean_length": 673.69921875,
      "completions/mean_terminated_length": 679.00390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 267.0,
      "epoch": 0.14293333333333333,
      "grad_norm": 0.005468668416142464,
      "learning_rate": 1.8333333333333333e-06,
      "loss": 0.0097,
      "num_tokens": 33399877.0,
      "reward": 1.2890727519989014,
      "reward_std": 0.4803275167942047,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.5664268136024475,
      "rewards/format_reward_step": 0.95703125,
      "step": 134
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5161622879946792,
      "calib/avg_num_step_conf": 10.34765625,
      "calib/ece": 0.35341269841269835,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.9325396825396826,
      "calib/gap": -0.0033269038909209403,
      "calib/mean_conf": 0.9596031746031746,
      "calib/mu_c": 0.9583225806451614,
      "calib/mu_w": 0.9616494845360823,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3489682539682539,
      "calib/std_conf": 0.06291718913249854,
      "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": 2542.0,
      "completions/max_terminated_length": 2542.0,
      "completions/mean_length": 627.140625,
      "completions/mean_terminated_length": 629.6000366210938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 204.0,
      "epoch": 0.144,
      "grad_norm": 0.005314249079674482,
      "learning_rate": 1.8055555555555557e-06,
      "loss": 0.0276,
      "num_tokens": 33666305.0,
      "reward": 1.41217041015625,
      "reward_std": 0.3732190430164337,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.6290280818939209,
      "rewards/format_reward_step": 0.984375,
      "step": 135
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.4833798449612403,
      "calib/avg_num_step_conf": 10.05859375,
      "calib/ece": 0.479782283464567,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.9566929133858267,
      "calib/gap": -0.0059959255813955,
      "calib/mean_conf": 0.9563515748031497,
      "calib/mu_c": 0.9533063999999999,
      "calib/mu_w": 0.9593023255813954,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.47200393700787413,
      "calib/std_conf": 0.10249409613053076,
      "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": 2903.0,
      "completions/max_terminated_length": 2903.0,
      "completions/mean_length": 607.0859375,
      "completions/mean_terminated_length": 607.0859375,
      "completions/min_length": 246.0,
      "completions/min_terminated_length": 246.0,
      "epoch": 0.14506666666666668,
      "grad_norm": 0.005611001048237085,
      "learning_rate": 1.777777777777778e-06,
      "loss": 0.0166,
      "num_tokens": 33930207.0,
      "reward": 1.2385427951812744,
      "reward_std": 0.41841921210289,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.5122418403625488,
      "rewards/format_reward_step": 0.98046875,
      "step": 136
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5514235544964254,
      "calib/avg_num_step_conf": 10.6875,
      "calib/ece": 0.43175889328063244,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.9169960474308301,
      "calib/gap": 0.01304151511350804,
      "calib/mean_conf": 0.9614031620553359,
      "calib/mu_c": 0.9675373134328357,
      "calib/mu_w": 0.9544957983193276,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.43175889328063244,
      "calib/std_conf": 0.0681815432199302,
      "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": 2791.0,
      "completions/max_terminated_length": 2791.0,
      "completions/mean_length": 624.45703125,
      "completions/mean_terminated_length": 624.45703125,
      "completions/min_length": 259.0,
      "completions/min_terminated_length": 259.0,
      "epoch": 0.14613333333333334,
      "grad_norm": 0.004993810784071684,
      "learning_rate": 1.75e-06,
      "loss": 0.0374,
      "num_tokens": 34197052.0,
      "reward": 1.2950917482376099,
      "reward_std": 0.3534542918205261,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.5589334964752197,
      "rewards/format_reward_step": 0.984375,
      "step": 137
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5824561403508771,
      "calib/avg_num_step_conf": 9.80078125,
      "calib/ece": 0.4074015748031496,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.9291338582677166,
      "calib/gap": 0.026552631578947383,
      "calib/mean_conf": 0.9585826771653543,
      "calib/mu_c": 0.9705,
      "calib/mu_w": 0.9439473684210526,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4074015748031496,
      "calib/std_conf": 0.08088194380563357,
      "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": 1978.0,
      "completions/max_terminated_length": 1978.0,
      "completions/mean_length": 596.3125,
      "completions/mean_terminated_length": 596.3125,
      "completions/min_length": 213.0,
      "completions/min_terminated_length": 213.0,
      "epoch": 0.1472,
      "grad_norm": 0.007067054510116577,
      "learning_rate": 1.7222222222222224e-06,
      "loss": 0.0312,
      "num_tokens": 34454044.0,
      "reward": 1.341176986694336,
      "reward_std": 0.48980024456977844,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.5886040925979614,
      "rewards/format_reward_step": 0.9921875,
      "step": 138
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.554967105263158,
      "calib/avg_num_step_conf": 9.2265625,
      "calib/ece": 0.3339145098039215,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.9647058823529412,
      "calib/gap": 0.028209736842105593,
      "calib/mean_conf": 0.9613654901960783,
      "calib/mu_c": 0.971875,
      "calib/mu_w": 0.9436652631578945,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3339145098039215,
      "calib/std_conf": 0.08533164312772462,
      "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": 1365.0,
      "completions/max_terminated_length": 1365.0,
      "completions/mean_length": 557.48828125,
      "completions/mean_terminated_length": 559.674560546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 221.0,
      "epoch": 0.14826666666666666,
      "grad_norm": 0.005635618232190609,
      "learning_rate": 1.6944444444444446e-06,
      "loss": 0.0034,
      "num_tokens": 34699857.0,
      "reward": 1.4520823955535889,
      "reward_std": 0.3601229190826416,
      "rewards/accuracy_reward_step": 0.625,
      "rewards/final_brier_reward_step": 0.6580711007118225,
      "rewards/format_reward_step": 0.99609375,
      "step": 139
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5678921568627451,
      "calib/avg_num_step_conf": 9.73046875,
      "calib/ece": 0.3163177865612648,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.9762845849802372,
      "calib/gap": -0.0021319327731089688,
      "calib/mean_conf": 0.9647019762845851,
      "calib/mu_c": 0.9639857142857143,
      "calib/mu_w": 0.9661176470588233,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.30849407114624505,
      "calib/std_conf": 0.08933853961267735,
      "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": 2502.0,
      "completions/max_terminated_length": 2502.0,
      "completions/mean_length": 607.1875,
      "completions/mean_terminated_length": 609.5686645507812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 223.0,
      "epoch": 0.14933333333333335,
      "grad_norm": 0.005778952967375517,
      "learning_rate": 1.6666666666666667e-06,
      "loss": -0.0124,
      "num_tokens": 34960313.0,
      "reward": 1.485205888748169,
      "reward_std": 0.32548052072525024,
      "rewards/accuracy_reward_step": 0.65625,
      "rewards/final_brier_reward_step": 0.6696308255195618,
      "rewards/format_reward_step": 0.98828125,
      "step": 140
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6089225240168636,
      "calib/avg_num_step_conf": 11.28125,
      "calib/ece": 0.33322000000000007,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.956,
      "calib/gap": 0.02745559471974568,
      "calib/mean_conf": 0.9692200000000001,
      "calib/mu_c": 0.9792138364779874,
      "calib/mu_w": 0.9517582417582418,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.33322000000000007,
      "calib/std_conf": 0.061144841156061565,
      "calib/step_conf_rate": 0.9765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 2215.0,
      "completions/max_terminated_length": 2215.0,
      "completions/mean_length": 672.8515625,
      "completions/mean_terminated_length": 680.830078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 255.0,
      "epoch": 0.1504,
      "grad_norm": 0.004524087067693472,
      "learning_rate": 1.638888888888889e-06,
      "loss": -0.0155,
      "num_tokens": 35239659.0,
      "reward": 1.4273971319198608,
      "reward_std": 0.28316599130630493,
      "rewards/accuracy_reward_step": 0.62109375,
      "rewards/final_brier_reward_step": 0.647763192653656,
      "rewards/format_reward_step": 0.96484375,
      "step": 141
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5219184027777778,
      "calib/avg_num_step_conf": 11.35546875,
      "calib/ece": 0.48552637795275594,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.9763779527559056,
      "calib/gap": -0.016247284226190484,
      "calib/mean_conf": 0.9697885826771654,
      "calib/mu_c": 0.9617289062500001,
      "calib/mu_w": 0.9779761904761906,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.4756889763779528,
      "calib/std_conf": 0.09118466460222968,
      "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": 2958.0,
      "completions/max_terminated_length": 2958.0,
      "completions/mean_length": 666.45703125,
      "completions/mean_terminated_length": 669.0706176757812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 244.0,
      "epoch": 0.15146666666666667,
      "grad_norm": 0.005157998763024807,
      "learning_rate": 1.6111111111111113e-06,
      "loss": 0.014,
      "num_tokens": 35515432.0,
      "reward": 1.248449683189392,
      "reward_std": 0.31556376814842224,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.5086181163787842,
      "rewards/format_reward_step": 0.98828125,
      "step": 142
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5193288590604027,
      "calib/avg_num_step_conf": 11.20703125,
      "calib/ece": 0.396444076305221,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.9558232931726908,
      "calib/gap": -0.015604161073825673,
      "calib/mean_conf": 0.9514675702811246,
      "calib/mu_c": 0.9452008389261745,
      "calib/mu_w": 0.9608050000000001,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.37475903614457845,
      "calib/std_conf": 0.16117320310964403,
      "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": 2955.0,
      "completions/max_terminated_length": 2955.0,
      "completions/mean_length": 677.8125,
      "completions/mean_terminated_length": 677.8125,
      "completions/min_length": 287.0,
      "completions/min_terminated_length": 287.0,
      "epoch": 0.15253333333333333,
      "grad_norm": 0.004777104593813419,
      "learning_rate": 1.5833333333333333e-06,
      "loss": 0.07,
      "num_tokens": 35796288.0,
      "reward": 1.3570444583892822,
      "reward_std": 0.4084314703941345,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.5812764763832092,
      "rewards/format_reward_step": 0.96875,
      "step": 143
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5465800273597812,
      "calib/avg_num_step_conf": 11.62890625,
      "calib/ece": 0.3121875000000001,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.98046875,
      "calib/gap": 0.011863201094391207,
      "calib/mean_conf": 0.9762500000000001,
      "calib/mu_c": 0.980235294117647,
      "calib/mu_w": 0.9683720930232558,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3121875000000001,
      "calib/std_conf": 0.058563533875612396,
      "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": 1914.0,
      "completions/max_terminated_length": 1914.0,
      "completions/mean_length": 648.1875,
      "completions/mean_terminated_length": 650.7294311523438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 281.0,
      "epoch": 0.1536,
      "grad_norm": 0.00496349623426795,
      "learning_rate": 1.5555555555555558e-06,
      "loss": -0.0065,
      "num_tokens": 36066352.0,
      "reward": 1.502691388130188,
      "reward_std": 0.3970443606376648,
      "rewards/accuracy_reward_step": 0.6640625,
      "rewards/final_brier_reward_step": 0.6811640858650208,
      "rewards/format_reward_step": 0.99609375,
      "step": 144
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5132053687400778,
      "calib/avg_num_step_conf": 12.39453125,
      "calib/ece": 0.3027482071713149,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.9641434262948207,
      "calib/gap": 0.022921150238129684,
      "calib/mean_conf": 0.9606780876494024,
      "calib/mu_c": 0.9681662721893491,
      "calib/mu_w": 0.9452451219512195,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2950597609561755,
      "calib/std_conf": 0.13028630149214565,
      "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": 2736.0,
      "completions/max_terminated_length": 2736.0,
      "completions/mean_length": 669.22265625,
      "completions/mean_terminated_length": 671.8471069335938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 254.0,
      "epoch": 0.15466666666666667,
      "grad_norm": 0.004919286817312241,
      "learning_rate": 1.527777777777778e-06,
      "loss": 0.0783,
      "num_tokens": 36340377.0,
      "reward": 1.4850218296051025,
      "reward_std": 0.48813188076019287,
      "rewards/accuracy_reward_step": 0.66015625,
      "rewards/final_brier_reward_step": 0.6731687784194946,
      "rewards/format_reward_step": 0.9765625,
      "step": 145
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5369150246305419,
      "calib/avg_num_step_conf": 10.796875,
      "calib/ece": 0.5162109375,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.95703125,
      "calib/gap": 0.0185541871921181,
      "calib/mean_conf": 0.9693359375000001,
      "calib/mu_c": 0.9794827586206896,
      "calib/mu_w": 0.9609285714285715,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.5162109375,
      "calib/std_conf": 0.08117273031009918,
      "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": 1681.0,
      "completions/max_terminated_length": 1681.0,
      "completions/mean_length": 630.99609375,
      "completions/mean_terminated_length": 633.4706420898438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 283.0,
      "epoch": 0.15573333333333333,
      "grad_norm": 0.005344673991203308,
      "learning_rate": 1.5e-06,
      "loss": 0.0048,
      "num_tokens": 36609128.0,
      "reward": 1.1972899436950684,
      "reward_std": 0.4087854027748108,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.48833006620407104,
      "rewards/format_reward_step": 1.0,
      "step": 146
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5084745762711864,
      "calib/avg_num_step_conf": 12.453125,
      "calib/ece": 0.5325149606299213,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.9566929133858267,
      "calib/gap": -0.02970306580259241,
      "calib/mean_conf": 0.9614141732283465,
      "calib/mu_c": 0.9455101694915253,
      "calib/mu_w": 0.9752132352941177,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.5146811023622048,
      "calib/std_conf": 0.13888806274243998,
      "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": 2747.0,
      "completions/max_terminated_length": 2747.0,
      "completions/mean_length": 692.4921875,
      "completions/mean_terminated_length": 692.4921875,
      "completions/min_length": 232.0,
      "completions/min_terminated_length": 232.0,
      "epoch": 0.1568,
      "grad_norm": 0.005198162980377674,
      "learning_rate": 1.4722222222222225e-06,
      "loss": 0.0296,
      "num_tokens": 36890086.0,
      "reward": 1.1903597116470337,
      "reward_std": 0.3408493995666504,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.46665698289871216,
      "rewards/format_reward_step": 0.9921875,
      "step": 147
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5111999423797176,
      "calib/avg_num_step_conf": 10.98828125,
      "calib/ece": 0.2854609375,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.9921875,
      "calib/gap": 0.007035148372227296,
      "calib/mean_conf": 0.9767890625000002,
      "calib/mu_c": 0.9789325842696629,
      "calib/mu_w": 0.9718974358974356,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.28346875,
      "calib/std_conf": 0.06638796393451973,
      "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": 1346.0,
      "completions/max_terminated_length": 1346.0,
      "completions/mean_length": 609.5234375,
      "completions/mean_terminated_length": 611.9137573242188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 196.0,
      "epoch": 0.15786666666666666,
      "grad_norm": 0.006085440516471863,
      "learning_rate": 1.4444444444444445e-06,
      "loss": 0.0038,
      "num_tokens": 37151236.0,
      "reward": 1.5469896793365479,
      "reward_std": 0.3135468661785126,
      "rewards/accuracy_reward_step": 0.6953125,
      "rewards/final_brier_reward_step": 0.7072604894638062,
      "rewards/format_reward_step": 0.99609375,
      "step": 148
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5226355611601512,
      "calib/avg_num_step_conf": 12.81640625,
      "calib/ece": 0.4488531746031745,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.9484126984126984,
      "calib/gap": 0.03717339218158888,
      "calib/mean_conf": 0.9612341269841269,
      "calib/mu_c": 0.9792307692307693,
      "calib/mu_w": 0.9420573770491805,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.44710714285714276,
      "calib/std_conf": 0.12230895674005653,
      "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": 2997.0,
      "completions/max_terminated_length": 2997.0,
      "completions/mean_length": 707.796875,
      "completions/mean_terminated_length": 716.1897583007812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 220.0,
      "epoch": 0.15893333333333334,
      "grad_norm": 0.0053066350519657135,
      "learning_rate": 1.4166666666666667e-06,
      "loss": -0.0465,
      "num_tokens": 37436888.0,
      "reward": 1.2656923532485962,
      "reward_std": 0.3937472105026245,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5391972661018372,
      "rewards/format_reward_step": 0.9765625,
      "step": 149
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5774102079395086,
      "calib/avg_num_step_conf": 11.36328125,
      "calib/ece": 0.43694822134387373,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.9723320158102767,
      "calib/gap": 0.008735362318840756,
      "calib/mean_conf": 0.9613916996047432,
      "calib/mu_c": 0.9653623188405799,
      "calib/mu_w": 0.9566269565217391,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4264426877470358,
      "calib/std_conf": 0.1391820122709795,
      "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": 2024.0,
      "completions/max_terminated_length": 2024.0,
      "completions/mean_length": 616.73046875,
      "completions/mean_terminated_length": 624.0435180664062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 205.0,
      "epoch": 0.16,
      "grad_norm": 0.006326043512672186,
      "learning_rate": 1.3888888888888892e-06,
      "loss": -0.0309,
      "num_tokens": 37699731.0,
      "reward": 1.3119094371795654,
      "reward_std": 0.4123223125934601,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5574127435684204,
      "rewards/format_reward_step": 0.98828125,
      "step": 150
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.551948051948052,
      "calib/avg_num_step_conf": 12.76171875,
      "calib/ece": 0.537176,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.964,
      "calib/gap": 0.006650000000000045,
      "calib/mean_conf": 0.9771760000000002,
      "calib/mu_c": 0.9808999999999999,
      "calib/mu_w": 0.9742499999999998,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.537176,
      "calib/std_conf": 0.03331019399523215,
      "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": 0.0078125,
      "completions/max_length": 2972.0,
      "completions/max_terminated_length": 2972.0,
      "completions/mean_length": 721.48828125,
      "completions/mean_terminated_length": 727.1693115234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 329.0,
      "epoch": 0.16106666666666666,
      "grad_norm": 0.004929741378873587,
      "learning_rate": 1.3611111111111112e-06,
      "loss": 0.0403,
      "num_tokens": 37991456.0,
      "reward": 1.1352100372314453,
      "reward_std": 0.42847442626953125,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.4501076936721802,
      "rewards/format_reward_step": 0.9609375,
      "step": 151
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5288242566204698,
      "calib/avg_num_step_conf": 12.8671875,
      "calib/ece": 0.5419180000000001,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.968,
      "calib/gap": 0.005414177890558758,
      "calib/mean_conf": 0.9701620000000001,
      "calib/mu_c": 0.9732155963302751,
      "calib/mu_w": 0.9678014184397163,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.5380400000000001,
      "calib/std_conf": 0.101737577895289,
      "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": 2504.0,
      "completions/max_terminated_length": 2504.0,
      "completions/mean_length": 694.65234375,
      "completions/mean_terminated_length": 700.1220703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 285.0,
      "epoch": 0.16213333333333332,
      "grad_norm": 0.005465598776936531,
      "learning_rate": 1.3333333333333334e-06,
      "loss": 0.01,
      "num_tokens": 38274679.0,
      "reward": 1.1406099796295166,
      "reward_std": 0.4606872797012329,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.4491886496543884,
      "rewards/format_reward_step": 0.97265625,
      "step": 152
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5483958122255995,
      "calib/avg_num_step_conf": 14.19921875,
      "calib/ece": 0.40882113821138233,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.967479674796748,
      "calib/gap": 0.001979736575481228,
      "calib/mean_conf": 0.9753252032520326,
      "calib/mu_c": 0.9761702127659575,
      "calib/mu_w": 0.9741904761904763,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.405487804878049,
      "calib/std_conf": 0.058104335079578955,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 2963.0,
      "completions/max_terminated_length": 2963.0,
      "completions/mean_length": 742.359375,
      "completions/mean_terminated_length": 760.176025390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 319.0,
      "epoch": 0.1632,
      "grad_norm": 0.005114991217851639,
      "learning_rate": 1.3055555555555556e-06,
      "loss": -0.0446,
      "num_tokens": 38572043.0,
      "reward": 1.3114057779312134,
      "reward_std": 0.4579676389694214,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.5642179250717163,
      "rewards/format_reward_step": 0.95703125,
      "step": 153
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5874609130706692,
      "calib/avg_num_step_conf": 12.41015625,
      "calib/ece": 0.4963003952569171,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.9762845849802372,
      "calib/gap": 0.007604940587867426,
      "calib/mean_conf": 0.973612648221344,
      "calib/mu_c": 0.977520325203252,
      "calib/mu_w": 0.9699153846153846,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.49187351778656135,
      "calib/std_conf": 0.09395912069491512,
      "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": 2010.0,
      "completions/max_terminated_length": 2010.0,
      "completions/mean_length": 664.15625,
      "completions/mean_terminated_length": 669.3858032226562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 300.0,
      "epoch": 0.16426666666666667,
      "grad_norm": 0.005510884802788496,
      "learning_rate": 1.2777777777777779e-06,
      "loss": 0.0088,
      "num_tokens": 38846507.0,
      "reward": 1.2234225273132324,
      "reward_std": 0.4460986852645874,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.5015326738357544,
      "rewards/format_reward_step": 0.984375,
      "step": 154
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5880905511811023,
      "calib/avg_num_step_conf": 11.30078125,
      "calib/ece": 0.4793854901960785,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.9529411764705882,
      "calib/gap": 0.012873252952755654,
      "calib/mean_conf": 0.9646145098039216,
      "calib/mu_c": 0.9710763779527557,
      "calib/mu_w": 0.958203125,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.47298039215686283,
      "calib/std_conf": 0.11337729821375137,
      "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": 1501.0,
      "completions/max_terminated_length": 1501.0,
      "completions/mean_length": 646.44921875,
      "completions/mean_terminated_length": 648.984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 330.0,
      "epoch": 0.16533333333333333,
      "grad_norm": 0.0056763202883303165,
      "learning_rate": 1.25e-06,
      "loss": 0.0139,
      "num_tokens": 39119214.0,
      "reward": 1.2522039413452148,
      "reward_std": 0.3926677703857422,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5200329422950745,
      "rewards/format_reward_step": 0.9921875,
      "step": 155
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.531457916155287,
      "calib/avg_num_step_conf": 12.58984375,
      "calib/ece": 0.4307556,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.956,
      "calib/gap": 0.008647910341709197,
      "calib/mean_conf": 0.9708364000000002,
      "calib/mu_c": 0.9747452554744526,
      "calib/mu_w": 0.9660973451327434,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.426796,
      "calib/std_conf": 0.09248727736851162,
      "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": 2979.0,
      "completions/max_terminated_length": 2979.0,
      "completions/mean_length": 739.82421875,
      "completions/mean_terminated_length": 739.82421875,
      "completions/min_length": 269.0,
      "completions/min_terminated_length": 269.0,
      "epoch": 0.1664,
      "grad_norm": 0.004702582489699125,
      "learning_rate": 1.2222222222222223e-06,
      "loss": 0.0402,
      "num_tokens": 39413369.0,
      "reward": 1.289594292640686,
      "reward_std": 0.3829977810382843,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5440323948860168,
      "rewards/format_reward_step": 0.95703125,
      "step": 156
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5537280701754386,
      "calib/avg_num_step_conf": 14.6171875,
      "calib/ece": 0.3767529880478087,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.9721115537848606,
      "calib/gap": 0.007298644338117999,
      "calib/mean_conf": 0.9724501992031873,
      "calib/mu_c": 0.975328947368421,
      "calib/mu_w": 0.968030303030303,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.37181274900398403,
      "calib/std_conf": 0.08823226608698645,
      "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": 2668.0,
      "completions/max_terminated_length": 2668.0,
      "completions/mean_length": 750.16015625,
      "completions/mean_terminated_length": 750.16015625,
      "completions/min_length": 271.0,
      "completions/min_terminated_length": 271.0,
      "epoch": 0.16746666666666668,
      "grad_norm": 0.004526448901742697,
      "learning_rate": 1.1944444444444446e-06,
      "loss": 0.0226,
      "num_tokens": 39709138.0,
      "reward": 1.3890341520309448,
      "reward_std": 0.38522738218307495,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.6100994944572449,
      "rewards/format_reward_step": 0.98046875,
      "step": 157
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5598376623376624,
      "calib/avg_num_step_conf": 11.703125,
      "calib/ece": 0.3835826771653544,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.9763779527559056,
      "calib/gap": -0.007674025974026,
      "calib/mean_conf": 0.9770472440944882,
      "calib/mu_c": 0.974025974025974,
      "calib/mu_w": 0.9817,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3771653543307087,
      "calib/std_conf": 0.07349233805130645,
      "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": 2600.0,
      "completions/max_terminated_length": 2600.0,
      "completions/mean_length": 687.171875,
      "completions/mean_terminated_length": 687.171875,
      "completions/min_length": 286.0,
      "completions/min_terminated_length": 286.0,
      "epoch": 0.16853333333333334,
      "grad_norm": 0.004906547721475363,
      "learning_rate": 1.1666666666666668e-06,
      "loss": 0.0358,
      "num_tokens": 39990294.0,
      "reward": 1.4026451110839844,
      "reward_std": 0.3692792057991028,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.6099777221679688,
      "rewards/format_reward_step": 0.9921875,
      "step": 158
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5207578358446182,
      "calib/avg_num_step_conf": 12.1953125,
      "calib/ece": 0.41045275590551195,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.9763779527559056,
      "calib/gap": -0.005047682624451766,
      "calib/mean_conf": 0.981003937007874,
      "calib/mu_c": 0.9788775510204082,
      "calib/mu_w": 0.98392523364486,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.40635826771653555,
      "calib/std_conf": 0.03595540742479003,
      "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": 2050.0,
      "completions/max_terminated_length": 2050.0,
      "completions/mean_length": 680.3984375,
      "completions/mean_terminated_length": 683.0667114257812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 265.0,
      "epoch": 0.1696,
      "grad_norm": 0.005360930692404509,
      "learning_rate": 1.138888888888889e-06,
      "loss": -0.0095,
      "num_tokens": 40269260.0,
      "reward": 1.352133870124817,
      "reward_std": 0.32214194536209106,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.5753616094589233,
      "rewards/format_reward_step": 0.98046875,
      "step": 159
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5068873221047134,
      "calib/avg_num_step_conf": 13.7421875,
      "calib/ece": 0.43642369477911636,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.9879518072289156,
      "calib/gap": -0.00706844888366609,
      "calib/mean_conf": 0.981487951807229,
      "calib/mu_c": 0.9783369565217391,
      "calib/mu_w": 0.9854054054054052,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.4318473895582328,
      "calib/std_conf": 0.06343509992924913,
      "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": 2988.0,
      "completions/max_terminated_length": 2988.0,
      "completions/mean_length": 756.81640625,
      "completions/mean_terminated_length": 765.79052734375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 327.0,
      "epoch": 0.17066666666666666,
      "grad_norm": 0.004979386460036039,
      "learning_rate": 1.111111111111111e-06,
      "loss": 0.012,
      "num_tokens": 40567845.0,
      "reward": 1.2913702726364136,
      "reward_std": 0.42069438099861145,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5397717952728271,
      "rewards/format_reward_step": 0.96484375,
      "step": 160
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5072447447447448,
      "calib/avg_num_step_conf": 12.1015625,
      "calib/ece": 0.28397598425196857,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.9645669291338582,
      "calib/gap": -0.0014889339339341223,
      "calib/mean_conf": 0.9723232283464568,
      "calib/mu_c": 0.9718894444444443,
      "calib/mu_w": 0.9733783783783784,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2738188976377953,
      "calib/std_conf": 0.09951728123091752,
      "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": 1905.0,
      "completions/max_terminated_length": 1905.0,
      "completions/mean_length": 668.1484375,
      "completions/mean_terminated_length": 673.409423828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 268.0,
      "epoch": 0.17173333333333332,
      "grad_norm": 0.0053687444888055325,
      "learning_rate": 1.0833333333333335e-06,
      "loss": 0.0337,
      "num_tokens": 40842811.0,
      "reward": 1.5395169258117676,
      "reward_std": 0.32825204730033875,
      "rewards/accuracy_reward_step": 0.703125,
      "rewards/final_brier_reward_step": 0.6962214708328247,
      "rewards/format_reward_step": 0.9765625,
      "step": 161
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4775086505190311,
      "calib/avg_num_step_conf": 13.33984375,
      "calib/ece": 0.315764705882353,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.9764705882352941,
      "calib/gap": 0.016235294117647348,
      "calib/mean_conf": 0.9743529411764705,
      "calib/mu_c": 0.979764705882353,
      "calib/mu_w": 0.9635294117647056,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3117254901960785,
      "calib/std_conf": 0.10062682438329817,
      "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": 1819.0,
      "completions/max_terminated_length": 1819.0,
      "completions/mean_length": 708.5,
      "completions/mean_terminated_length": 711.2785034179688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 328.0,
      "epoch": 0.1728,
      "grad_norm": 0.00477219931781292,
      "learning_rate": 1.0555555555555557e-06,
      "loss": 0.0052,
      "num_tokens": 41128331.0,
      "reward": 1.4949811697006226,
      "reward_std": 0.33206310868263245,
      "rewards/accuracy_reward_step": 0.6640625,
      "rewards/final_brier_reward_step": 0.6735562086105347,
      "rewards/format_reward_step": 0.98828125,
      "step": 162
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6011690647482015,
      "calib/avg_num_step_conf": 13.21484375,
      "calib/ece": 0.4339760956175301,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.9760956175298805,
      "calib/gap": -0.0008663283658788545,
      "calib/mean_conf": 0.9803505976095618,
      "calib/mu_c": 0.9799640287769784,
      "calib/mu_w": 0.9808303571428573,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4302709163346615,
      "calib/std_conf": 0.060573065010961304,
      "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": 743.7265625,
      "completions/mean_terminated_length": 749.5827026367188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 281.0,
      "epoch": 0.17386666666666667,
      "grad_norm": 0.004371706862002611,
      "learning_rate": 1.0277777777777777e-06,
      "loss": 0.0162,
      "num_tokens": 41423557.0,
      "reward": 1.3090941905975342,
      "reward_std": 0.34967905282974243,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.5556883811950684,
      "rewards/format_reward_step": 0.9765625,
      "step": 163
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4665488347735009,
      "calib/avg_num_step_conf": 14.33984375,
      "calib/ece": 0.4535483870967743,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.967741935483871,
      "calib/gap": -0.019934537837130284,
      "calib/mean_conf": 0.9737903225806454,
      "calib/mu_c": 0.9646268656716417,
      "calib/mu_w": 0.984561403508772,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.4435080645161292,
      "calib/std_conf": 0.09169631453824054,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 3043.0,
      "completions/max_terminated_length": 3043.0,
      "completions/mean_length": 778.26953125,
      "completions/mean_terminated_length": 790.623046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 391.0,
      "epoch": 0.17493333333333333,
      "grad_norm": 0.005314142443239689,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0191,
      "num_tokens": 41728930.0,
      "reward": 1.2675137519836426,
      "reward_std": 0.4203830361366272,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.5272148251533508,
      "rewards/format_reward_step": 0.9609375,
      "step": 164
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5550458715596329,
      "calib/avg_num_step_conf": 12.74609375,
      "calib/ece": 0.553452380952381,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.9682539682539683,
      "calib/gap": -0.006286649130685773,
      "calib/mean_conf": 0.967420634920635,
      "calib/mu_c": 0.9638532110091743,
      "calib/mu_w": 0.9701398601398601,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.5441666666666667,
      "calib/std_conf": 0.12179907036777422,
      "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": 3025.0,
      "completions/max_terminated_length": 3025.0,
      "completions/mean_length": 740.8125,
      "completions/mean_terminated_length": 746.6456909179688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 272.0,
      "epoch": 0.176,
      "grad_norm": 0.005232121329754591,
      "learning_rate": 9.722222222222224e-07,
      "loss": 0.0414,
      "num_tokens": 42024154.0,
      "reward": 1.1319527626037598,
      "reward_std": 0.3429035246372223,
      "rewards/accuracy_reward_step": 0.42578125,
      "rewards/final_brier_reward_step": 0.43578046560287476,
      "rewards/format_reward_step": 0.9765625,
      "step": 165
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5238732708612227,
      "calib/avg_num_step_conf": 12.90625,
      "calib/ece": 0.31248097165991906,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.9635627530364372,
      "calib/gap": -0.007145381526104666,
      "calib/mean_conf": 0.9765311740890688,
      "calib/mu_c": 0.9741879518072288,
      "calib/mu_w": 0.9813333333333335,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.30847368421052634,
      "calib/std_conf": 0.06627414997796408,
      "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": 3045.0,
      "completions/max_terminated_length": 3045.0,
      "completions/mean_length": 759.67578125,
      "completions/mean_terminated_length": 768.683837890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 380.0,
      "epoch": 0.17706666666666668,
      "grad_norm": 0.004472343251109123,
      "learning_rate": 9.444444444444445e-07,
      "loss": 0.0544,
      "num_tokens": 42324815.0,
      "reward": 1.4526236057281494,
      "reward_std": 0.3695136606693268,
      "rewards/accuracy_reward_step": 0.6484375,
      "rewards/final_brier_reward_step": 0.6552475094795227,
      "rewards/format_reward_step": 0.953125,
      "step": 166
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5892742453436095,
      "calib/avg_num_step_conf": 13.49609375,
      "calib/ece": 0.2994488188976378,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.9881889763779528,
      "calib/gap": 0.019660315421394348,
      "calib/mean_conf": 0.9805511811023622,
      "calib/mu_c": 0.9868208092485548,
      "calib/mu_w": 0.9671604938271604,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2994488188976378,
      "calib/std_conf": 0.05669099953025043,
      "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": 2116.0,
      "completions/max_terminated_length": 2116.0,
      "completions/mean_length": 740.69140625,
      "completions/mean_terminated_length": 746.5236206054688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 363.0,
      "epoch": 0.17813333333333334,
      "grad_norm": 0.004243503324687481,
      "learning_rate": 9.166666666666666e-07,
      "loss": 0.0064,
      "num_tokens": 42620040.0,
      "reward": 1.5143901109695435,
      "reward_std": 0.30221521854400635,
      "rewards/accuracy_reward_step": 0.67578125,
      "rewards/final_brier_reward_step": 0.692842960357666,
      "rewards/format_reward_step": 0.984375,
      "step": 167
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.4753119913185025,
      "calib/avg_num_step_conf": 13.609375,
      "calib/ece": 0.36682730923694784,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.9839357429718876,
      "calib/gap": 0.015961068909386866,
      "calib/mean_conf": 0.9772690763052209,
      "calib/mu_c": 0.9834868421052633,
      "calib/mu_w": 0.9675257731958764,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.36682730923694784,
      "calib/std_conf": 0.07769632399670272,
      "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": 2553.0,
      "completions/max_terminated_length": 2553.0,
      "completions/mean_length": 782.87109375,
      "completions/mean_terminated_length": 795.2976684570312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 277.0,
      "epoch": 0.1792,
      "grad_norm": 0.004571868572384119,
      "learning_rate": 8.88888888888889e-07,
      "loss": 0.0065,
      "num_tokens": 42925127.0,
      "reward": 1.3821645975112915,
      "reward_std": 0.4252123534679413,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.6080793142318726,
      "rewards/format_reward_step": 0.96875,
      "step": 168
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.573837358684481,
      "calib/avg_num_step_conf": 12.76171875,
      "calib/ece": 0.4312709163346614,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.9362549800796812,
      "calib/gap": 0.004950089928057344,
      "calib/mean_conf": 0.9405537848605579,
      "calib/mu_c": 0.9427625899280574,
      "calib/mu_w": 0.9378125,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.40901992031872514,
      "calib/std_conf": 0.19500897950419732,
      "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": 2873.0,
      "completions/max_terminated_length": 2873.0,
      "completions/mean_length": 767.828125,
      "completions/mean_terminated_length": 773.8740234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 239.0,
      "epoch": 0.18026666666666666,
      "grad_norm": 0.005218811333179474,
      "learning_rate": 8.611111111111112e-07,
      "loss": 0.0088,
      "num_tokens": 43225875.0,
      "reward": 1.3115191459655762,
      "reward_std": 0.36574727296829224,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.5566322207450867,
      "rewards/format_reward_step": 0.98046875,
      "step": 169
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5763374485596708,
      "calib/avg_num_step_conf": 14.75,
      "calib/ece": 0.34105158730158724,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.9801587301587301,
      "calib/gap": 0.009709876543210094,
      "calib/mean_conf": 0.9839087301587301,
      "calib/mu_c": 0.9873765432098767,
      "calib/mu_w": 0.9776666666666666,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.34105158730158724,
      "calib/std_conf": 0.025344836109849323,
      "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": 2817.0,
      "completions/max_terminated_length": 2817.0,
      "completions/mean_length": 787.78125,
      "completions/mean_terminated_length": 793.9842529296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 395.0,
      "epoch": 0.18133333333333335,
      "grad_norm": 0.004583289381116629,
      "learning_rate": 8.333333333333333e-07,
      "loss": -0.0022,
      "num_tokens": 43531699.0,
      "reward": 1.4404544830322266,
      "reward_std": 0.4104905426502228,
      "rewards/accuracy_reward_step": 0.6328125,
      "rewards/final_brier_reward_step": 0.6426275968551636,
      "rewards/format_reward_step": 0.97265625,
      "step": 170
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.563218031968032,
      "calib/avg_num_step_conf": 12.4375,
      "calib/ece": 0.5390980392156863,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.984313725490196,
      "calib/gap": 0.013037587412587648,
      "calib/mean_conf": 0.9783137254901961,
      "calib/mu_c": 0.9856250000000001,
      "calib/mu_w": 0.9725874125874124,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.5390980392156863,
      "calib/std_conf": 0.06584518092333413,
      "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": 1634.0,
      "completions/max_terminated_length": 1634.0,
      "completions/mean_length": 705.03515625,
      "completions/mean_terminated_length": 707.800048828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 312.0,
      "epoch": 0.1824,
      "grad_norm": 0.005645171273499727,
      "learning_rate": 8.055555555555557e-07,
      "loss": 0.0225,
      "num_tokens": 43819084.0,
      "reward": 1.1672158241271973,
      "reward_std": 0.4768272042274475,
      "rewards/accuracy_reward_step": 0.4375,
      "rewards/final_brier_reward_step": 0.46333789825439453,
      "rewards/format_reward_step": 0.99609375,
      "step": 171
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5296686746987951,
      "calib/avg_num_step_conf": 13.9140625,
      "calib/ece": 0.3175487804878051,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.9552845528455285,
      "calib/gap": -0.010208885542168478,
      "calib/mean_conf": 0.9712235772357725,
      "calib/mu_c": 0.9679036144578315,
      "calib/mu_w": 0.9781124999999999,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.30698780487804905,
      "calib/std_conf": 0.09647310386484517,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 2968.0,
      "completions/max_terminated_length": 2968.0,
      "completions/mean_length": 714.68359375,
      "completions/mean_terminated_length": 728.9203491210938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 332.0,
      "epoch": 0.18346666666666667,
      "grad_norm": 0.004878144711256027,
      "learning_rate": 7.777777777777779e-07,
      "loss": 0.0115,
      "num_tokens": 44105395.0,
      "reward": 1.445297360420227,
      "reward_std": 0.3872257471084595,
      "rewards/accuracy_reward_step": 0.6484375,
      "rewards/final_brier_reward_step": 0.6445009708404541,
      "rewards/format_reward_step": 0.94921875,
      "step": 172
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.4876770152505447,
      "calib/avg_num_step_conf": 13.21875,
      "calib/ece": 0.3702610441767071,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.9759036144578314,
      "calib/gap": 0.0055453431372546325,
      "calib/mean_conf": 0.9758032128514057,
      "calib/mu_c": 0.9779411764705882,
      "calib/mu_w": 0.9723958333333336,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3658032128514059,
      "calib/std_conf": 0.08385274330610008,
      "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": 2552.0,
      "completions/max_terminated_length": 2552.0,
      "completions/mean_length": 758.0703125,
      "completions/mean_terminated_length": 764.0393676757812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 279.0,
      "epoch": 0.18453333333333333,
      "grad_norm": 0.0059276907704770565,
      "learning_rate": 7.5e-07,
      "loss": 0.0431,
      "num_tokens": 44402621.0,
      "reward": 1.3855817317962646,
      "reward_std": 0.3574729561805725,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.6071008443832397,
      "rewards/format_reward_step": 0.96875,
      "step": 173
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5388752723311546,
      "calib/avg_num_step_conf": 13.3515625,
      "calib/ece": 0.41520491803278686,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.9590163934426229,
      "calib/gap": 0.025065359477124516,
      "calib/mean_conf": 0.9725819672131147,
      "calib/mu_c": 0.9836764705882355,
      "calib/mu_w": 0.958611111111111,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.41520491803278686,
      "calib/std_conf": 0.08657587785080939,
      "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": 0.015625,
      "completions/max_length": 3015.0,
      "completions/max_terminated_length": 3015.0,
      "completions/mean_length": 818.7421875,
      "completions/mean_terminated_length": 831.7381591796875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 369.0,
      "epoch": 0.1856,
      "grad_norm": 0.0047086006961762905,
      "learning_rate": 7.222222222222222e-07,
      "loss": 0.0036,
      "num_tokens": 44716451.0,
      "reward": 1.282027006149292,
      "reward_std": 0.5039170980453491,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.5562413930892944,
      "rewards/format_reward_step": 0.9453125,
      "step": 174
    },
    {
      "calib/answer_extract_rate": 0.9453125,
      "calib/auroc": 0.5316115702479338,
      "calib/avg_num_step_conf": 13.703125,
      "calib/ece": 0.5224586776859506,
      "calib/final_conf_rate": 0.9453125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.9545454545454546,
      "calib/gap": 0.009106060606060562,
      "calib/mean_conf": 0.9680785123966944,
      "calib/mu_c": 0.9730454545454545,
      "calib/mu_w": 0.963939393939394,
      "calib/nonempty_final_conf_rate": 0.9453125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.5179958677685952,
      "calib/std_conf": 0.0969031106005789,
      "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": 0.0234375,
      "completions/max_length": 3007.0,
      "completions/max_terminated_length": 3007.0,
      "completions/mean_length": 788.08203125,
      "completions/mean_terminated_length": 806.9960327148438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 332.0,
      "epoch": 0.18666666666666668,
      "grad_norm": 0.004811226390302181,
      "learning_rate": 6.944444444444446e-07,
      "loss": -0.012,
      "num_tokens": 45024024.0,
      "reward": 1.128366231918335,
      "reward_std": 0.40532705187797546,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.4559510350227356,
      "rewards/format_reward_step": 0.94140625,
      "step": 175
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5092234824878474,
      "calib/avg_num_step_conf": 13.39453125,
      "calib/ece": 0.4220392156862746,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.9921568627450981,
      "calib/gap": 0.017257260376417616,
      "calib/mean_conf": 0.9789019607843139,
      "calib/mu_c": 0.9865492957746477,
      "calib/mu_w": 0.9692920353982301,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4220392156862746,
      "calib/std_conf": 0.08299876089364717,
      "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": 2317.0,
      "completions/max_terminated_length": 2317.0,
      "completions/mean_length": 770.65234375,
      "completions/mean_terminated_length": 773.674560546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 293.0,
      "epoch": 0.18773333333333334,
      "grad_norm": 0.00466120382770896,
      "learning_rate": 6.666666666666667e-07,
      "loss": 0.0033,
      "num_tokens": 45325375.0,
      "reward": 1.3379881381988525,
      "reward_std": 0.4600631594657898,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.5744138360023499,
      "rewards/format_reward_step": 0.9921875,
      "step": 176
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5415127357573402,
      "calib/avg_num_step_conf": 12.40625,
      "calib/ece": 0.42604,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.98,
      "calib/gap": 0.0025024304880421067,
      "calib/mean_conf": 0.98004,
      "calib/mu_c": 0.9811510791366906,
      "calib/mu_w": 0.9786486486486485,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.42504,
      "calib/std_conf": 0.026525429308495652,
      "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": 2635.0,
      "completions/max_terminated_length": 2635.0,
      "completions/mean_length": 726.93359375,
      "completions/mean_terminated_length": 735.5534057617188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 379.0,
      "epoch": 0.1888,
      "grad_norm": 0.00513953622430563,
      "learning_rate": 6.388888888888889e-07,
      "loss": 0.0125,
      "num_tokens": 45615302.0,
      "reward": 1.3037654161453247,
      "reward_std": 0.335764080286026,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.5528433322906494,
      "rewards/format_reward_step": 0.96875,
      "step": 177
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5136054421768707,
      "calib/avg_num_step_conf": 12.98828125,
      "calib/ece": 0.3749003984063746,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.9880478087649402,
      "calib/gap": 0.000624249699880064,
      "calib/mean_conf": 0.9844621513944224,
      "calib/mu_c": 0.9847058823529413,
      "calib/mu_w": 0.9840816326530613,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3749003984063746,
      "calib/std_conf": 0.015968393887289656,
      "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": 3050.0,
      "completions/max_terminated_length": 3050.0,
      "completions/mean_length": 736.109375,
      "completions/mean_terminated_length": 738.99609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 333.0,
      "epoch": 0.18986666666666666,
      "grad_norm": 0.004597798455506563,
      "learning_rate": 6.111111111111112e-07,
      "loss": 0.042,
      "num_tokens": 45909818.0,
      "reward": 1.3925690650939941,
      "reward_std": 0.4295831024646759,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.6093569993972778,
      "rewards/format_reward_step": 0.98046875,
      "step": 178
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6071184821184822,
      "calib/avg_num_step_conf": 12.37109375,
      "calib/ece": 0.38009176788124155,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.9554655870445344,
      "calib/gap": 0.015821184821184886,
      "calib/mean_conf": 0.9738839406207827,
      "calib/mu_c": 0.9802252252252251,
      "calib/mu_w": 0.9644040404040403,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3773927125506073,
      "calib/std_conf": 0.05663986846172311,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 2866.0,
      "completions/max_terminated_length": 2866.0,
      "completions/mean_length": 749.6796875,
      "completions/mean_terminated_length": 761.5794067382812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 331.0,
      "epoch": 0.19093333333333334,
      "grad_norm": 0.005119526293128729,
      "learning_rate": 5.833333333333334e-07,
      "loss": 0.0142,
      "num_tokens": 46208000.0,
      "reward": 1.3591045141220093,
      "reward_std": 0.4644799828529358,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.6010216474533081,
      "rewards/format_reward_step": 0.9609375,
      "step": 179
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5116135328562135,
      "calib/avg_num_step_conf": 14.2734375,
      "calib/ece": 0.4028685258964144,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.9760956175298805,
      "calib/gap": 0.008314248536109181,
      "calib/mean_conf": 0.9805577689243029,
      "calib/mu_c": 0.9840689655172413,
      "calib/mu_w": 0.9757547169811321,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4028685258964144,
      "calib/std_conf": 0.0643856010835871,
      "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": 2615.0,
      "completions/max_terminated_length": 2615.0,
      "completions/mean_length": 811.69140625,
      "completions/mean_terminated_length": 818.0827026367188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 347.0,
      "epoch": 0.192,
      "grad_norm": 0.005072887521237135,
      "learning_rate": 5.555555555555555e-07,
      "loss": 0.0256,
      "num_tokens": 46519649.0,
      "reward": 1.3476653099060059,
      "reward_std": 0.31839191913604736,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.5820492506027222,
      "rewards/format_reward_step": 0.98046875,
      "step": 180
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5585551806024247,
      "calib/avg_num_step_conf": 12.28515625,
      "calib/ece": 0.47905138339920955,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.9762845849802372,
      "calib/gap": 0.006711348581427412,
      "calib/mean_conf": 0.9810276679841897,
      "calib/mu_c": 0.9843700787401575,
      "calib/mu_w": 0.9776587301587301,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.47905138339920955,
      "calib/std_conf": 0.019528311901133372,
      "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": 2601.0,
      "completions/max_terminated_length": 2601.0,
      "completions/mean_length": 724.99609375,
      "completions/mean_terminated_length": 730.7047119140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 366.0,
      "epoch": 0.19306666666666666,
      "grad_norm": 0.005564880557358265,
      "learning_rate": 5.277777777777779e-07,
      "loss": -0.019,
      "num_tokens": 46811512.0,
      "reward": 1.24691903591156,
      "reward_std": 0.421791672706604,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5172755718231201,
      "rewards/format_reward_step": 0.984375,
      "step": 181
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5355902777777777,
      "calib/avg_num_step_conf": 13.62890625,
      "calib/ece": 0.3652579365079366,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.9880952380952381,
      "calib/gap": 0.0014022435897435015,
      "calib/mean_conf": 0.9843055555555557,
      "calib/mu_c": 0.9848397435897436,
      "calib/mu_w": 0.9834375000000001,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3652579365079366,
      "calib/std_conf": 0.015046682941715125,
      "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": 2640.0,
      "completions/max_terminated_length": 2640.0,
      "completions/mean_length": 757.38671875,
      "completions/mean_terminated_length": 766.3676147460938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 377.0,
      "epoch": 0.19413333333333332,
      "grad_norm": 0.004975506104528904,
      "learning_rate": 5.000000000000001e-07,
      "loss": -0.0333,
      "num_tokens": 47111563.0,
      "reward": 1.4083222150802612,
      "reward_std": 0.34334421157836914,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.6174256801605225,
      "rewards/format_reward_step": 0.98046875,
      "step": 182
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.500189393939394,
      "calib/avg_num_step_conf": 13.6015625,
      "calib/ece": 0.4109251968503936,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.968503937007874,
      "calib/gap": 0.0006250000000000977,
      "calib/mean_conf": 0.9778543307086615,
      "calib/mu_c": 0.9781250000000001,
      "calib/mu_w": 0.9775,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4109251968503936,
      "calib/std_conf": 0.029806172419749376,
      "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": 2796.0,
      "completions/max_terminated_length": 2796.0,
      "completions/mean_length": 782.46484375,
      "completions/mean_terminated_length": 782.46484375,
      "completions/min_length": 330.0,
      "completions/min_terminated_length": 330.0,
      "epoch": 0.1952,
      "grad_norm": 0.005101703107357025,
      "learning_rate": 4.7222222222222226e-07,
      "loss": 0.0187,
      "num_tokens": 47418554.0,
      "reward": 1.348827600479126,
      "reward_std": 0.4252949655056,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.5804678201675415,
      "rewards/format_reward_step": 0.9921875,
      "step": 183
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5591919609057746,
      "calib/avg_num_step_conf": 13.23828125,
      "calib/ece": 0.32352362204724405,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.9763779527559056,
      "calib/gap": 0.004149287631633247,
      "calib/mean_conf": 0.981003937007874,
      "calib/mu_c": 0.9824251497005988,
      "calib/mu_w": 0.9782758620689656,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.32352362204724405,
      "calib/std_conf": 0.020075545909267006,
      "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": 2541.0,
      "completions/max_terminated_length": 2541.0,
      "completions/mean_length": 757.546875,
      "completions/mean_terminated_length": 760.5177001953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 272.0,
      "epoch": 0.19626666666666667,
      "grad_norm": 0.004986181855201721,
      "learning_rate": 4.444444444444445e-07,
      "loss": 0.0007,
      "num_tokens": 47717766.0,
      "reward": 1.4816131591796875,
      "reward_std": 0.2942894697189331,
      "rewards/accuracy_reward_step": 0.65234375,
      "rewards/final_brier_reward_step": 0.6663514971733093,
      "rewards/format_reward_step": 0.9921875,
      "step": 184
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.6218245614035087,
      "calib/avg_num_step_conf": 14.19921875,
      "calib/ece": 0.3641224489795918,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.8979591836734694,
      "calib/gap": 0.008778947368420775,
      "calib/mean_conf": 0.9637959183673469,
      "calib/mu_c": 0.9671999999999997,
      "calib/mu_w": 0.958421052631579,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.35783673469387745,
      "calib/std_conf": 0.07952331954016516,
      "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": 3070.0,
      "completions/max_terminated_length": 3070.0,
      "completions/mean_length": 822.60546875,
      "completions/mean_terminated_length": 822.60546875,
      "completions/min_length": 311.0,
      "completions/min_terminated_length": 311.0,
      "epoch": 0.19733333333333333,
      "grad_norm": 0.004520988091826439,
      "learning_rate": 4.1666666666666667e-07,
      "loss": 0.0194,
      "num_tokens": 48035273.0,
      "reward": 1.3592960834503174,
      "reward_std": 0.3480474054813385,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.6014046669006348,
      "rewards/format_reward_step": 0.9453125,
      "step": 185
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5727453399122807,
      "calib/avg_num_step_conf": 14.375,
      "calib/ece": 0.3745927419354841,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.9637096774193549,
      "calib/gap": -0.00312609649122797,
      "calib/mean_conf": 0.9729798387096775,
      "calib/mu_c": 0.9717697368421054,
      "calib/mu_w": 0.9748958333333334,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.36733467741935505,
      "calib/std_conf": 0.08144211528385956,
      "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": 2744.0,
      "completions/max_terminated_length": 2744.0,
      "completions/mean_length": 796.85546875,
      "completions/mean_terminated_length": 806.3043823242188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 311.0,
      "epoch": 0.1984,
      "grad_norm": 0.004679948557168245,
      "learning_rate": 3.8888888888888895e-07,
      "loss": 0.0232,
      "num_tokens": 48344308.0,
      "reward": 1.3730634450912476,
      "reward_std": 0.3464033007621765,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.5976893901824951,
      "rewards/format_reward_step": 0.9609375,
      "step": 186
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.55846881663113,
      "calib/avg_num_step_conf": 14.91015625,
      "calib/ece": 0.44058943089430896,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.991869918699187,
      "calib/gap": 0.0037559968017056855,
      "calib/mean_conf": 0.9853048780487804,
      "calib/mu_c": 0.9870149253731343,
      "calib/mu_w": 0.9832589285714286,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.44058943089430896,
      "calib/std_conf": 0.014199813085100618,
      "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": 2854.0,
      "completions/max_terminated_length": 2854.0,
      "completions/mean_length": 816.32421875,
      "completions/mean_terminated_length": 826.0039672851562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 299.0,
      "epoch": 0.19946666666666665,
      "grad_norm": 0.005040613003075123,
      "learning_rate": 3.611111111111111e-07,
      "loss": 0.0234,
      "num_tokens": 48654831.0,
      "reward": 1.2702534198760986,
      "reward_std": 0.47981759905815125,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.5366007089614868,
      "rewards/format_reward_step": 0.95703125,
      "step": 187
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.4861683734539167,
      "calib/avg_num_step_conf": 13.28515625,
      "calib/ece": 0.39433734939759035,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.9718875502008032,
      "calib/gap": -0.0018160659662185274,
      "calib/mean_conf": 0.9806827309236947,
      "calib/mu_c": 0.9799315068493151,
      "calib/mu_w": 0.9817475728155336,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.39433734939759035,
      "calib/std_conf": 0.021563759548728724,
      "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": 2649.0,
      "completions/max_terminated_length": 2649.0,
      "completions/mean_length": 772.3671875,
      "completions/mean_terminated_length": 775.3961181640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 291.0,
      "epoch": 0.20053333333333334,
      "grad_norm": 0.005349774844944477,
      "learning_rate": 3.3333333333333335e-07,
      "loss": 0.0388,
      "num_tokens": 48956629.0,
      "reward": 1.3448392152786255,
      "reward_std": 0.33842799067497253,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.5803034901618958,
      "rewards/format_reward_step": 0.96875,
      "step": 188
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5526046986721144,
      "calib/avg_num_step_conf": 11.64453125,
      "calib/ece": 0.3324015748031495,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.9960629921259843,
      "calib/gap": 0.002968335035750891,
      "calib/mean_conf": 0.9811417322834645,
      "calib/mu_c": 0.9821818181818182,
      "calib/mu_w": 0.9792134831460673,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3319685039370078,
      "calib/std_conf": 0.01633421818188331,
      "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": 2131.0,
      "completions/max_terminated_length": 2131.0,
      "completions/mean_length": 682.56640625,
      "completions/mean_terminated_length": 685.2431640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 289.0,
      "epoch": 0.2016,
      "grad_norm": 0.005616314243525267,
      "learning_rate": 3.055555555555556e-07,
      "loss": 0.0066,
      "num_tokens": 49239134.0,
      "reward": 1.4676647186279297,
      "reward_std": 0.37088748812675476,
      "rewards/accuracy_reward_step": 0.64453125,
      "rewards/final_brier_reward_step": 0.6579858660697937,
      "rewards/format_reward_step": 0.98828125,
      "step": 189
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.569698703279939,
      "calib/avg_num_step_conf": 13.71875,
      "calib/ece": 0.445370634920635,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.9722222222222222,
      "calib/gap": -0.014109305873379219,
      "calib/mean_conf": 0.9711769841269843,
      "calib/mu_c": 0.9647942028985506,
      "calib/mu_w": 0.9789035087719298,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.4344642857142858,
      "calib/std_conf": 0.09984325943395767,
      "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": 2784.0,
      "completions/max_terminated_length": 2784.0,
      "completions/mean_length": 829.5703125,
      "completions/mean_terminated_length": 832.8236083984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 283.0,
      "epoch": 0.20266666666666666,
      "grad_norm": 0.004342732951045036,
      "learning_rate": 2.7777777777777776e-07,
      "loss": 0.0434,
      "num_tokens": 49557112.0,
      "reward": 1.3048603534698486,
      "reward_std": 0.31585729122161865,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5472208857536316,
      "rewards/format_reward_step": 0.984375,
      "step": 190
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.48879999999999996,
      "calib/avg_num_step_conf": 13.609375,
      "calib/ece": 0.48379200000000006,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.988,
      "calib/gap": -0.001024000000000247,
      "calib/mean_conf": 0.9837920000000001,
      "calib/mu_c": 0.9832799999999999,
      "calib/mu_w": 0.9843040000000002,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.48379200000000006,
      "calib/std_conf": 0.014890155674135848,
      "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": 2983.0,
      "completions/max_terminated_length": 2983.0,
      "completions/mean_length": 760.3828125,
      "completions/mean_terminated_length": 769.3992309570312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 320.0,
      "epoch": 0.20373333333333332,
      "grad_norm": 0.005018308758735657,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0128,
      "num_tokens": 49855938.0,
      "reward": 1.228130578994751,
      "reward_std": 0.3813462257385254,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.5031362771987915,
      "rewards/format_reward_step": 0.9765625,
      "step": 191
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.588122332859175,
      "calib/avg_num_step_conf": 13.0859375,
      "calib/ece": 0.36819135802469133,
      "calib/final_conf_rate": 0.94921875,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.9835390946502057,
      "calib/gap": 0.017762873399715606,
      "calib/mean_conf": 0.9772448559670782,
      "calib/mu_c": 0.9841891891891892,
      "calib/mu_w": 0.9664263157894736,
      "calib/nonempty_final_conf_rate": 0.94921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.36819135802469133,
      "calib/std_conf": 0.06522785896110676,
      "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": 0.03125,
      "completions/max_length": 2817.0,
      "completions/max_terminated_length": 2817.0,
      "completions/mean_length": 741.56640625,
      "completions/mean_terminated_length": 765.4878540039062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 324.0,
      "epoch": 0.2048,
      "grad_norm": 0.0051372176967561245,
      "learning_rate": 2.2222222222222224e-07,
      "loss": -0.0055,
      "num_tokens": 50150755.0,
      "reward": 1.348089575767517,
      "reward_std": 0.41244158148765564,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.5907105207443237,
      "rewards/format_reward_step": 0.94140625,
      "step": 192
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5299048013245033,
      "calib/avg_num_step_conf": 14.00390625,
      "calib/ece": 0.36805668016194343,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.9676113360323887,
      "calib/gap": 0.009059740618101109,
      "calib/mean_conf": 0.9793927125506071,
      "calib/mu_c": 0.9829139072847679,
      "calib/mu_w": 0.9738541666666668,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.36805668016194343,
      "calib/std_conf": 0.038410151053830825,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 2966.0,
      "completions/max_terminated_length": 2966.0,
      "completions/mean_length": 757.72265625,
      "completions/mean_terminated_length": 775.9080200195312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 405.0,
      "epoch": 0.20586666666666667,
      "grad_norm": 0.004883711691945791,
      "learning_rate": 1.9444444444444447e-07,
      "loss": -0.019,
      "num_tokens": 50450444.0,
      "reward": 1.3639910221099854,
      "reward_std": 0.5335198640823364,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.5990757942199707,
      "rewards/format_reward_step": 0.94921875,
      "step": 193
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6402847021779061,
      "calib/avg_num_step_conf": 12.921875,
      "calib/ece": 0.38382470119521916,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.9402390438247012,
      "calib/gap": 0.02168525321437953,
      "calib/mean_conf": 0.9734661354581674,
      "calib/mu_c": 0.9823648648648647,
      "calib/mu_w": 0.9606796116504852,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.38382470119521916,
      "calib/std_conf": 0.0386053926703595,
      "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": 2622.0,
      "completions/max_terminated_length": 2622.0,
      "completions/mean_length": 734.34765625,
      "completions/mean_terminated_length": 737.2274780273438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 338.0,
      "epoch": 0.20693333333333333,
      "grad_norm": 0.004884345922619104,
      "learning_rate": 1.6666666666666668e-07,
      "loss": -0.0004,
      "num_tokens": 50744381.0,
      "reward": 1.372166395187378,
      "reward_std": 0.28283581137657166,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.6076140403747559,
      "rewards/format_reward_step": 0.98046875,
      "step": 194
    },
    {
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.559592822636301,
      "calib/avg_num_step_conf": 13.25390625,
      "calib/ece": 0.40950617283950624,
      "calib/final_conf_rate": 0.94921875,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.9670781893004116,
      "calib/gap": 0.00868322981366454,
      "calib/mean_conf": 0.9774074074074075,
      "calib/mu_c": 0.9811594202898551,
      "calib/mu_w": 0.9724761904761906,
      "calib/nonempty_final_conf_rate": 0.94921875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.40950617283950624,
      "calib/std_conf": 0.028850877176519178,
      "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": 0.01953125,
      "completions/max_length": 3022.0,
      "completions/max_terminated_length": 3022.0,
      "completions/mean_length": 791.9609375,
      "completions/mean_terminated_length": 807.737060546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 333.0,
      "epoch": 0.208,
      "grad_norm": 0.005530741531401873,
      "learning_rate": 1.3888888888888888e-07,
      "loss": -0.0004,
      "num_tokens": 51053107.0,
      "reward": 1.2913596630096436,
      "reward_std": 0.40997639298439026,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5592820048332214,
      "rewards/format_reward_step": 0.9453125,
      "step": 195
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5305325987144169,
      "calib/avg_num_step_conf": 12.5234375,
      "calib/ece": 0.37438735177865623,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.9881422924901185,
      "calib/gap": 0.004069264069264222,
      "calib/mean_conf": 0.9830830039525693,
      "calib/mu_c": 0.9846753246753248,
      "calib/mu_w": 0.9806060606060606,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.37438735177865623,
      "calib/std_conf": 0.019559906653369336,
      "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": 1653.0,
      "completions/max_terminated_length": 1653.0,
      "completions/mean_length": 688.078125,
      "completions/mean_terminated_length": 690.7764892578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 364.0,
      "epoch": 0.20906666666666668,
      "grad_norm": 0.004695164505392313,
      "learning_rate": 1.1111111111111112e-07,
      "loss": 0.0017,
      "num_tokens": 51331799.0,
      "reward": 1.407560110092163,
      "reward_std": 0.3021622896194458,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.615901529788971,
      "rewards/format_reward_step": 0.98828125,
      "step": 196
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5234830418653947,
      "calib/avg_num_step_conf": 13.2890625,
      "calib/ece": 0.43194331983805667,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.9838056680161943,
      "calib/gap": -0.0001126126126125282,
      "calib/mean_conf": 0.9825506072874494,
      "calib/mu_c": 0.9825,
      "calib/mu_w": 0.9826126126126126,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.43194331983805667,
      "calib/std_conf": 0.018842315781338333,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 2906.0,
      "completions/max_terminated_length": 2906.0,
      "completions/mean_length": 721.5546875,
      "completions/mean_terminated_length": 733.0079956054688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 362.0,
      "epoch": 0.21013333333333334,
      "grad_norm": 0.005680236965417862,
      "learning_rate": 8.333333333333334e-08,
      "loss": -0.006,
      "num_tokens": 51621573.0,
      "reward": 1.2806005477905273,
      "reward_std": 0.4925641715526581,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.5416699051856995,
      "rewards/format_reward_step": 0.95703125,
      "step": 197
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.638334026910806,
      "calib/avg_num_step_conf": 12.09375,
      "calib/ece": 0.330996015936255,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.9721115537848606,
      "calib/gap": 0.019513108614232277,
      "calib/mean_conf": 0.9764143426294822,
      "calib/mu_c": 0.9833333333333334,
      "calib/mu_w": 0.9638202247191011,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.330996015936255,
      "calib/std_conf": 0.04828422277298219,
      "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": 2635.0,
      "completions/max_terminated_length": 2635.0,
      "completions/mean_length": 688.01171875,
      "completions/mean_terminated_length": 696.1699829101562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 273.0,
      "epoch": 0.2112,
      "grad_norm": 0.005235373042523861,
      "learning_rate": 5.555555555555556e-08,
      "loss": 0.0059,
      "num_tokens": 51903088.0,
      "reward": 1.450615644454956,
      "reward_std": 0.3237333297729492,
      "rewards/accuracy_reward_step": 0.6328125,
      "rewards/final_brier_reward_step": 0.6551374793052673,
      "rewards/format_reward_step": 0.98046875,
      "step": 198
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5294384057971014,
      "calib/avg_num_step_conf": 13.5703125,
      "calib/ece": 0.43600000000000017,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.972,
      "calib/gap": -0.00765010351966855,
      "calib/mean_conf": 0.9779200000000001,
      "calib/mu_c": 0.9744927536231885,
      "calib/mu_w": 0.9821428571428571,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.4309600000000002,
      "calib/std_conf": 0.06697815763366444,
      "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": 2956.0,
      "completions/max_terminated_length": 2956.0,
      "completions/mean_length": 788.546875,
      "completions/mean_terminated_length": 794.7559204101562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 310.0,
      "epoch": 0.21226666666666666,
      "grad_norm": 0.00531103927642107,
      "learning_rate": 2.777777777777778e-08,
      "loss": -0.0174,
      "num_tokens": 52209156.0,
      "reward": 1.2944998741149902,
      "reward_std": 0.4788621962070465,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5421249866485596,
      "rewards/format_reward_step": 0.96875,
      "step": 199
    },
    {
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.6175240929705216,
      "calib/avg_num_step_conf": 12.734375,
      "calib/ece": 0.36910000000000015,
      "calib/final_conf_rate": 0.94921875,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.9547325102880658,
      "calib/gap": 0.021978380102040562,
      "calib/mean_conf": 0.9740382716049383,
      "calib/mu_c": 0.982721088435374,
      "calib/mu_w": 0.9607427083333334,
      "calib/nonempty_final_conf_rate": 0.94921875,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.36910000000000015,
      "calib/std_conf": 0.0670588137560041,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 2834.0,
      "completions/max_terminated_length": 2834.0,
      "completions/mean_length": 767.05859375,
      "completions/mean_terminated_length": 785.468017578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 238.0,
      "epoch": 0.21333333333333335,
      "grad_norm": 0.00476312730461359,
      "learning_rate": 0.0,
      "loss": -0.0217,
      "num_tokens": 52513571.0,
      "reward": 1.3442987203598022,
      "reward_std": 0.33491945266723633,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.5948475003242493,
      "rewards/format_reward_step": 0.9453125,
      "step": 200
    },
    {
      "epoch": 0.21333333333333335,
      "step": 200,
      "total_flos": 0.0,
      "train_loss": 0.02728243867168203,
      "train_runtime": 9826.6555,
      "train_samples_per_second": 5.21,
      "train_steps_per_second": 0.02
    }
  ],
  "logging_steps": 1,
  "max_steps": 200,
  "num_input_tokens_seen": 52513571,
  "num_train_epochs": 1,
  "save_steps": 20,
  "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
}