{
  "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": 0.0040498161688447,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0322,
      "num_tokens": 264685.0,
      "reward": 0.055236753076314926,
      "reward_std": 0.11281141638755798,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.01655624993145466,
      "rewards/format_reward_step": 0.04296875,
      "rewards/stepwise_brier_reward": 0.024703249335289,
      "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": 0.006203852593898773,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0643,
      "num_tokens": 533467.0,
      "reward": 0.11156807839870453,
      "reward_std": 0.21452845633029938,
      "rewards/accuracy_reward_step": 0.03125,
      "rewards/final_brier_reward_step": 0.02965039201080799,
      "rewards/format_reward_step": 0.08984375,
      "rewards/stepwise_brier_reward": 0.04943438619375229,
      "step": 2
    },
    {
      "calib/answer_extract_rate": 0.06640625,
      "calib/auroc": 0.5555555555555556,
      "calib/avg_num_step_conf": 0.3359375,
      "calib/ece": 0.553853846153846,
      "calib/final_conf_rate": 0.05078125,
      "calib/format_rate": 0.04296875,
      "calib/frac_conf_gt_0.9": 0.7692307692307693,
      "calib/gap": 0.09887777777777784,
      "calib/mean_conf": 0.8615461538461537,
      "calib/mu_c": 0.9299999999999999,
      "calib/mu_w": 0.8311222222222221,
      "calib/nonempty_final_conf_rate": 0.05078125,
      "calib/nonempty_reasoning_rate": 0.08984375,
      "calib/nonempty_step_conf_rate": 0.0703125,
      "calib/pce": 0.553853846153846,
      "calib/std_conf": 0.24775902254900398,
      "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": 0.0859375,
      "completions/max_length": 2872.0,
      "completions/max_terminated_length": 2872.0,
      "completions/mean_length": 696.01953125,
      "completions/mean_terminated_length": 761.457275390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0032,
      "grad_norm": 0.004721686244010925,
      "learning_rate": 7.5e-07,
      "loss": 0.0425,
      "num_tokens": 816904.0,
      "reward": 0.05670829862356186,
      "reward_std": 0.12842696905136108,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.02178749069571495,
      "rewards/format_reward_step": 0.04296875,
      "rewards/stepwise_brier_reward": 0.02535819262266159,
      "step": 3
    },
    {
      "calib/answer_extract_rate": 0.078125,
      "calib/auroc": 0.5375,
      "calib/avg_num_step_conf": 0.29296875,
      "calib/ece": 0.575,
      "calib/final_conf_rate": 0.0546875,
      "calib/format_rate": 0.046875,
      "calib/frac_conf_gt_0.9": 0.8571428571428571,
      "calib/gap": 0.139,
      "calib/mean_conf": 0.8607142857142858,
      "calib/mu_c": 0.96,
      "calib/mu_w": 0.821,
      "calib/nonempty_final_conf_rate": 0.0546875,
      "calib/nonempty_reasoning_rate": 0.09765625,
      "calib/nonempty_step_conf_rate": 0.0703125,
      "calib/pce": 0.575,
      "calib/std_conf": 0.27101566337744826,
      "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": 3063.0,
      "completions/max_terminated_length": 3063.0,
      "completions/mean_length": 659.09375,
      "completions/mean_terminated_length": 724.1544799804688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.004266666666666667,
      "grad_norm": 0.004437054041773081,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0262,
      "num_tokens": 1091800.0,
      "reward": 0.05978678539395332,
      "reward_std": 0.1366736888885498,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.0242253877222538,
      "rewards/format_reward_step": 0.046875,
      "rewards/stepwise_brier_reward": 0.027421751990914345,
      "step": 4
    },
    {
      "calib/answer_extract_rate": 0.05078125,
      "calib/auroc": 0.25,
      "calib/avg_num_step_conf": 0.16015625,
      "calib/ece": 0.7177777777777776,
      "calib/final_conf_rate": 0.03515625,
      "calib/format_rate": 0.0234375,
      "calib/frac_conf_gt_0.9": 0.6666666666666666,
      "calib/gap": 0.07999999999999996,
      "calib/mean_conf": 0.8288888888888889,
      "calib/mu_c": 0.9,
      "calib/mu_w": 0.8200000000000001,
      "calib/nonempty_final_conf_rate": 0.03515625,
      "calib/nonempty_reasoning_rate": 0.05078125,
      "calib/nonempty_step_conf_rate": 0.03125,
      "calib/pce": 0.7177777777777776,
      "calib/std_conf": 0.3111785641162194,
      "calib/step_conf_rate": 0.03125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.09375,
      "completions/max_length": 3051.0,
      "completions/max_terminated_length": 3051.0,
      "completions/mean_length": 713.015625,
      "completions/mean_terminated_length": 786.77587890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.005333333333333333,
      "grad_norm": 0.002612976124510169,
      "learning_rate": 1.25e-06,
      "loss": 0.0354,
      "num_tokens": 1381020.0,
      "reward": 0.022981680929660797,
      "reward_std": 0.060142092406749725,
      "rewards/accuracy_reward_step": 0.00390625,
      "rewards/final_brier_reward_step": 0.006355468649417162,
      "rewards/format_reward_step": 0.0234375,
      "rewards/stepwise_brier_reward": 0.015258748084306717,
      "step": 5
    },
    {
      "calib/answer_extract_rate": 0.1171875,
      "calib/auroc": 0.531578947368421,
      "calib/avg_num_step_conf": 0.375,
      "calib/ece": 0.7041666666666667,
      "calib/final_conf_rate": 0.09375,
      "calib/format_rate": 0.0625,
      "calib/frac_conf_gt_0.9": 0.875,
      "calib/gap": 0.05999999999999994,
      "calib/mean_conf": 0.9125,
      "calib/mu_c": 0.96,
      "calib/mu_w": 0.9,
      "calib/nonempty_final_conf_rate": 0.09375,
      "calib/nonempty_reasoning_rate": 0.1328125,
      "calib/nonempty_step_conf_rate": 0.08203125,
      "calib/pce": 0.7041666666666667,
      "calib/std_conf": 0.15476460620352875,
      "calib/step_conf_rate": 0.08203125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0703125,
      "completions/max_length": 3014.0,
      "completions/max_terminated_length": 3014.0,
      "completions/mean_length": 609.15234375,
      "completions/mean_terminated_length": 655.2227172851562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0064,
      "grad_norm": 0.003538590855896473,
      "learning_rate": 1.5e-06,
      "loss": 0.0299,
      "num_tokens": 1642915.0,
      "reward": 0.0765293687582016,
      "reward_std": 0.13818004727363586,
      "rewards/accuracy_reward_step": 0.01953125,
      "rewards/final_brier_reward_step": 0.02597109228372574,
      "rewards/format_reward_step": 0.0625,
      "rewards/stepwise_brier_reward": 0.03795885294675827,
      "step": 6
    },
    {
      "calib/answer_extract_rate": 0.1015625,
      "calib/auroc": 0.8125,
      "calib/avg_num_step_conf": 0.4921875,
      "calib/ece": 0.8647058823529413,
      "calib/final_conf_rate": 0.06640625,
      "calib/format_rate": 0.05859375,
      "calib/frac_conf_gt_0.9": 0.7058823529411765,
      "calib/gap": 0.06000000000000005,
      "calib/mean_conf": 0.9235294117647058,
      "calib/mu_c": 0.98,
      "calib/mu_w": 0.9199999999999999,
      "calib/nonempty_final_conf_rate": 0.06640625,
      "calib/nonempty_reasoning_rate": 0.1171875,
      "calib/nonempty_step_conf_rate": 0.0859375,
      "calib/pce": 0.8647058823529413,
      "calib/std_conf": 0.10312420703270675,
      "calib/step_conf_rate": 0.0859375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.09765625,
      "completions/max_length": 3027.0,
      "completions/max_terminated_length": 3027.0,
      "completions/mean_length": 743.33984375,
      "completions/mean_terminated_length": 823.7879028320312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.007466666666666667,
      "grad_norm": 0.004745212383568287,
      "learning_rate": 1.75e-06,
      "loss": 0.0266,
      "num_tokens": 1940634.0,
      "reward": 0.05298828333616257,
      "reward_std": 0.09799638390541077,
      "rewards/accuracy_reward_step": 0.0078125,
      "rewards/final_brier_reward_step": 0.011421484872698784,
      "rewards/format_reward_step": 0.05859375,
      "rewards/stepwise_brier_reward": 0.03646914288401604,
      "step": 7
    },
    {
      "calib/answer_extract_rate": 0.0859375,
      "calib/auroc": 0.8506493506493507,
      "calib/avg_num_step_conf": 0.3671875,
      "calib/ece": 0.4780555555555557,
      "calib/final_conf_rate": 0.0703125,
      "calib/format_rate": 0.06640625,
      "calib/frac_conf_gt_0.9": 0.7222222222222222,
      "calib/gap": 0.1756493506493506,
      "calib/mean_conf": 0.8669444444444444,
      "calib/mu_c": 0.9742857142857143,
      "calib/mu_w": 0.7986363636363637,
      "calib/nonempty_final_conf_rate": 0.0703125,
      "calib/nonempty_reasoning_rate": 0.09765625,
      "calib/nonempty_step_conf_rate": 0.08203125,
      "calib/pce": 0.4780555555555557,
      "calib/std_conf": 0.24804311278983163,
      "calib/step_conf_rate": 0.08203125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1171875,
      "completions/max_length": 3021.0,
      "completions/max_terminated_length": 3021.0,
      "completions/mean_length": 641.24609375,
      "completions/mean_terminated_length": 726.3672485351562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.008533333333333334,
      "grad_norm": 0.004859757609665394,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0115,
      "num_tokens": 2211305.0,
      "reward": 0.092274010181427,
      "reward_std": 0.18566857278347015,
      "rewards/accuracy_reward_step": 0.02734375,
      "rewards/final_brier_reward_step": 0.03517138585448265,
      "rewards/format_reward_step": 0.06640625,
      "rewards/stepwise_brier_reward": 0.03704963997006416,
      "step": 8
    },
    {
      "calib/answer_extract_rate": 0.1015625,
      "calib/auroc": 0.926470588235294,
      "calib/avg_num_step_conf": 0.328125,
      "calib/ece": 0.8073684210526315,
      "calib/final_conf_rate": 0.07421875,
      "calib/format_rate": 0.0625,
      "calib/frac_conf_gt_0.9": 0.8421052631578947,
      "calib/gap": 0.09205882352941175,
      "calib/mean_conf": 0.9126315789473682,
      "calib/mu_c": 0.995,
      "calib/mu_w": 0.9029411764705882,
      "calib/nonempty_final_conf_rate": 0.07421875,
      "calib/nonempty_reasoning_rate": 0.1015625,
      "calib/nonempty_step_conf_rate": 0.0625,
      "calib/pce": 0.8073684210526315,
      "calib/std_conf": 0.19086622775521844,
      "calib/step_conf_rate": 0.0625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2959.0,
      "completions/max_terminated_length": 2959.0,
      "completions/mean_length": 589.00390625,
      "completions/mean_terminated_length": 658.4497680664062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 5.0,
      "epoch": 0.0096,
      "grad_norm": 0.007626960054039955,
      "learning_rate": 2.25e-06,
      "loss": 0.0175,
      "num_tokens": 2469626.0,
      "reward": 0.05566889047622681,
      "reward_std": 0.12854988873004913,
      "rewards/accuracy_reward_step": 0.0078125,
      "rewards/final_brier_reward_step": 0.015421093441545963,
      "rewards/format_reward_step": 0.0625,
      "rewards/stepwise_brier_reward": 0.03537946939468384,
      "step": 9
    },
    {
      "calib/answer_extract_rate": 0.1171875,
      "calib/auroc": 0.49404761904761907,
      "calib/avg_num_step_conf": 0.40234375,
      "calib/ece": 0.7312000000000001,
      "calib/final_conf_rate": 0.09765625,
      "calib/format_rate": 0.08203125,
      "calib/frac_conf_gt_0.9": 0.84,
      "calib/gap": 0.06630952380952371,
      "calib/mean_conf": 0.8768,
      "calib/mu_c": 0.9325,
      "calib/mu_w": 0.8661904761904763,
      "calib/nonempty_final_conf_rate": 0.09765625,
      "calib/nonempty_reasoning_rate": 0.125,
      "calib/nonempty_step_conf_rate": 0.08984375,
      "calib/pce": 0.724,
      "calib/std_conf": 0.23929429579494785,
      "calib/step_conf_rate": 0.08984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05859375,
      "completions/max_length": 3026.0,
      "completions/max_terminated_length": 3026.0,
      "completions/mean_length": 666.9453125,
      "completions/mean_terminated_length": 708.4564819335938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.010666666666666666,
      "grad_norm": 0.0056964014656841755,
      "learning_rate": 2.5e-06,
      "loss": 0.0414,
      "num_tokens": 2747164.0,
      "reward": 0.08224216103553772,
      "reward_std": 0.1923098862171173,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.02598554641008377,
      "rewards/format_reward_step": 0.08203125,
      "rewards/stepwise_brier_reward": 0.045170582830905914,
      "step": 10
    },
    {
      "calib/answer_extract_rate": 0.1875,
      "calib/auroc": 0.453125,
      "calib/avg_num_step_conf": 0.7421875,
      "calib/ece": 0.7353657894736841,
      "calib/final_conf_rate": 0.1484375,
      "calib/format_rate": 0.1171875,
      "calib/frac_conf_gt_0.9": 0.7105263157894737,
      "calib/gap": 0.04404479166666664,
      "calib/mean_conf": 0.8795763157894736,
      "calib/mu_c": 0.9166666666666666,
      "calib/mu_w": 0.872621875,
      "calib/nonempty_final_conf_rate": 0.1484375,
      "calib/nonempty_reasoning_rate": 0.23046875,
      "calib/nonempty_step_conf_rate": 0.171875,
      "calib/pce": 0.7285236842105263,
      "calib/std_conf": 0.20181194925006496,
      "calib/step_conf_rate": 0.171875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.04296875,
      "completions/max_length": 3025.0,
      "completions/max_terminated_length": 3025.0,
      "completions/mean_length": 707.6015625,
      "completions/mean_terminated_length": 739.3713989257812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.011733333333333333,
      "grad_norm": 0.006040748208761215,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.075,
      "num_tokens": 3032790.0,
      "reward": 0.11954192072153091,
      "reward_std": 0.18748214840888977,
      "rewards/accuracy_reward_step": 0.0234375,
      "rewards/final_brier_reward_step": 0.03433149307966232,
      "rewards/format_reward_step": 0.1171875,
      "rewards/stepwise_brier_reward": 0.06883619725704193,
      "step": 11
    },
    {
      "calib/answer_extract_rate": 0.19140625,
      "calib/auroc": 0.5789473684210527,
      "calib/avg_num_step_conf": 0.64453125,
      "calib/ece": 0.5094484848484849,
      "calib/final_conf_rate": 0.12890625,
      "calib/format_rate": 0.11328125,
      "calib/frac_conf_gt_0.9": 0.8181818181818182,
      "calib/gap": 0.05772481203007518,
      "calib/mean_conf": 0.912478787878788,
      "calib/mu_c": 0.9457142857142857,
      "calib/mu_w": 0.8879894736842106,
      "calib/nonempty_final_conf_rate": 0.12890625,
      "calib/nonempty_reasoning_rate": 0.23046875,
      "calib/nonempty_step_conf_rate": 0.15625,
      "calib/pce": 0.4988424242424243,
      "calib/std_conf": 0.17290449940265368,
      "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": 0.125,
      "completions/max_length": 2957.0,
      "completions/max_terminated_length": 2957.0,
      "completions/mean_length": 500.68359375,
      "completions/mean_terminated_length": 572.2098388671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0128,
      "grad_norm": 0.007509665098041296,
      "learning_rate": 3e-06,
      "loss": 0.065,
      "num_tokens": 3265141.0,
      "reward": 0.17071044445037842,
      "reward_std": 0.330117404460907,
      "rewards/accuracy_reward_step": 0.0546875,
      "rewards/final_brier_reward_step": 0.0605458989739418,
      "rewards/format_reward_step": 0.11328125,
      "rewards/stepwise_brier_reward": 0.06760834902524948,
      "step": 12
    },
    {
      "calib/answer_extract_rate": 0.24609375,
      "calib/auroc": 0.4939271255060728,
      "calib/avg_num_step_conf": 1.32421875,
      "calib/ece": 0.6212941176470589,
      "calib/final_conf_rate": 0.19921875,
      "calib/format_rate": 0.1640625,
      "calib/frac_conf_gt_0.9": 0.7058823529411765,
      "calib/gap": 0.042352226720647757,
      "calib/mean_conf": 0.8530588235294121,
      "calib/mu_c": 0.8846153846153846,
      "calib/mu_w": 0.8422631578947368,
      "calib/nonempty_final_conf_rate": 0.19921875,
      "calib/nonempty_reasoning_rate": 0.296875,
      "calib/nonempty_step_conf_rate": 0.24609375,
      "calib/pce": 0.6097254901960785,
      "calib/std_conf": 0.24945976230723188,
      "calib/step_conf_rate": 0.24609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05859375,
      "completions/max_length": 3009.0,
      "completions/max_terminated_length": 3009.0,
      "completions/mean_length": 682.77734375,
      "completions/mean_terminated_length": 725.27392578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 8.0,
      "epoch": 0.013866666666666666,
      "grad_norm": 0.006930575706064701,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.1169,
      "num_tokens": 3544524.0,
      "reward": 0.21188601851463318,
      "reward_std": 0.38299399614334106,
      "rewards/accuracy_reward_step": 0.05859375,
      "rewards/final_brier_reward_step": 0.06761399656534195,
      "rewards/format_reward_step": 0.1640625,
      "rewards/stepwise_brier_reward": 0.10024251788854599,
      "step": 13
    },
    {
      "calib/answer_extract_rate": 0.359375,
      "calib/auroc": 0.5449236298292902,
      "calib/avg_num_step_conf": 1.4375,
      "calib/ece": 0.6217297297297297,
      "calib/final_conf_rate": 0.2890625,
      "calib/format_rate": 0.234375,
      "calib/frac_conf_gt_0.9": 0.7297297297297297,
      "calib/gap": 0.04615633423180576,
      "calib/mean_conf": 0.9055135135135135,
      "calib/mu_c": 0.9385714285714284,
      "calib/mu_w": 0.8924150943396226,
      "calib/nonempty_final_conf_rate": 0.2890625,
      "calib/nonempty_reasoning_rate": 0.40625,
      "calib/nonempty_step_conf_rate": 0.30859375,
      "calib/pce": 0.6217297297297297,
      "calib/std_conf": 0.15345800052618,
      "calib/step_conf_rate": 0.30859375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2908.0,
      "completions/max_terminated_length": 2908.0,
      "completions/mean_length": 591.48046875,
      "completions/mean_terminated_length": 615.5243530273438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.014933333333333333,
      "grad_norm": 0.00944006908684969,
      "learning_rate": 3.5e-06,
      "loss": 0.1012,
      "num_tokens": 3801343.0,
      "reward": 0.2951757311820984,
      "reward_std": 0.4583866000175476,
      "rewards/accuracy_reward_step": 0.08203125,
      "rewards/final_brier_reward_step": 0.0882023349404335,
      "rewards/format_reward_step": 0.234375,
      "rewards/stepwise_brier_reward": 0.13156315684318542,
      "step": 14
    },
    {
      "calib/answer_extract_rate": 0.4296875,
      "calib/auroc": 0.5465277777777778,
      "calib/avg_num_step_conf": 1.7265625,
      "calib/ece": 0.6830434782608696,
      "calib/final_conf_rate": 0.359375,
      "calib/format_rate": 0.3203125,
      "calib/frac_conf_gt_0.9": 0.717391304347826,
      "calib/gap": 0.012222222222222356,
      "calib/mean_conf": 0.9004347826086957,
      "calib/mu_c": 0.9100000000000001,
      "calib/mu_w": 0.8977777777777778,
      "calib/nonempty_final_conf_rate": 0.359375,
      "calib/nonempty_reasoning_rate": 0.46484375,
      "calib/nonempty_step_conf_rate": 0.37109375,
      "calib/pce": 0.6830434782608696,
      "calib/std_conf": 0.16801858750356122,
      "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": 0.0234375,
      "completions/max_length": 2582.0,
      "completions/max_terminated_length": 2582.0,
      "completions/mean_length": 517.28515625,
      "completions/mean_terminated_length": 529.7000122070312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.016,
      "grad_norm": 0.012086746282875538,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.0694,
      "num_tokens": 4041648.0,
      "reward": 0.35359853506088257,
      "reward_std": 0.40532323718070984,
      "rewards/accuracy_reward_step": 0.078125,
      "rewards/final_brier_reward_step": 0.1023242175579071,
      "rewards/format_reward_step": 0.3203125,
      "rewards/stepwise_brier_reward": 0.20269489288330078,
      "step": 15
    },
    {
      "calib/answer_extract_rate": 0.48046875,
      "calib/auroc": 0.45916666666666667,
      "calib/avg_num_step_conf": 2.203125,
      "calib/ece": 0.7230227272727273,
      "calib/final_conf_rate": 0.4296875,
      "calib/format_rate": 0.3515625,
      "calib/frac_conf_gt_0.9": 0.7636363636363637,
      "calib/gap": 0.005455555555555547,
      "calib/mean_conf": 0.8880863636363635,
      "calib/mu_c": 0.89255,
      "calib/mu_w": 0.8870944444444444,
      "calib/nonempty_final_conf_rate": 0.4296875,
      "calib/nonempty_reasoning_rate": 0.578125,
      "calib/nonempty_step_conf_rate": 0.4765625,
      "calib/pce": 0.7146454545454546,
      "calib/std_conf": 0.22062873853737872,
      "calib/step_conf_rate": 0.4765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03515625,
      "completions/max_length": 2850.0,
      "completions/max_terminated_length": 2850.0,
      "completions/mean_length": 540.2578125,
      "completions/mean_terminated_length": 559.943359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.017066666666666667,
      "grad_norm": 0.010672827251255512,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.1363,
      "num_tokens": 4288802.0,
      "reward": 0.3784298300743103,
      "reward_std": 0.49684762954711914,
      "rewards/accuracy_reward_step": 0.08203125,
      "rewards/final_brier_reward_step": 0.10257644206285477,
      "rewards/format_reward_step": 0.3515625,
      "rewards/stepwise_brier_reward": 0.21583032608032227,
      "step": 16
    },
    {
      "calib/answer_extract_rate": 0.6171875,
      "calib/auroc": 0.48190045248868774,
      "calib/avg_num_step_conf": 3.13671875,
      "calib/ece": 0.6582450331125829,
      "calib/final_conf_rate": 0.58984375,
      "calib/format_rate": 0.5,
      "calib/frac_conf_gt_0.9": 0.7019867549668874,
      "calib/gap": 0.004967320261437869,
      "calib/mean_conf": 0.8705629139072847,
      "calib/mu_c": 0.8744117647058822,
      "calib/mu_w": 0.8694444444444444,
      "calib/nonempty_final_conf_rate": 0.58984375,
      "calib/nonempty_reasoning_rate": 0.70703125,
      "calib/nonempty_step_conf_rate": 0.62890625,
      "calib/pce": 0.6518211920529803,
      "calib/std_conf": 0.2147164640849067,
      "calib/step_conf_rate": 0.62890625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3003.0,
      "completions/max_terminated_length": 3003.0,
      "completions/mean_length": 440.71875,
      "completions/mean_terminated_length": 451.2960205078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 8.0,
      "epoch": 0.018133333333333335,
      "grad_norm": 0.009812280535697937,
      "learning_rate": 4.25e-06,
      "loss": 0.1454,
      "num_tokens": 4505154.0,
      "reward": 0.5809885859489441,
      "reward_std": 0.6631735563278198,
      "rewards/accuracy_reward_step": 0.13671875,
      "rewards/final_brier_reward_step": 0.19295614957809448,
      "rewards/format_reward_step": 0.5,
      "rewards/stepwise_brier_reward": 0.3106856942176819,
      "step": 17
    },
    {
      "calib/answer_extract_rate": 0.65234375,
      "calib/auroc": 0.5294930875576036,
      "calib/avg_num_step_conf": 3.41015625,
      "calib/ece": 0.6736477987421385,
      "calib/final_conf_rate": 0.62109375,
      "calib/format_rate": 0.546875,
      "calib/frac_conf_gt_0.9": 0.7610062893081762,
      "calib/gap": 0.022638248847925868,
      "calib/mean_conf": 0.8937735849056603,
      "calib/mu_c": 0.9114285714285713,
      "calib/mu_w": 0.8887903225806454,
      "calib/nonempty_final_conf_rate": 0.62109375,
      "calib/nonempty_reasoning_rate": 0.734375,
      "calib/nonempty_step_conf_rate": 0.6640625,
      "calib/pce": 0.6736477987421385,
      "calib/std_conf": 0.18813443355326082,
      "calib/step_conf_rate": 0.6640625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 2636.0,
      "completions/max_terminated_length": 2636.0,
      "completions/mean_length": 445.26953125,
      "completions/mean_terminated_length": 455.9560241699219,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 26.0,
      "epoch": 0.0192,
      "grad_norm": 0.009522214531898499,
      "learning_rate": 4.5e-06,
      "loss": 0.154,
      "num_tokens": 4729863.0,
      "reward": 0.6203250885009766,
      "reward_std": 0.6184340715408325,
      "rewards/accuracy_reward_step": 0.140625,
      "rewards/final_brier_reward_step": 0.19318124651908875,
      "rewards/format_reward_step": 0.546875,
      "rewards/stepwise_brier_reward": 0.3506191074848175,
      "step": 18
    },
    {
      "calib/answer_extract_rate": 0.85546875,
      "calib/auroc": 0.5221840833968981,
      "calib/avg_num_step_conf": 4.39453125,
      "calib/ece": 0.6783364055299539,
      "calib/final_conf_rate": 0.84765625,
      "calib/format_rate": 0.76953125,
      "calib/frac_conf_gt_0.9": 0.6866359447004609,
      "calib/gap": 0.015556318332061947,
      "calib/mean_conf": 0.8816543778801844,
      "calib/mu_c": 0.8939130434782608,
      "calib/mu_w": 0.8783567251461989,
      "calib/nonempty_final_conf_rate": 0.84765625,
      "calib/nonempty_reasoning_rate": 0.953125,
      "calib/nonempty_step_conf_rate": 0.91015625,
      "calib/pce": 0.6740046082949308,
      "calib/std_conf": 0.18007124906241606,
      "calib/step_conf_rate": 0.91015625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3008.0,
      "completions/max_terminated_length": 3008.0,
      "completions/mean_length": 323.453125,
      "completions/mean_terminated_length": 327.2885437011719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 8.0,
      "epoch": 0.020266666666666665,
      "grad_norm": 0.01093797292560339,
      "learning_rate": 4.75e-06,
      "loss": 0.0717,
      "num_tokens": 4917427.0,
      "reward": 0.8620531558990479,
      "reward_std": 0.6407575607299805,
      "rewards/accuracy_reward_step": 0.1875,
      "rewards/final_brier_reward_step": 0.27811408042907715,
      "rewards/format_reward_step": 0.76953125,
      "rewards/stepwise_brier_reward": 0.5060364007949829,
      "step": 19
    },
    {
      "calib/answer_extract_rate": 0.86328125,
      "calib/auroc": 0.5285938654841094,
      "calib/avg_num_step_conf": 4.4765625,
      "calib/ece": 0.6103956521739131,
      "calib/final_conf_rate": 0.8984375,
      "calib/format_rate": 0.8203125,
      "calib/frac_conf_gt_0.9": 0.717391304347826,
      "calib/gap": 0.011736511456023613,
      "calib/mean_conf": 0.8793434782608696,
      "calib/mu_c": 0.8877121212121212,
      "calib/mu_w": 0.8759756097560976,
      "calib/nonempty_final_conf_rate": 0.8984375,
      "calib/nonempty_reasoning_rate": 0.96875,
      "calib/nonempty_step_conf_rate": 0.92578125,
      "calib/pce": 0.6013913043478262,
      "calib/std_conf": 0.19598049886590194,
      "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": 0.0078125,
      "completions/max_length": 1490.0,
      "completions/max_terminated_length": 1490.0,
      "completions/mean_length": 270.5390625,
      "completions/mean_terminated_length": 272.6692810058594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 7.0,
      "epoch": 0.021333333333333333,
      "grad_norm": 0.010510188527405262,
      "learning_rate": 5e-06,
      "loss": 0.0665,
      "num_tokens": 5091557.0,
      "reward": 1.0149328708648682,
      "reward_std": 0.7371776103973389,
      "rewards/accuracy_reward_step": 0.2578125,
      "rewards/final_brier_reward_step": 0.3442583680152893,
      "rewards/format_reward_step": 0.8203125,
      "rewards/stepwise_brier_reward": 0.5279731154441833,
      "step": 20
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4460788542834958,
      "calib/avg_num_step_conf": 4.90625,
      "calib/ece": 0.6471679166666667,
      "calib/final_conf_rate": 0.9375,
      "calib/format_rate": 0.921875,
      "calib/frac_conf_gt_0.9": 0.7791666666666667,
      "calib/gap": -0.024085937365197263,
      "calib/mean_conf": 0.9161679166666669,
      "calib/mu_c": 0.8988059701492537,
      "calib/mu_w": 0.9228919075144509,
      "calib/nonempty_final_conf_rate": 0.9375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.6420845833333334,
      "calib/std_conf": 0.1259816409071565,
      "calib/step_conf_rate": 0.9765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2262.0,
      "completions/max_terminated_length": 2262.0,
      "completions/mean_length": 283.921875,
      "completions/mean_terminated_length": 285.0353088378906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 51.0,
      "epoch": 0.0224,
      "grad_norm": 0.008625438436865807,
      "learning_rate": 4.9722222222222224e-06,
      "loss": 0.0561,
      "num_tokens": 5267201.0,
      "reward": 1.100550889968872,
      "reward_std": 0.6891491413116455,
      "rewards/accuracy_reward_step": 0.26953125,
      "rewards/final_brier_reward_step": 0.343247652053833,
      "rewards/format_reward_step": 0.921875,
      "rewards/stepwise_brier_reward": 0.5980186462402344,
      "step": 21
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.49097638057218895,
      "calib/avg_num_step_conf": 4.65234375,
      "calib/ece": 0.603175732217573,
      "calib/final_conf_rate": 0.93359375,
      "calib/format_rate": 0.9140625,
      "calib/frac_conf_gt_0.9": 0.698744769874477,
      "calib/gap": 0.020667664670658614,
      "calib/mean_conf": 0.8950585774058577,
      "calib/mu_c": 0.9095,
      "calib/mu_w": 0.8888323353293414,
      "calib/nonempty_final_conf_rate": 0.93359375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.5984895397489538,
      "calib/std_conf": 0.1637264171260149,
      "calib/step_conf_rate": 0.97265625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1551.0,
      "completions/max_terminated_length": 1551.0,
      "completions/mean_length": 259.1953125,
      "completions/mean_terminated_length": 259.1953125,
      "completions/min_length": 7.0,
      "completions/min_terminated_length": 7.0,
      "epoch": 0.023466666666666667,
      "grad_norm": 0.008928249590098858,
      "learning_rate": 4.944444444444445e-06,
      "loss": -0.0152,
      "num_tokens": 5435371.0,
      "reward": 1.1315412521362305,
      "reward_std": 0.7189508676528931,
      "rewards/accuracy_reward_step": 0.28515625,
      "rewards/final_brier_reward_step": 0.3857576251029968,
      "rewards/format_reward_step": 0.9140625,
      "rewards/stepwise_brier_reward": 0.6013451814651489,
      "step": 22
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4944663692518874,
      "calib/avg_num_step_conf": 4.55859375,
      "calib/ece": 0.64112,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9296875,
      "calib/frac_conf_gt_0.9": 0.716,
      "calib/gap": 0.018452299245024117,
      "calib/mean_conf": 0.87064,
      "calib/mu_c": 0.8845161290322581,
      "calib/mu_w": 0.866063829787234,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.63188,
      "calib/std_conf": 0.2155801252434927,
      "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": 1238.0,
      "completions/max_terminated_length": 1238.0,
      "completions/mean_length": 262.15234375,
      "completions/mean_terminated_length": 262.15234375,
      "completions/min_length": 88.0,
      "completions/min_terminated_length": 88.0,
      "epoch": 0.024533333333333334,
      "grad_norm": 0.008457643911242485,
      "learning_rate": 4.9166666666666665e-06,
      "loss": -0.0006,
      "num_tokens": 5606418.0,
      "reward": 1.0740993022918701,
      "reward_std": 0.6285284757614136,
      "rewards/accuracy_reward_step": 0.2421875,
      "rewards/final_brier_reward_step": 0.36159294843673706,
      "rewards/format_reward_step": 0.9296875,
      "rewards/stepwise_brier_reward": 0.6223044991493225,
      "step": 23
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.4774484536082475,
      "calib/avg_num_step_conf": 4.9296875,
      "calib/ece": 0.6777235772357724,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.93359375,
      "calib/frac_conf_gt_0.9": 0.6951219512195121,
      "calib/gap": 0.02600713719270409,
      "calib/mean_conf": 0.8891056910569107,
      "calib/mu_c": 0.9096153846153846,
      "calib/mu_w": 0.8836082474226805,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.6777235772357724,
      "calib/std_conf": 0.16910336536862294,
      "calib/step_conf_rate": 0.97265625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2538.0,
      "completions/max_terminated_length": 2538.0,
      "completions/mean_length": 281.55859375,
      "completions/mean_terminated_length": 281.55859375,
      "completions/min_length": 48.0,
      "completions/min_terminated_length": 48.0,
      "epoch": 0.0256,
      "grad_norm": 0.008437077514827251,
      "learning_rate": 4.888888888888889e-06,
      "loss": 0.0738,
      "num_tokens": 5783009.0,
      "reward": 1.0127873420715332,
      "reward_std": 0.5980093479156494,
      "rewards/accuracy_reward_step": 0.203125,
      "rewards/final_brier_reward_step": 0.33446913957595825,
      "rewards/format_reward_step": 0.93359375,
      "rewards/stepwise_brier_reward": 0.6307430863380432,
      "step": 24
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5340866571877573,
      "calib/avg_num_step_conf": 5.234375,
      "calib/ece": 0.5391618852459015,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.6229508196721312,
      "calib/gap": 0.04392161191349242,
      "calib/mean_conf": 0.8685430327868853,
      "calib/mu_c": 0.8975240963855422,
      "calib/mu_w": 0.8536024844720498,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.5337704918032785,
      "calib/std_conf": 0.19359585140432667,
      "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": 1198.0,
      "completions/max_terminated_length": 1198.0,
      "completions/mean_length": 280.28515625,
      "completions/mean_terminated_length": 281.38433837890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 49.0,
      "epoch": 0.02666666666666667,
      "grad_norm": 0.007779384031891823,
      "learning_rate": 4.861111111111111e-06,
      "loss": -0.0124,
      "num_tokens": 5957986.0,
      "reward": 1.2369064092636108,
      "reward_std": 0.6545590162277222,
      "rewards/accuracy_reward_step": 0.328125,
      "rewards/final_brier_reward_step": 0.44903552532196045,
      "rewards/format_reward_step": 0.94140625,
      "rewards/stepwise_brier_reward": 0.6470277309417725,
      "step": 25
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.49043005671077505,
      "calib/avg_num_step_conf": 5.14453125,
      "calib/ece": 0.6145981554677207,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.6442687747035574,
      "calib/gap": -0.015380434782608643,
      "calib/mean_conf": 0.8832147562582344,
      "calib/mu_c": 0.8720289855072464,
      "calib/mu_w": 0.887409420289855,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.6125428194993412,
      "calib/std_conf": 0.1627337088137336,
      "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": 716.0,
      "completions/max_terminated_length": 716.0,
      "completions/mean_length": 275.27734375,
      "completions/mean_terminated_length": 276.35687255859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 100.0,
      "epoch": 0.027733333333333332,
      "grad_norm": 0.008365216664969921,
      "learning_rate": 4.833333333333333e-06,
      "loss": -0.0109,
      "num_tokens": 6133697.0,
      "reward": 1.1731226444244385,
      "reward_std": 0.5622432231903076,
      "rewards/accuracy_reward_step": 0.26953125,
      "rewards/final_brier_reward_step": 0.3909871578216553,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.7155659198760986,
      "step": 26
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4591040589111949,
      "calib/avg_num_step_conf": 5.33203125,
      "calib/ece": 0.6002741935483872,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.5282258064516129,
      "calib/gap": -0.0024202682563337907,
      "calib/mean_conf": 0.8306774193548387,
      "calib/mu_c": 0.8288524590163934,
      "calib/mu_w": 0.8312727272727272,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.592491935483871,
      "calib/std_conf": 0.21991851674255056,
      "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": 3055.0,
      "completions/max_terminated_length": 3055.0,
      "completions/mean_length": 327.75,
      "completions/mean_terminated_length": 327.75,
      "completions/min_length": 83.0,
      "completions/min_terminated_length": 83.0,
      "epoch": 0.0288,
      "grad_norm": 0.007940828800201416,
      "learning_rate": 4.805555555555556e-06,
      "loss": 0.071,
      "num_tokens": 6322817.0,
      "reward": 1.118673324584961,
      "reward_std": 0.5638723373413086,
      "rewards/accuracy_reward_step": 0.23828125,
      "rewards/final_brier_reward_step": 0.4043359160423279,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.7266073822975159,
      "step": 27
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5241018171729781,
      "calib/avg_num_step_conf": 5.2890625,
      "calib/ece": 0.44848605577689243,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.4063745019920319,
      "calib/gap": 0.028539325842696583,
      "calib/mean_conf": 0.7801195219123506,
      "calib/mu_c": 0.7985393258426966,
      "calib/mu_w": 0.77,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.4370119521912351,
      "calib/std_conf": 0.22883645070809805,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1562.0,
      "completions/max_terminated_length": 1562.0,
      "completions/mean_length": 319.4375,
      "completions/mean_terminated_length": 319.4375,
      "completions/min_length": 103.0,
      "completions/min_terminated_length": 103.0,
      "epoch": 0.029866666666666666,
      "grad_norm": 0.0069170789793133736,
      "learning_rate": 4.777777777777778e-06,
      "loss": 0.0388,
      "num_tokens": 6511537.0,
      "reward": 1.3392189741134644,
      "reward_std": 0.617000937461853,
      "rewards/accuracy_reward_step": 0.3515625,
      "rewards/final_brier_reward_step": 0.5301804542541504,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.7876328229904175,
      "step": 28
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.460865561694291,
      "calib/avg_num_step_conf": 6.0078125,
      "calib/ece": 0.46424291497975717,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9296875,
      "calib/frac_conf_gt_0.9": 0.32793522267206476,
      "calib/gap": -0.0018699146157710178,
      "calib/mean_conf": 0.7272793522267207,
      "calib/mu_c": 0.725909090909091,
      "calib/mu_w": 0.727779005524862,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.4621578947368422,
      "calib/std_conf": 0.2607661223114103,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2038.0,
      "completions/max_terminated_length": 2038.0,
      "completions/mean_length": 396.26171875,
      "completions/mean_terminated_length": 397.8157043457031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 139.0,
      "epoch": 0.030933333333333334,
      "grad_norm": 0.005927573889493942,
      "learning_rate": 4.75e-06,
      "loss": 0.0884,
      "num_tokens": 6720108.0,
      "reward": 1.1746200323104858,
      "reward_std": 0.6304316520690918,
      "rewards/accuracy_reward_step": 0.26171875,
      "rewards/final_brier_reward_step": 0.4880704879760742,
      "rewards/format_reward_step": 0.9296875,
      "rewards/stepwise_brier_reward": 0.7807222604751587,
      "step": 29
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5158426966292136,
      "calib/avg_num_step_conf": 6.0234375,
      "calib/ece": 0.35924901185770747,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.22529644268774704,
      "calib/gap": 0.020675655430711593,
      "calib/mean_conf": 0.6333201581027668,
      "calib/mu_c": 0.6478666666666667,
      "calib/mu_w": 0.6271910112359551,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.3480632411067193,
      "calib/std_conf": 0.2838165294692594,
      "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": 1814.0,
      "completions/max_terminated_length": 1814.0,
      "completions/mean_length": 370.265625,
      "completions/mean_terminated_length": 370.265625,
      "completions/min_length": 91.0,
      "completions/min_terminated_length": 91.0,
      "epoch": 0.032,
      "grad_norm": 0.005835330579429865,
      "learning_rate": 4.722222222222222e-06,
      "loss": 0.0146,
      "num_tokens": 6921880.0,
      "reward": 1.2952570915222168,
      "reward_std": 0.52005535364151,
      "rewards/accuracy_reward_step": 0.296875,
      "rewards/final_brier_reward_step": 0.5941238403320312,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8525295257568359,
      "step": 30
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.47856310356310355,
      "calib/avg_num_step_conf": 6.13671875,
      "calib/ece": 0.36245967741935486,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.13709677419354838,
      "calib/gap": -0.019958374958375047,
      "calib/mean_conf": 0.5747983870967742,
      "calib/mu_c": 0.5601515151515152,
      "calib/mu_w": 0.5801098901098902,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.33556451612903226,
      "calib/std_conf": 0.277652819874921,
      "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": 2377.0,
      "completions/max_terminated_length": 2377.0,
      "completions/mean_length": 390.66796875,
      "completions/mean_terminated_length": 390.66796875,
      "completions/min_length": 115.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.03306666666666667,
      "grad_norm": 0.005797600839287043,
      "learning_rate": 4.694444444444445e-06,
      "loss": 0.0471,
      "num_tokens": 7127803.0,
      "reward": 1.2438130378723145,
      "reward_std": 0.5065557956695557,
      "rewards/accuracy_reward_step": 0.265625,
      "rewards/final_brier_reward_step": 0.6011230945587158,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.8506916165351868,
      "step": 31
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5199475623582767,
      "calib/avg_num_step_conf": 5.71875,
      "calib/ece": 0.23920634920634917,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.07142857142857142,
      "calib/gap": 0.018214285714285627,
      "calib/mean_conf": 0.4847619047619047,
      "calib/mu_c": 0.4969047619047619,
      "calib/mu_w": 0.4786904761904763,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.19531746031746028,
      "calib/std_conf": 0.2714210525274376,
      "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": 2415.0,
      "completions/max_terminated_length": 2415.0,
      "completions/mean_length": 369.02734375,
      "completions/mean_terminated_length": 369.02734375,
      "completions/min_length": 115.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.034133333333333335,
      "grad_norm": 0.006477469112724066,
      "learning_rate": 4.666666666666667e-06,
      "loss": 0.0744,
      "num_tokens": 7328978.0,
      "reward": 1.3680706024169922,
      "reward_std": 0.4866165518760681,
      "rewards/accuracy_reward_step": 0.328125,
      "rewards/final_brier_reward_step": 0.6714035272598267,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8790041208267212,
      "step": 32
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5659700452803901,
      "calib/avg_num_step_conf": 5.7890625,
      "calib/ece": 0.1629746031746032,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.051587301587301584,
      "calib/gap": 0.052554231974921484,
      "calib/mean_conf": 0.42996190476190477,
      "calib/mu_c": 0.46437241379310334,
      "calib/mu_w": 0.41181818181818186,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.12384920634920635,
      "calib/std_conf": 0.24329715806022356,
      "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": 1542.0,
      "completions/max_terminated_length": 1542.0,
      "completions/mean_length": 387.45703125,
      "completions/mean_terminated_length": 387.45703125,
      "completions/min_length": 132.0,
      "completions/min_terminated_length": 132.0,
      "epoch": 0.0352,
      "grad_norm": 0.005612695589661598,
      "learning_rate": 4.638888888888889e-06,
      "loss": 0.0391,
      "num_tokens": 7535039.0,
      "reward": 1.3998472690582275,
      "reward_std": 0.5582293272018433,
      "rewards/accuracy_reward_step": 0.33984375,
      "rewards/final_brier_reward_step": 0.716796875,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8825923204421997,
      "step": 33
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4201649175412294,
      "calib/avg_num_step_conf": 5.58984375,
      "calib/ece": 0.2755590551181103,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.04330708661417323,
      "calib/gap": -0.06942753623188402,
      "calib/mean_conf": 0.37822047244094487,
      "calib/mu_c": 0.34049999999999997,
      "calib/mu_w": 0.409927536231884,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.09854330708661418,
      "calib/std_conf": 0.23812968455758163,
      "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": 1212.0,
      "completions/max_terminated_length": 1212.0,
      "completions/mean_length": 342.25,
      "completions/mean_terminated_length": 343.5921630859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 85.0,
      "epoch": 0.03626666666666667,
      "grad_norm": 0.006945520173758268,
      "learning_rate": 4.611111111111112e-06,
      "loss": 0.005,
      "num_tokens": 7727767.0,
      "reward": 1.5494495630264282,
      "reward_std": 0.5377414226531982,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6412307024002075,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8846927285194397,
      "step": 34
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4798036465638149,
      "calib/avg_num_step_conf": 6.08984375,
      "calib/ece": 0.2348582995951417,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.032388663967611336,
      "calib/gap": -0.018398316970547002,
      "calib/mean_conf": 0.398502024291498,
      "calib/mu_c": 0.3869565217391305,
      "calib/mu_w": 0.4053548387096775,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.13044534412955466,
      "calib/std_conf": 0.23782708777620928,
      "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": 2501.0,
      "completions/max_terminated_length": 2501.0,
      "completions/mean_length": 460.26171875,
      "completions/mean_terminated_length": 460.26171875,
      "completions/min_length": 122.0,
      "completions/min_terminated_length": 122.0,
      "epoch": 0.037333333333333336,
      "grad_norm": 0.005454621743410826,
      "learning_rate": 4.583333333333333e-06,
      "loss": 0.1171,
      "num_tokens": 7954850.0,
      "reward": 1.403791904449463,
      "reward_std": 0.6024346947669983,
      "rewards/accuracy_reward_step": 0.36328125,
      "rewards/final_brier_reward_step": 0.6721000075340271,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.8493179082870483,
      "step": 35
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5275575447570333,
      "calib/avg_num_step_conf": 5.6640625,
      "calib/ece": 0.2941195219123506,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.01195219123505976,
      "calib/gap": 0.02005012787723781,
      "calib/mean_conf": 0.28787250996015934,
      "calib/mu_c": 0.29705882352941176,
      "calib/mu_w": 0.27700869565217395,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.020079681274900396,
      "calib/std_conf": 0.2075100911511454,
      "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": 1636.0,
      "completions/max_terminated_length": 1636.0,
      "completions/mean_length": 356.33984375,
      "completions/mean_terminated_length": 357.7372741699219,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 110.0,
      "epoch": 0.0384,
      "grad_norm": 0.0070097423158586025,
      "learning_rate": 4.555555555555556e-06,
      "loss": 0.0373,
      "num_tokens": 8148785.0,
      "reward": 1.6728756427764893,
      "reward_std": 0.44604384899139404,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.6388723254203796,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.888568103313446,
      "step": 36
    },
    {
      "calib/answer_extract_rate": 0.9453125,
      "calib/auroc": 0.4613553113553114,
      "calib/avg_num_step_conf": 5.44921875,
      "calib/ece": 0.22573651452282156,
      "calib/final_conf_rate": 0.94140625,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.008298755186721992,
      "calib/gap": -0.027555128205128204,
      "calib/mean_conf": 0.25061203319502073,
      "calib/mu_c": 0.23346153846153847,
      "calib/mu_w": 0.2610166666666667,
      "calib/nonempty_final_conf_rate": 0.94140625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.04937759336099585,
      "calib/std_conf": 0.1813629127488925,
      "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": 2560.0,
      "completions/max_terminated_length": 2560.0,
      "completions/mean_length": 440.265625,
      "completions/mean_terminated_length": 441.9921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 98.0,
      "epoch": 0.039466666666666664,
      "grad_norm": 0.005869188345968723,
      "learning_rate": 4.527777777777778e-06,
      "loss": 0.1705,
      "num_tokens": 8368589.0,
      "reward": 1.3879544734954834,
      "reward_std": 0.45529523491859436,
      "rewards/accuracy_reward_step": 0.359375,
      "rewards/final_brier_reward_step": 0.6618225574493408,
      "rewards/format_reward_step": 0.94140625,
      "rewards/stepwise_brier_reward": 0.8509328961372375,
      "step": 37
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.48789722785665984,
      "calib/avg_num_step_conf": 5.08203125,
      "calib/ece": 0.20344129554655874,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.004048582995951417,
      "calib/gap": 0.001409060175794452,
      "calib/mean_conf": 0.27113360323886637,
      "calib/mu_c": 0.2719607843137255,
      "calib/mu_w": 0.27055172413793105,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.030809716599190285,
      "calib/std_conf": 0.18329289619049413,
      "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": 2715.0,
      "completions/max_terminated_length": 2715.0,
      "completions/mean_length": 392.77734375,
      "completions/mean_terminated_length": 394.3176574707031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 95.0,
      "epoch": 0.04053333333333333,
      "grad_norm": 0.0068012080155313015,
      "learning_rate": 4.5e-06,
      "loss": 0.1799,
      "num_tokens": 8576028.0,
      "reward": 1.4690444469451904,
      "reward_std": 0.450836718082428,
      "rewards/accuracy_reward_step": 0.40234375,
      "rewards/final_brier_reward_step": 0.6724566221237183,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.8755960464477539,
      "step": 38
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.511598357985274,
      "calib/avg_num_step_conf": 4.78515625,
      "calib/ece": 0.23355436507936506,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.007936507936507936,
      "calib/gap": -0.0040859646836515495,
      "calib/mean_conf": 0.23882658730158732,
      "calib/mu_c": 0.2364106796116505,
      "calib/mu_w": 0.24049664429530204,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.031825396825396826,
      "calib/std_conf": 0.165378487399972,
      "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": 2279.0,
      "completions/max_terminated_length": 2279.0,
      "completions/mean_length": 358.375,
      "completions/mean_terminated_length": 358.375,
      "completions/min_length": 73.0,
      "completions/min_terminated_length": 73.0,
      "epoch": 0.0416,
      "grad_norm": 0.007424731273204088,
      "learning_rate": 4.472222222222223e-06,
      "loss": 0.1441,
      "num_tokens": 8773860.0,
      "reward": 1.4880990982055664,
      "reward_std": 0.42358171939849854,
      "rewards/accuracy_reward_step": 0.40234375,
      "rewards/final_brier_reward_step": 0.6842081546783447,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.9010008573532104,
      "step": 39
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5260904739432346,
      "calib/avg_num_step_conf": 4.2890625,
      "calib/ece": 0.15598425196850393,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.006416773410638482,
      "calib/mean_conf": 0.24346456692913387,
      "calib/mu_c": 0.24758241758241764,
      "calib/mu_w": 0.24116564417177916,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.020590551181102368,
      "calib/std_conf": 0.1314478906528968,
      "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": 2257.0,
      "completions/max_terminated_length": 2257.0,
      "completions/mean_length": 358.2890625,
      "completions/mean_terminated_length": 358.2890625,
      "completions/min_length": 66.0,
      "completions/min_terminated_length": 66.0,
      "epoch": 0.042666666666666665,
      "grad_norm": 0.007885615341365337,
      "learning_rate": 4.444444444444444e-06,
      "loss": 0.054,
      "num_tokens": 8972342.0,
      "reward": 1.439058780670166,
      "reward_std": 0.41545605659484863,
      "rewards/accuracy_reward_step": 0.35546875,
      "rewards/final_brier_reward_step": 0.7350699305534363,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.9117902517318726,
      "step": 40
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.48192687815329327,
      "calib/avg_num_step_conf": 3.69921875,
      "calib/ece": 0.42384,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.017163591125855332,
      "calib/mean_conf": 0.23304000000000002,
      "calib/mu_c": 0.22679245283018865,
      "calib/mu_w": 0.24395604395604398,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.01044,
      "calib/std_conf": 0.13034246583519893,
      "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": 2443.0,
      "completions/max_terminated_length": 2443.0,
      "completions/mean_length": 322.546875,
      "completions/mean_terminated_length": 322.546875,
      "completions/min_length": 115.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.04373333333333333,
      "grad_norm": 0.008142044767737389,
      "learning_rate": 4.416666666666667e-06,
      "loss": 0.0904,
      "num_tokens": 9162162.0,
      "reward": 1.7881989479064941,
      "reward_std": 0.5068119168281555,
      "rewards/accuracy_reward_step": 0.62109375,
      "rewards/final_brier_reward_step": 0.5658527612686157,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.9150683879852295,
      "step": 41
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.533406318423522,
      "calib/avg_num_step_conf": 3.50390625,
      "calib/ece": 0.2625223097112861,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.007874015748031496,
      "calib/gap": 0.007689750808049234,
      "calib/mean_conf": 0.2148005249343832,
      "calib/mu_c": 0.21900869565217396,
      "calib/mu_w": 0.21131894484412472,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.012283464566929131,
      "calib/std_conf": 0.13146999005453217,
      "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": 2084.0,
      "completions/max_terminated_length": 2084.0,
      "completions/mean_length": 275.0078125,
      "completions/mean_terminated_length": 275.0078125,
      "completions/min_length": 84.0,
      "completions/min_terminated_length": 84.0,
      "epoch": 0.0448,
      "grad_norm": 0.009993243031203747,
      "learning_rate": 4.388888888888889e-06,
      "loss": 0.0874,
      "num_tokens": 9336932.0,
      "reward": 1.5709275007247925,
      "reward_std": 0.37025392055511475,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.6732515692710876,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.938583493232727,
      "step": 42
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5023479980227384,
      "calib/avg_num_step_conf": 3.33984375,
      "calib/ece": 0.23219607843137252,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003823529411764698,
      "calib/mean_conf": 0.24678431372549017,
      "calib/mu_c": 0.2488235294117647,
      "calib/mu_w": 0.245,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.006156862745098041,
      "calib/std_conf": 0.12070646842090392,
      "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": 2040.0,
      "completions/max_terminated_length": 2040.0,
      "completions/mean_length": 305.984375,
      "completions/mean_terminated_length": 305.984375,
      "completions/min_length": 93.0,
      "completions/min_terminated_length": 93.0,
      "epoch": 0.04586666666666667,
      "grad_norm": 0.00938818883150816,
      "learning_rate": 4.361111111111112e-06,
      "loss": -0.006,
      "num_tokens": 9520488.0,
      "reward": 1.5983197689056396,
      "reward_std": 0.46041882038116455,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.6835503578186035,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9362914562225342,
      "step": 43
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5968408011563081,
      "calib/avg_num_step_conf": 2.80859375,
      "calib/ece": 0.09551181102362206,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0395691375868952,
      "calib/mean_conf": 0.2628346456692914,
      "calib/mu_c": 0.2888505747126437,
      "calib/mu_w": 0.2492814371257485,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.007913385826771657,
      "calib/std_conf": 0.1133866469786607,
      "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": 1992.0,
      "completions/max_terminated_length": 1992.0,
      "completions/mean_length": 298.0625,
      "completions/mean_terminated_length": 298.0625,
      "completions/min_length": 75.0,
      "completions/min_terminated_length": 75.0,
      "epoch": 0.046933333333333334,
      "grad_norm": 0.009420789778232574,
      "learning_rate": 4.333333333333334e-06,
      "loss": 0.0713,
      "num_tokens": 9703112.0,
      "reward": 1.4322527647018433,
      "reward_std": 0.35099905729293823,
      "rewards/accuracy_reward_step": 0.33984375,
      "rewards/final_brier_reward_step": 0.7673734426498413,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9381999969482422,
      "step": 44
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5263662511984659,
      "calib/avg_num_step_conf": 2.49609375,
      "calib/ece": 0.17224409448818898,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.007874015748031496,
      "calib/gap": 0.0072860338766378985,
      "calib/mean_conf": 0.27153543307086614,
      "calib/mu_c": 0.27580952380952384,
      "calib/mu_w": 0.26852348993288594,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.015196850393700778,
      "calib/std_conf": 0.1375016274936232,
      "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": 2572.0,
      "completions/max_terminated_length": 2572.0,
      "completions/mean_length": 291.578125,
      "completions/mean_terminated_length": 292.7215881347656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 70.0,
      "epoch": 0.048,
      "grad_norm": 0.010626120492815971,
      "learning_rate": 4.305555555555556e-06,
      "loss": 0.0287,
      "num_tokens": 9882804.0,
      "reward": 1.5220348834991455,
      "reward_std": 0.42932039499282837,
      "rewards/accuracy_reward_step": 0.41015625,
      "rewards/final_brier_reward_step": 0.7141109704971313,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.9365284442901611,
      "step": 45
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4949874686716792,
      "calib/avg_num_step_conf": 2.73046875,
      "calib/ece": 0.21122529644268775,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.005614661654135322,
      "calib/mean_conf": 0.273201581027668,
      "calib/mu_c": 0.27025,
      "calib/mu_w": 0.2758646616541353,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.005059288537549407,
      "calib/std_conf": 0.1139222867938069,
      "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": 2797.0,
      "completions/max_terminated_length": 2797.0,
      "completions/mean_length": 320.91796875,
      "completions/mean_terminated_length": 320.91796875,
      "completions/min_length": 73.0,
      "completions/min_terminated_length": 73.0,
      "epoch": 0.04906666666666667,
      "grad_norm": 0.010083785280585289,
      "learning_rate": 4.277777777777778e-06,
      "loss": 0.0758,
      "num_tokens": 10069727.0,
      "reward": 1.592787265777588,
      "reward_std": 0.49180591106414795,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6757726669311523,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.9297513961791992,
      "step": 46
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.6044896640826873,
      "calib/avg_num_step_conf": 2.13671875,
      "calib/ece": 0.24012048192771088,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.028325581395348853,
      "calib/mean_conf": 0.2546586345381526,
      "calib/mu_c": 0.2693333333333333,
      "calib/mu_w": 0.24100775193798446,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.00642570281124498,
      "calib/std_conf": 0.1037605364841029,
      "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": 2751.0,
      "completions/max_terminated_length": 2751.0,
      "completions/mean_length": 315.53125,
      "completions/mean_terminated_length": 316.7686462402344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 68.0,
      "epoch": 0.050133333333333335,
      "grad_norm": 0.010103489272296429,
      "learning_rate": 4.25e-06,
      "loss": 0.0558,
      "num_tokens": 10256479.0,
      "reward": 1.5880870819091797,
      "reward_std": 0.44787827134132385,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6811476945877075,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.9212011098861694,
      "step": 47
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5208504238446814,
      "calib/avg_num_step_conf": 1.8515625,
      "calib/ece": 0.14577689243027883,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.006519688269073093,
      "calib/mean_conf": 0.24880478087649405,
      "calib/mu_c": 0.2529347826086957,
      "calib/mu_w": 0.24641509433962258,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.014023904382470117,
      "calib/std_conf": 0.09218553598516582,
      "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": 1606.0,
      "completions/max_terminated_length": 1606.0,
      "completions/mean_length": 260.640625,
      "completions/mean_terminated_length": 262.6929016113281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 75.0,
      "epoch": 0.0512,
      "grad_norm": 0.012790133245289326,
      "learning_rate": 4.222222222222223e-06,
      "loss": 0.05,
      "num_tokens": 10426891.0,
      "reward": 1.4405946731567383,
      "reward_std": 0.4182698130607605,
      "rewards/accuracy_reward_step": 0.359375,
      "rewards/final_brier_reward_step": 0.7264277338981628,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.9343888759613037,
      "step": 48
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4991228070175438,
      "calib/avg_num_step_conf": 2.27734375,
      "calib/ece": 0.20771653543307084,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0014010025062656806,
      "calib/mean_conf": 0.2632283464566929,
      "calib/mu_c": 0.26245614035087717,
      "calib/mu_w": 0.26385714285714285,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.011062992125984257,
      "calib/std_conf": 0.09765665299222448,
      "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": 2046.0,
      "completions/max_terminated_length": 2046.0,
      "completions/mean_length": 268.12109375,
      "completions/mean_terminated_length": 270.2322692871094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 85.0,
      "epoch": 0.05226666666666667,
      "grad_norm": 0.01150052435696125,
      "learning_rate": 4.194444444444445e-06,
      "loss": 0.044,
      "num_tokens": 10600066.0,
      "reward": 1.5686070919036865,
      "reward_std": 0.315078467130661,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.6947246193885803,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9390791654586792,
      "step": 49
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5348462301587302,
      "calib/avg_num_step_conf": 1.9140625,
      "calib/ece": 0.2641732283464567,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.008423859126984151,
      "calib/mean_conf": 0.2451181102362205,
      "calib/mu_c": 0.24929687500000003,
      "calib/mu_w": 0.24087301587301588,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.002677165354330709,
      "calib/std_conf": 0.09437624602489562,
      "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": 1302.0,
      "completions/max_terminated_length": 1302.0,
      "completions/mean_length": 290.515625,
      "completions/mean_terminated_length": 291.6549072265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 64.0,
      "epoch": 0.05333333333333334,
      "grad_norm": 0.012038069777190685,
      "learning_rate": 4.166666666666667e-06,
      "loss": -0.0097,
      "num_tokens": 10779798.0,
      "reward": 1.6445038318634033,
      "reward_std": 0.4098130464553833,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.6655043363571167,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9437607526779175,
      "step": 50
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6397450532724506,
      "calib/avg_num_step_conf": 2.0546875,
      "calib/ece": 0.18196850393700786,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.048555301877219625,
      "calib/mean_conf": 0.24496062992125983,
      "calib/mu_c": 0.2728703703703703,
      "calib/mu_w": 0.2243150684931507,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0008661417322834646,
      "calib/std_conf": 0.09215165535059676,
      "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": 2313.0,
      "completions/max_terminated_length": 2313.0,
      "completions/mean_length": 284.765625,
      "completions/mean_terminated_length": 284.765625,
      "completions/min_length": 66.0,
      "completions/min_terminated_length": 66.0,
      "epoch": 0.0544,
      "grad_norm": 0.012309834361076355,
      "learning_rate": 4.138888888888889e-06,
      "loss": 0.0642,
      "num_tokens": 10961994.0,
      "reward": 1.550361156463623,
      "reward_std": 0.3914136290550232,
      "rewards/accuracy_reward_step": 0.421875,
      "rewards/final_brier_reward_step": 0.7325843572616577,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.953235387802124,
      "step": 51
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6822704081632652,
      "calib/avg_num_step_conf": 2.05859375,
      "calib/ece": 0.31448412698412687,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.05291071428571423,
      "calib/mean_conf": 0.24805555555555558,
      "calib/mu_c": 0.2715714285714285,
      "calib/mu_w": 0.2186607142857143,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.003492063492063492,
      "calib/std_conf": 0.09835042672703444,
      "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": 2819.0,
      "completions/max_terminated_length": 2819.0,
      "completions/mean_length": 284.51171875,
      "completions/mean_terminated_length": 285.6274719238281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 78.0,
      "epoch": 0.055466666666666664,
      "grad_norm": 0.012607217766344547,
      "learning_rate": 4.111111111111111e-06,
      "loss": 0.0793,
      "num_tokens": 11142781.0,
      "reward": 1.711374044418335,
      "reward_std": 0.4108741581439972,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6605429649353027,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.9427658319473267,
      "step": 52
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6280425640372117,
      "calib/avg_num_step_conf": 2.05859375,
      "calib/ece": 0.28673306772908363,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.03888619854721545,
      "calib/mean_conf": 0.24314741035856574,
      "calib/mu_c": 0.2614285714285714,
      "calib/mu_w": 0.22254237288135595,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0,
      "calib/std_conf": 0.09202921642913564,
      "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": 2553.0,
      "completions/max_terminated_length": 2553.0,
      "completions/mean_length": 310.7265625,
      "completions/mean_terminated_length": 311.94512939453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 69.0,
      "epoch": 0.05653333333333333,
      "grad_norm": 0.013702677562832832,
      "learning_rate": 4.083333333333334e-06,
      "loss": 0.0836,
      "num_tokens": 11328151.0,
      "reward": 1.6691620349884033,
      "reward_std": 0.46202585101127625,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.6626269221305847,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.9437087774276733,
      "step": 53
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6580402494695559,
      "calib/avg_num_step_conf": 2.38671875,
      "calib/ece": 0.3486614173228347,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.04364174114318775,
      "calib/mean_conf": 0.25448818897637804,
      "calib/mu_c": 0.2721854304635762,
      "calib/mu_w": 0.22854368932038843,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.004330708661417323,
      "calib/std_conf": 0.08621259806561991,
      "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": 1577.0,
      "completions/max_terminated_length": 1577.0,
      "completions/mean_length": 302.8125,
      "completions/mean_terminated_length": 302.8125,
      "completions/min_length": 60.0,
      "completions/min_terminated_length": 60.0,
      "epoch": 0.0576,
      "grad_norm": 0.01350472867488861,
      "learning_rate": 4.055555555555556e-06,
      "loss": 0.0647,
      "num_tokens": 11511903.0,
      "reward": 1.7834925651550293,
      "reward_std": 0.2862309515476227,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.6496000289916992,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.9453080892562866,
      "step": 54
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6802901292967518,
      "calib/avg_num_step_conf": 1.87890625,
      "calib/ece": 0.15003906249999993,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.044869757174392955,
      "calib/mean_conf": 0.26886718750000005,
      "calib/mu_c": 0.29533333333333334,
      "calib/mu_w": 0.2504635761589404,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.004375,
      "calib/std_conf": 0.09419418698008834,
      "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": 906.0,
      "completions/max_terminated_length": 906.0,
      "completions/mean_length": 272.95703125,
      "completions/mean_terminated_length": 274.0274658203125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 70.0,
      "epoch": 0.058666666666666666,
      "grad_norm": 0.014819178730249405,
      "learning_rate": 4.027777777777779e-06,
      "loss": 0.0191,
      "num_tokens": 11689604.0,
      "reward": 1.5323195457458496,
      "reward_std": 0.35186684131622314,
      "rewards/accuracy_reward_step": 0.41015625,
      "rewards/final_brier_reward_step": 0.7427335977554321,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.949044942855835,
      "step": 55
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6308443604284815,
      "calib/avg_num_step_conf": 2.00390625,
      "calib/ece": 0.1781818181818182,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.043971014492753646,
      "calib/mean_conf": 0.28679841897233205,
      "calib/mu_c": 0.3107826086956522,
      "calib/mu_w": 0.26681159420289857,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.005217391304347826,
      "calib/std_conf": 0.09575385768898549,
      "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": 2281.0,
      "completions/max_terminated_length": 2281.0,
      "completions/mean_length": 305.90625,
      "completions/mean_terminated_length": 305.90625,
      "completions/min_length": 77.0,
      "completions/min_terminated_length": 77.0,
      "epoch": 0.05973333333333333,
      "grad_norm": 0.013396499678492546,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.04,
      "num_tokens": 11874756.0,
      "reward": 1.5809111595153809,
      "reward_std": 0.42818328738212585,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.7240632772445679,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9355190396308899,
      "step": 56
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.680624286258089,
      "calib/avg_num_step_conf": 1.69921875,
      "calib/ece": 0.251699604743083,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.05620733409465811,
      "calib/mean_conf": 0.3095652173913044,
      "calib/mu_c": 0.3342253521126761,
      "calib/mu_w": 0.278018018018018,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.08552678368165743,
      "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": 2786.0,
      "completions/max_terminated_length": 2786.0,
      "completions/mean_length": 284.23828125,
      "completions/mean_terminated_length": 284.23828125,
      "completions/min_length": 94.0,
      "completions/min_terminated_length": 94.0,
      "epoch": 0.0608,
      "grad_norm": 0.01563195139169693,
      "learning_rate": 3.972222222222223e-06,
      "loss": 0.0421,
      "num_tokens": 12054313.0,
      "reward": 1.7328195571899414,
      "reward_std": 0.4466170072555542,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.7000335454940796,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9343697428703308,
      "step": 57
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5945839874411304,
      "calib/avg_num_step_conf": 1.86328125,
      "calib/ece": 0.08501992031872514,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.02580913134484558,
      "calib/mean_conf": 0.3405577689243028,
      "calib/mu_c": 0.35567307692307687,
      "calib/mu_w": 0.3298639455782313,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.005617529880478088,
      "calib/std_conf": 0.09560297033906312,
      "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": 0.01171875,
      "completions/max_length": 1508.0,
      "completions/max_terminated_length": 1508.0,
      "completions/mean_length": 308.59765625,
      "completions/mean_terminated_length": 312.2569274902344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 104.0,
      "epoch": 0.06186666666666667,
      "grad_norm": 0.011491983197629452,
      "learning_rate": 3.944444444444445e-06,
      "loss": 0.037,
      "num_tokens": 12239634.0,
      "reward": 1.5119364261627197,
      "reward_std": 0.47350311279296875,
      "rewards/accuracy_reward_step": 0.40625,
      "rewards/final_brier_reward_step": 0.7385351657867432,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.9185852408409119,
      "step": 58
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5770900520962541,
      "calib/avg_num_step_conf": 1.6484375,
      "calib/ece": 0.08125490196078433,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.022528528901017142,
      "calib/mean_conf": 0.3803921568627451,
      "calib/mu_c": 0.39267241379310347,
      "calib/mu_w": 0.37014388489208633,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0033725490196078456,
      "calib/std_conf": 0.08218966072835476,
      "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": 950.0,
      "completions/max_terminated_length": 950.0,
      "completions/mean_length": 285.0,
      "completions/mean_terminated_length": 286.11767578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 76.0,
      "epoch": 0.06293333333333333,
      "grad_norm": 0.013964063487946987,
      "learning_rate": 3.916666666666667e-06,
      "loss": 0.0,
      "num_tokens": 12418842.0,
      "reward": 1.5842854976654053,
      "reward_std": 0.39020073413848877,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.7388551235198975,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9107863903045654,
      "step": 59
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6187847027263085,
      "calib/avg_num_step_conf": 1.96484375,
      "calib/ece": 0.09157480314960624,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.03150040551500405,
      "calib/mean_conf": 0.3807874015748031,
      "calib/mu_c": 0.39777777777777784,
      "calib/mu_w": 0.3662773722627738,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.005866141732283469,
      "calib/std_conf": 0.08282993010738718,
      "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": 1115.0,
      "completions/max_terminated_length": 1115.0,
      "completions/mean_length": 322.37890625,
      "completions/mean_terminated_length": 323.6431579589844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 107.0,
      "epoch": 0.064,
      "grad_norm": 0.013174435123801231,
      "learning_rate": 3.88888888888889e-06,
      "loss": 0.0401,
      "num_tokens": 12610227.0,
      "reward": 1.5859081745147705,
      "reward_std": 0.4918421506881714,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.7384917736053467,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.9098280668258667,
      "step": 60
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6021793852676206,
      "calib/avg_num_step_conf": 2.19140625,
      "calib/ece": 0.17679999999999996,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.020727344992050867,
      "calib/mean_conf": 0.4208,
      "calib/mu_c": 0.4292567567567568,
      "calib/mu_w": 0.4085294117647059,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0028,
      "calib/std_conf": 0.07755875192394474,
      "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": 851.0,
      "completions/max_terminated_length": 851.0,
      "completions/mean_length": 296.55859375,
      "completions/mean_terminated_length": 298.8937072753906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 91.0,
      "epoch": 0.06506666666666666,
      "grad_norm": 0.014439865946769714,
      "learning_rate": 3.861111111111112e-06,
      "loss": -0.0231,
      "num_tokens": 12790210.0,
      "reward": 1.7645703554153442,
      "reward_std": 0.41605138778686523,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7159687876701355,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8969999551773071,
      "step": 61
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5541274201988488,
      "calib/avg_num_step_conf": 2.1953125,
      "calib/ece": 0.05904382470119514,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.010890894819466312,
      "calib/mean_conf": 0.42573705179282867,
      "calib/mu_c": 0.4321153846153847,
      "calib/mu_w": 0.42122448979591837,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.03521912350597604,
      "calib/std_conf": 0.0894146223105681,
      "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": 1442.0,
      "completions/max_terminated_length": 1442.0,
      "completions/mean_length": 363.63671875,
      "completions/mean_terminated_length": 365.0627746582031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.06613333333333334,
      "grad_norm": 0.012277968227863312,
      "learning_rate": 3.833333333333334e-06,
      "loss": 0.0016,
      "num_tokens": 12990381.0,
      "reward": 1.510021686553955,
      "reward_std": 0.48530229926109314,
      "rewards/accuracy_reward_step": 0.40625,
      "rewards/final_brier_reward_step": 0.7397617101669312,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.901887834072113,
      "step": 62
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5801273634453782,
      "calib/avg_num_step_conf": 2.19140625,
      "calib/ece": 0.0829554655870445,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.029581144957983252,
      "calib/mean_conf": 0.43315789473684213,
      "calib/mu_c": 0.4484873949579833,
      "calib/mu_w": 0.41890625000000004,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.017165991902834014,
      "calib/std_conf": 0.08432962226115784,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2682.0,
      "completions/max_terminated_length": 2682.0,
      "completions/mean_length": 399.46875,
      "completions/mean_terminated_length": 401.0353088378906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 103.0,
      "epoch": 0.0672,
      "grad_norm": 0.010792006738483906,
      "learning_rate": 3.8055555555555556e-06,
      "loss": 0.0388,
      "num_tokens": 13201285.0,
      "reward": 1.5720921754837036,
      "reward_std": 0.5513401031494141,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.7202058434486389,
      "rewards/format_reward_step": 0.953125,
      "rewards/stepwise_brier_reward": 0.8728504180908203,
      "step": 63
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5626911314984709,
      "calib/avg_num_step_conf": 2.18359375,
      "calib/ece": 0.09184426229508194,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.033306829765545276,
      "calib/mean_conf": 0.46356557377049185,
      "calib/mu_c": 0.47844444444444445,
      "calib/mu_w": 0.4451376146788992,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.0010655737704918034,
      "calib/std_conf": 0.09594222029173002,
      "calib/step_conf_rate": 0.9765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 3028.0,
      "completions/max_terminated_length": 3028.0,
      "completions/mean_length": 380.2890625,
      "completions/mean_terminated_length": 381.7804260253906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 127.0,
      "epoch": 0.06826666666666667,
      "grad_norm": 0.012527639046311378,
      "learning_rate": 3.777777777777778e-06,
      "loss": 0.0231,
      "num_tokens": 13402415.0,
      "reward": 1.65690279006958,
      "reward_std": 0.5391913652420044,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.707330048084259,
      "rewards/format_reward_step": 0.94140625,
      "rewards/stepwise_brier_reward": 0.8499686121940613,
      "step": 64
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5567651142244099,
      "calib/avg_num_step_conf": 2.4453125,
      "calib/ece": 0.08656126482213439,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.013242458664647128,
      "calib/mean_conf": 0.47604743083003953,
      "calib/mu_c": 0.48201438848920863,
      "calib/mu_w": 0.4687719298245615,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.00660079051383399,
      "calib/std_conf": 0.08234333383939904,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1972.0,
      "completions/max_terminated_length": 1972.0,
      "completions/mean_length": 341.32421875,
      "completions/mean_terminated_length": 342.6627502441406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 123.0,
      "epoch": 0.06933333333333333,
      "grad_norm": 0.013505609706044197,
      "learning_rate": 3.7500000000000005e-06,
      "loss": -0.012,
      "num_tokens": 13594818.0,
      "reward": 1.7095803022384644,
      "reward_std": 0.30704957246780396,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.7325637340545654,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8870077133178711,
      "step": 65
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.6078224101479915,
      "calib/avg_num_step_conf": 2.47265625,
      "calib/ece": 0.039399999999999956,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.03412582484464094,
      "calib/mean_conf": 0.47892000000000007,
      "calib/mu_c": 0.49652892561983475,
      "calib/mu_w": 0.4624031007751938,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.017159999999999998,
      "calib/std_conf": 0.10243453323952818,
      "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": 0.0078125,
      "completions/max_length": 2699.0,
      "completions/max_terminated_length": 2699.0,
      "completions/mean_length": 424.5078125,
      "completions/mean_terminated_length": 427.85040283203125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 133.0,
      "epoch": 0.0704,
      "grad_norm": 0.012787997722625732,
      "learning_rate": 3.7222222222222225e-06,
      "loss": 0.0821,
      "num_tokens": 13809844.0,
      "reward": 1.5854953527450562,
      "reward_std": 0.48220348358154297,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.7278234362602234,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.8641577959060669,
      "step": 66
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6565740622199463,
      "calib/avg_num_step_conf": 2.828125,
      "calib/ece": 0.05438735177865616,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.06275380873127645,
      "calib/mean_conf": 0.5335968379446641,
      "calib/mu_c": 0.5601369863013699,
      "calib/mu_w": 0.4973831775700934,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.005454545454545452,
      "calib/std_conf": 0.11480223375604197,
      "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": 1222.0,
      "completions/max_terminated_length": 1222.0,
      "completions/mean_length": 399.72265625,
      "completions/mean_terminated_length": 402.8700866699219,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 172.0,
      "epoch": 0.07146666666666666,
      "grad_norm": 0.011641331948339939,
      "learning_rate": 3.694444444444445e-06,
      "loss": -0.0325,
      "num_tokens": 14017181.0,
      "reward": 1.7519131898880005,
      "reward_std": 0.28649717569351196,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.7556644678115845,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8691759705543518,
      "step": 67
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5697100362454692,
      "calib/avg_num_step_conf": 2.859375,
      "calib/ece": 0.1192094861660079,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.027567804024496767,
      "calib/mean_conf": 0.5072727272727273,
      "calib/mu_c": 0.5211111111111111,
      "calib/mu_w": 0.4935433070866143,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.06422924901185773,
      "calib/std_conf": 0.10645227886321025,
      "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": 1862.0,
      "completions/max_terminated_length": 1862.0,
      "completions/mean_length": 415.78515625,
      "completions/mean_terminated_length": 419.0590515136719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.07253333333333334,
      "grad_norm": 0.01155012845993042,
      "learning_rate": 3.6666666666666666e-06,
      "loss": 0.0184,
      "num_tokens": 14227710.0,
      "reward": 1.6360442638397217,
      "reward_std": 0.3913213014602661,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.7412973046302795,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8810046911239624,
      "step": 68
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5540638606676344,
      "calib/avg_num_step_conf": 3.01953125,
      "calib/ece": 0.08558232931726908,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.008032128514056224,
      "calib/gap": 0.010903813167964183,
      "calib/mean_conf": 0.511285140562249,
      "calib/mu_c": 0.5175471698113208,
      "calib/mu_w": 0.5066433566433566,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.08558232931726908,
      "calib/std_conf": 0.1164637723119289,
      "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": 0.01171875,
      "completions/max_length": 2306.0,
      "completions/max_terminated_length": 2306.0,
      "completions/mean_length": 479.484375,
      "completions/mean_terminated_length": 485.16998291015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 153.0,
      "epoch": 0.0736,
      "grad_norm": 0.010634348727762699,
      "learning_rate": 3.638888888888889e-06,
      "loss": 0.0674,
      "num_tokens": 14454954.0,
      "reward": 1.4911199808120728,
      "reward_std": 0.4545786380767822,
      "rewards/accuracy_reward_step": 0.4140625,
      "rewards/final_brier_reward_step": 0.706221878528595,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.8598206043243408,
      "step": 69
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6284987277353689,
      "calib/avg_num_step_conf": 3.15625,
      "calib/ece": 0.029043824701195244,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.05819147582697187,
      "calib/mean_conf": 0.5035458167330678,
      "calib/mu_c": 0.5339166666666666,
      "calib/mu_w": 0.4757251908396947,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.027250996015936294,
      "calib/std_conf": 0.11774346181943768,
      "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": 2717.0,
      "completions/max_terminated_length": 2717.0,
      "completions/mean_length": 476.625,
      "completions/mean_terminated_length": 478.494140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 142.0,
      "epoch": 0.07466666666666667,
      "grad_norm": 0.010567680932581425,
      "learning_rate": 3.6111111111111115e-06,
      "loss": 0.0672,
      "num_tokens": 14683962.0,
      "reward": 1.5975029468536377,
      "reward_std": 0.410000205039978,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.7483577728271484,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8760287165641785,
      "step": 70
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6081660986999874,
      "calib/avg_num_step_conf": 3.4921875,
      "calib/ece": 0.10715415019762846,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.03986684336741142,
      "calib/mean_conf": 0.5569565217391306,
      "calib/mu_c": 0.5788596491228071,
      "calib/mu_w": 0.5389928057553957,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.10675889328063241,
      "calib/std_conf": 0.10736433707912163,
      "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": 2378.0,
      "completions/max_terminated_length": 2378.0,
      "completions/mean_length": 473.33984375,
      "completions/mean_terminated_length": 475.19610595703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 182.0,
      "epoch": 0.07573333333333333,
      "grad_norm": 0.0112195685505867,
      "learning_rate": 3.5833333333333335e-06,
      "loss": 0.021,
      "num_tokens": 14909545.0,
      "reward": 1.567267894744873,
      "reward_std": 0.528032660484314,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.7405582070350647,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8800756931304932,
      "step": 71
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6617248123146804,
      "calib/avg_num_step_conf": 3.53515625,
      "calib/ece": 0.05758730158730162,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.015873015873015872,
      "calib/gap": 0.06981692006813456,
      "calib/mean_conf": 0.5544920634920636,
      "calib/mu_c": 0.5880152671755725,
      "calib/mu_w": 0.518198347107438,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.04611904761904761,
      "calib/std_conf": 0.14388952833887278,
      "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": 2528.0,
      "completions/max_terminated_length": 2528.0,
      "completions/mean_length": 461.5859375,
      "completions/mean_terminated_length": 461.5859375,
      "completions/min_length": 174.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.0768,
      "grad_norm": 0.012084675952792168,
      "learning_rate": 3.555555555555556e-06,
      "loss": 0.0357,
      "num_tokens": 15132119.0,
      "reward": 1.665956974029541,
      "reward_std": 0.4454992413520813,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.7514150738716125,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8733506798744202,
      "step": 72
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5393425605536333,
      "calib/avg_num_step_conf": 3.7578125,
      "calib/ece": 0.1385098039215686,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.00784313725490196,
      "calib/gap": 0.01276470588235279,
      "calib/mean_conf": 0.5752156862745098,
      "calib/mu_c": 0.5794705882352942,
      "calib/mu_w": 0.5667058823529414,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.02352941176470588,
      "calib/std_conf": 0.1267506914898553,
      "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": 2663.0,
      "completions/max_terminated_length": 2663.0,
      "completions/mean_length": 454.59765625,
      "completions/mean_terminated_length": 454.59765625,
      "completions/min_length": 163.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.07786666666666667,
      "grad_norm": 0.010945611633360386,
      "learning_rate": 3.5277777777777784e-06,
      "loss": 0.0264,
      "num_tokens": 15355528.0,
      "reward": 1.8979418277740479,
      "reward_std": 0.4913485646247864,
      "rewards/accuracy_reward_step": 0.6640625,
      "rewards/final_brier_reward_step": 0.7496898174285889,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8811398148536682,
      "step": 73
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5702736318407959,
      "calib/avg_num_step_conf": 3.68359375,
      "calib/ece": 0.061653225806451606,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.016129032258064516,
      "calib/gap": 0.04322204765645454,
      "calib/mean_conf": 0.5732661290322582,
      "calib/mu_c": 0.593134328358209,
      "calib/mu_w": 0.5499122807017545,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.96875,
      "calib/pce": 0.047298387096774176,
      "calib/std_conf": 0.12196731508700083,
      "calib/step_conf_rate": 0.96875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 3026.0,
      "completions/max_terminated_length": 3026.0,
      "completions/mean_length": 460.96875,
      "completions/mean_terminated_length": 466.434814453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.07893333333333333,
      "grad_norm": 0.0119440583512187,
      "learning_rate": 3.5e-06,
      "loss": 0.0263,
      "num_tokens": 15577464.0,
      "reward": 1.656386375427246,
      "reward_std": 0.3758211135864258,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.7251109480857849,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.8457469344139099,
      "step": 74
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5690184049079754,
      "calib/avg_num_step_conf": 4.08984375,
      "calib/ece": 0.05826086956521743,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.003952569169960474,
      "calib/gap": 0.033805725971370104,
      "calib/mean_conf": 0.6194466403162056,
      "calib/mu_c": 0.6314723926380369,
      "calib/mu_w": 0.5976666666666668,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.01671936758893279,
      "calib/std_conf": 0.12093055743654747,
      "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": 1472.0,
      "completions/max_terminated_length": 1472.0,
      "completions/mean_length": 442.0859375,
      "completions/mean_terminated_length": 443.81964111328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 177.0,
      "epoch": 0.08,
      "grad_norm": 0.011777788400650024,
      "learning_rate": 3.4722222222222224e-06,
      "loss": -0.0068,
      "num_tokens": 15795390.0,
      "reward": 1.8508899211883545,
      "reward_std": 0.4994873106479645,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.7589175701141357,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8555798530578613,
      "step": 75
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6155647382920111,
      "calib/avg_num_step_conf": 4.125,
      "calib/ece": 0.050395256916996076,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.023715415019762844,
      "calib/gap": 0.04662878787878788,
      "calib/mean_conf": 0.6134782608695653,
      "calib/mu_c": 0.6296969696969698,
      "calib/mu_w": 0.5830681818181819,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0058498023715415045,
      "calib/std_conf": 0.1204264348357623,
      "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": 3035.0,
      "completions/max_terminated_length": 3035.0,
      "completions/mean_length": 487.5625,
      "completions/mean_terminated_length": 489.47454833984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 198.0,
      "epoch": 0.08106666666666666,
      "grad_norm": 0.011100761592388153,
      "learning_rate": 3.444444444444445e-06,
      "loss": 0.0342,
      "num_tokens": 16023262.0,
      "reward": 1.8673896789550781,
      "reward_std": 0.3366937041282654,
      "rewards/accuracy_reward_step": 0.64453125,
      "rewards/final_brier_reward_step": 0.7664664387702942,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8671547174453735,
      "step": 76
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5162422037422038,
      "calib/avg_num_step_conf": 4.3671875,
      "calib/ece": 0.12722222222222224,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.031746031746031744,
      "calib/gap": 0.010109147609147384,
      "calib/mean_conf": 0.6042063492063492,
      "calib/mu_c": 0.6083783783783783,
      "calib/mu_w": 0.5982692307692309,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.07206349206349204,
      "calib/std_conf": 0.13086921360961365,
      "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": 2711.0,
      "completions/max_terminated_length": 2711.0,
      "completions/mean_length": 515.22265625,
      "completions/mean_terminated_length": 521.33203125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 199.0,
      "epoch": 0.08213333333333334,
      "grad_norm": 0.010618972592055798,
      "learning_rate": 3.416666666666667e-06,
      "loss": -0.0007,
      "num_tokens": 16259823.0,
      "reward": 1.756075143814087,
      "reward_std": 0.45902228355407715,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.7307425737380981,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8638701438903809,
      "step": 77
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5577266922094508,
      "calib/avg_num_step_conf": 4.9140625,
      "calib/ece": 0.09317667984189726,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.03162055335968379,
      "calib/gap": 0.011251621966794412,
      "calib/mean_conf": 0.6279300395256917,
      "calib/mu_c": 0.6327331034482759,
      "calib/mu_w": 0.6214814814814815,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0739920948616601,
      "calib/std_conf": 0.15163950974534543,
      "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": 1562.0,
      "completions/max_terminated_length": 1562.0,
      "completions/mean_length": 539.32421875,
      "completions/mean_terminated_length": 541.4392700195312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 221.0,
      "epoch": 0.0832,
      "grad_norm": 0.010913941077888012,
      "learning_rate": 3.3888888888888893e-06,
      "loss": 0.0031,
      "num_tokens": 16505914.0,
      "reward": 1.7397754192352295,
      "reward_std": 0.5139069557189941,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.7233127355575562,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8686016798019409,
      "step": 78
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6401579986833443,
      "calib/avg_num_step_conf": 4.875,
      "calib/ece": 0.09833992094861659,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.07114624505928854,
      "calib/gap": 0.06974522712310727,
      "calib/mean_conf": 0.6607905138339921,
      "calib/mu_c": 0.6878064516129032,
      "calib/mu_w": 0.618061224489796,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0732411067193676,
      "calib/std_conf": 0.1553936973459181,
      "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": 1514.0,
      "completions/max_terminated_length": 1514.0,
      "completions/mean_length": 532.37109375,
      "completions/mean_terminated_length": 534.4588623046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 223.0,
      "epoch": 0.08426666666666667,
      "grad_norm": 0.010380618274211884,
      "learning_rate": 3.3611111111111117e-06,
      "loss": 0.0087,
      "num_tokens": 16748577.0,
      "reward": 1.799513339996338,
      "reward_std": 0.4986400902271271,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.7558000087738037,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8485037088394165,
      "step": 79
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6805725971370142,
      "calib/avg_num_step_conf": 5.2421875,
      "calib/ece": 0.07553359683794474,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.1067193675889328,
      "calib/gap": 0.08949488752556234,
      "calib/mean_conf": 0.70399209486166,
      "calib/mu_c": 0.7358282208588957,
      "calib/mu_w": 0.6463333333333333,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.06762845849802375,
      "calib/std_conf": 0.14632072607194188,
      "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": 1265.0,
      "completions/max_terminated_length": 1265.0,
      "completions/mean_length": 486.0078125,
      "completions/mean_terminated_length": 487.91375732421875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 247.0,
      "epoch": 0.08533333333333333,
      "grad_norm": 0.01169226597994566,
      "learning_rate": 3.3333333333333333e-06,
      "loss": -0.0085,
      "num_tokens": 16975155.0,
      "reward": 1.8564797639846802,
      "reward_std": 0.43828481435775757,
      "rewards/accuracy_reward_step": 0.640625,
      "rewards/final_brier_reward_step": 0.7751379013061523,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8382811546325684,
      "step": 80
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5273809523809525,
      "calib/avg_num_step_conf": 4.95703125,
      "calib/ece": 0.1275708502024292,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.032388663967611336,
      "calib/gap": 0.005582312925169974,
      "calib/mean_conf": 0.6635222672064778,
      "calib/mu_c": 0.6657823129251701,
      "calib/mu_w": 0.6602000000000001,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.09797570850202432,
      "calib/std_conf": 0.13415254516637715,
      "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": 2979.0,
      "completions/max_terminated_length": 2979.0,
      "completions/mean_length": 538.74609375,
      "completions/mean_terminated_length": 545.1343994140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 224.0,
      "epoch": 0.0864,
      "grad_norm": 0.011217792518436909,
      "learning_rate": 3.3055555555555558e-06,
      "loss": 0.0133,
      "num_tokens": 17219322.0,
      "reward": 1.7309539318084717,
      "reward_std": 0.40973007678985596,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.7130863666534424,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.8357298374176025,
      "step": 81
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5857142857142857,
      "calib/avg_num_step_conf": 5.0625,
      "calib/ece": 0.07434782608695656,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.03162055335968379,
      "calib/gap": 0.03528442728442727,
      "calib/mean_conf": 0.6445454545454545,
      "calib/mu_c": 0.6591891891891892,
      "calib/mu_w": 0.623904761904762,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0669565217391304,
      "calib/std_conf": 0.1319431055862712,
      "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": 2933.0,
      "completions/max_terminated_length": 2933.0,
      "completions/mean_length": 523.328125,
      "completions/mean_terminated_length": 523.328125,
      "completions/min_length": 253.0,
      "completions/min_terminated_length": 253.0,
      "epoch": 0.08746666666666666,
      "grad_norm": 0.010938949882984161,
      "learning_rate": 3.277777777777778e-06,
      "loss": 0.0451,
      "num_tokens": 17458846.0,
      "reward": 1.768110752105713,
      "reward_std": 0.41400328278541565,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7428593635559082,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8686458468437195,
      "step": 82
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.527170868347339,
      "calib/avg_num_step_conf": 5.05859375,
      "calib/ece": 0.12930314960629918,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.05511811023622047,
      "calib/gap": 0.020962029256146764,
      "calib/mean_conf": 0.6515866141732283,
      "calib/mu_c": 0.6614074074074073,
      "calib/mu_w": 0.6404453781512606,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.12469685039370072,
      "calib/std_conf": 0.14549304708457206,
      "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": 2630.0,
      "completions/max_terminated_length": 2630.0,
      "completions/mean_length": 587.734375,
      "completions/mean_terminated_length": 590.0392456054688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 274.0,
      "epoch": 0.08853333333333334,
      "grad_norm": 0.010783905163407326,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.0689,
      "num_tokens": 17716570.0,
      "reward": 1.6735692024230957,
      "reward_std": 0.3873692750930786,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.7130337357521057,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8562429547309875,
      "step": 83
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.622803666921314,
      "calib/avg_num_step_conf": 5.4921875,
      "calib/ece": 0.1501179282868526,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.07569721115537849,
      "calib/gap": 0.06809253883371513,
      "calib/mean_conf": 0.6459776892430279,
      "calib/mu_c": 0.678260606060606,
      "calib/mu_w": 0.6101680672268909,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.13509960159362547,
      "calib/std_conf": 0.16050626125057085,
      "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": 1755.0,
      "completions/max_terminated_length": 1755.0,
      "completions/mean_length": 501.8515625,
      "completions/mean_terminated_length": 507.8023986816406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 182.0,
      "epoch": 0.0896,
      "grad_norm": 0.012522642500698566,
      "learning_rate": 3.2222222222222227e-06,
      "loss": -0.0018,
      "num_tokens": 17950964.0,
      "reward": 1.6547077894210815,
      "reward_std": 0.45172953605651855,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.7247183322906494,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.8550503253936768,
      "step": 84
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6427714646464646,
      "calib/avg_num_step_conf": 5.515625,
      "calib/ece": 0.12132539682539682,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.04365079365079365,
      "calib/gap": 0.07116363636363643,
      "calib/mean_conf": 0.6216428571428572,
      "calib/mu_c": 0.655530303030303,
      "calib/mu_w": 0.5843666666666666,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.10957936507936504,
      "calib/std_conf": 0.15777913482297817,
      "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": 2914.0,
      "completions/max_terminated_length": 2914.0,
      "completions/mean_length": 558.80078125,
      "completions/mean_terminated_length": 563.2008056640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 205.0,
      "epoch": 0.09066666666666667,
      "grad_norm": 0.0107113691046834,
      "learning_rate": 3.1944444444444443e-06,
      "loss": 0.0019,
      "num_tokens": 18201841.0,
      "reward": 1.6682491302490234,
      "reward_std": 0.46407395601272583,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.7398586273193359,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8706381320953369,
      "step": 85
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6746666666666665,
      "calib/avg_num_step_conf": 5.3671875,
      "calib/ece": 0.11095617529880475,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.00796812749003984,
      "calib/gap": 0.08622095238095251,
      "calib/mean_conf": 0.6094422310756973,
      "calib/mu_c": 0.6523809523809525,
      "calib/mu_w": 0.56616,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.10920318725099598,
      "calib/std_conf": 0.1339578620841114,
      "calib/step_conf_rate": 0.97265625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2788.0,
      "completions/max_terminated_length": 2788.0,
      "completions/mean_length": 577.26953125,
      "completions/mean_terminated_length": 577.26953125,
      "completions/min_length": 241.0,
      "completions/min_terminated_length": 241.0,
      "epoch": 0.09173333333333333,
      "grad_norm": 0.010854917578399181,
      "learning_rate": 3.1666666666666667e-06,
      "loss": 0.0634,
      "num_tokens": 18455134.0,
      "reward": 1.6256598234176636,
      "reward_std": 0.42664429545402527,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.7426445484161377,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.8693699836730957,
      "step": 86
    },
    {
      "calib/answer_extract_rate": 0.9453125,
      "calib/auroc": 0.562534435261708,
      "calib/avg_num_step_conf": 6.09765625,
      "calib/ece": 0.12314049586776857,
      "calib/final_conf_rate": 0.9453125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.02066115702479339,
      "calib/gap": 0.026233766233766276,
      "calib/mean_conf": 0.6025619834710744,
      "calib/mu_c": 0.6109090909090908,
      "calib/mu_w": 0.5846753246753246,
      "calib/nonempty_final_conf_rate": 0.9453125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.021942148760330567,
      "calib/std_conf": 0.1339392296245931,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 2985.0,
      "completions/max_terminated_length": 2985.0,
      "completions/mean_length": 551.4140625,
      "completions/mean_terminated_length": 564.6480102539062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 180.0,
      "epoch": 0.0928,
      "grad_norm": 0.011104823090136051,
      "learning_rate": 3.138888888888889e-06,
      "loss": -0.0038,
      "num_tokens": 18701792.0,
      "reward": 1.8306856155395508,
      "reward_std": 0.44113269448280334,
      "rewards/accuracy_reward_step": 0.64453125,
      "rewards/final_brier_reward_step": 0.726388692855835,
      "rewards/format_reward_step": 0.94140625,
      "rewards/stepwise_brier_reward": 0.8463533520698547,
      "step": 87
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5883070209215762,
      "calib/avg_num_step_conf": 6.16015625,
      "calib/ece": 0.13539525691699608,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.03557312252964427,
      "calib/gap": 0.03860672570915158,
      "calib/mean_conf": 0.5782806324110672,
      "calib/mu_c": 0.5944557823129252,
      "calib/mu_w": 0.5558490566037736,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.06632411067193678,
      "calib/std_conf": 0.1581033905867141,
      "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": 2336.0,
      "completions/max_terminated_length": 2336.0,
      "completions/mean_length": 573.1015625,
      "completions/mean_terminated_length": 573.1015625,
      "completions/min_length": 303.0,
      "completions/min_terminated_length": 303.0,
      "epoch": 0.09386666666666667,
      "grad_norm": 0.010853197425603867,
      "learning_rate": 3.1111111111111116e-06,
      "loss": 0.0224,
      "num_tokens": 18958354.0,
      "reward": 1.760920763015747,
      "reward_std": 0.4879763424396515,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.7380582094192505,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8915627002716064,
      "step": 88
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.6754781788793104,
      "calib/avg_num_step_conf": 6.64453125,
      "calib/ece": 0.1219959016393443,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.93359375,
      "calib/frac_conf_gt_0.9": 0.04918032786885246,
      "calib/gap": 0.08492914870689638,
      "calib/mean_conf": 0.6147254098360656,
      "calib/mu_c": 0.6551015625,
      "calib/mu_w": 0.5701724137931036,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.9609375,
      "calib/pce": 0.10606557377049185,
      "calib/std_conf": 0.15300354386845894,
      "calib/step_conf_rate": 0.9609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 2446.0,
      "completions/max_terminated_length": 2446.0,
      "completions/mean_length": 618.3125,
      "completions/mean_terminated_length": 630.6295166015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.09493333333333333,
      "grad_norm": 0.010586491785943508,
      "learning_rate": 3.0833333333333336e-06,
      "loss": -0.0082,
      "num_tokens": 19225530.0,
      "reward": 1.6018019914627075,
      "reward_std": 0.37154436111450195,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.7094407677650452,
      "rewards/format_reward_step": 0.93359375,
      "rewards/stepwise_brier_reward": 0.8305795192718506,
      "step": 89
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5019526248399488,
      "calib/avg_num_step_conf": 7.43359375,
      "calib/ece": 0.14103174603174606,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.06349206349206349,
      "calib/gap": -0.006066581306017893,
      "calib/mean_conf": 0.6491269841269841,
      "calib/mu_c": 0.6464788732394365,
      "calib/mu_w": 0.6525454545454544,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.11333333333333337,
      "calib/std_conf": 0.14228029479773335,
      "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": 2055.0,
      "completions/max_terminated_length": 2055.0,
      "completions/mean_length": 583.7578125,
      "completions/mean_terminated_length": 588.3543090820312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 219.0,
      "epoch": 0.096,
      "grad_norm": 0.00980350561439991,
      "learning_rate": 3.055555555555556e-06,
      "loss": 0.0095,
      "num_tokens": 19478292.0,
      "reward": 1.7174973487854004,
      "reward_std": 0.47956204414367676,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.7093156576156616,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8716117143630981,
      "step": 90
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5488736793142402,
      "calib/avg_num_step_conf": 7.3046875,
      "calib/ece": 0.07418666666666668,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.024,
      "calib/gap": 0.02385762066139474,
      "calib/mean_conf": 0.6261333333333333,
      "calib/mu_c": 0.6357718120805368,
      "calib/mu_w": 0.611914191419142,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.05216,
      "calib/std_conf": 0.12623427954930996,
      "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": 2756.0,
      "completions/max_terminated_length": 2756.0,
      "completions/mean_length": 626.83203125,
      "completions/mean_terminated_length": 631.7677001953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 273.0,
      "epoch": 0.09706666666666666,
      "grad_norm": 0.009960450232028961,
      "learning_rate": 3.0277777777777776e-06,
      "loss": 0.0244,
      "num_tokens": 19746473.0,
      "reward": 1.7600347995758057,
      "reward_std": 0.4663000702857971,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7324761152267456,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.8701634407043457,
      "step": 91
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6787819889860706,
      "calib/avg_num_step_conf": 7.16015625,
      "calib/ece": 0.0928571428571429,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.04365079365079365,
      "calib/gap": 0.08712925170068031,
      "calib/mean_conf": 0.6153968253968256,
      "calib/mu_c": 0.6517006802721088,
      "calib/mu_w": 0.5645714285714285,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.062460317460317466,
      "calib/std_conf": 0.15089495344128934,
      "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": 1871.0,
      "completions/max_terminated_length": 1871.0,
      "completions/mean_length": 561.0859375,
      "completions/mean_terminated_length": 563.2863159179688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 255.0,
      "epoch": 0.09813333333333334,
      "grad_norm": 0.010641069151461124,
      "learning_rate": 3e-06,
      "loss": 0.0441,
      "num_tokens": 19996831.0,
      "reward": 1.762449026107788,
      "reward_std": 0.4561864733695984,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.7633843421936035,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8723491430282593,
      "step": 92
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5030359795878819,
      "calib/avg_num_step_conf": 7.390625,
      "calib/ece": 0.138484,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.072,
      "calib/gap": 0.0032170402428783884,
      "calib/mean_conf": 0.657524,
      "calib/mu_c": 0.658978102189781,
      "calib/mu_w": 0.6557610619469026,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.124004,
      "calib/std_conf": 0.14997290896691975,
      "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": 1824.0,
      "completions/max_terminated_length": 1824.0,
      "completions/mean_length": 612.09375,
      "completions/mean_terminated_length": 619.351806640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 254.0,
      "epoch": 0.0992,
      "grad_norm": 0.010396548546850681,
      "learning_rate": 2.9722222222222225e-06,
      "loss": 0.0022,
      "num_tokens": 20259303.0,
      "reward": 1.6816561222076416,
      "reward_std": 0.5101436376571655,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.702549159526825,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8600127696990967,
      "step": 93
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6179757343550447,
      "calib/avg_num_step_conf": 6.43359375,
      "calib/ece": 0.08928286852589642,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.01195219123505976,
      "calib/gap": 0.05509259259259247,
      "calib/mean_conf": 0.6271314741035857,
      "calib/mu_c": 0.6525925925925925,
      "calib/mu_w": 0.5975,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.08928286852589642,
      "calib/std_conf": 0.13229712215801656,
      "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": 2211.0,
      "completions/max_terminated_length": 2211.0,
      "completions/mean_length": 548.9296875,
      "completions/mean_terminated_length": 553.251953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 278.0,
      "epoch": 0.10026666666666667,
      "grad_norm": 0.012168372049927711,
      "learning_rate": 2.944444444444445e-06,
      "loss": 0.0315,
      "num_tokens": 20508509.0,
      "reward": 1.681666612625122,
      "reward_std": 0.43118709325790405,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.7372269630432129,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8722525835037231,
      "step": 94
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6383513931888545,
      "calib/avg_num_step_conf": 7.109375,
      "calib/ece": 0.08560236220472447,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.07480314960629922,
      "calib/gap": 0.08595614035087717,
      "calib/mean_conf": 0.6512716535433071,
      "calib/mu_c": 0.6857894736842106,
      "calib/mu_w": 0.5998333333333334,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.06922440944881895,
      "calib/std_conf": 0.17228837590525065,
      "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": 1856.0,
      "completions/max_terminated_length": 1856.0,
      "completions/mean_length": 599.09765625,
      "completions/mean_terminated_length": 601.4470825195312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 245.0,
      "epoch": 0.10133333333333333,
      "grad_norm": 0.009556005708873272,
      "learning_rate": 2.916666666666667e-06,
      "loss": 0.0179,
      "num_tokens": 20768006.0,
      "reward": 1.7951675653457642,
      "reward_std": 0.4173787236213684,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.7625200748443604,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8712753057479858,
      "step": 95
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6868736714242333,
      "calib/avg_num_step_conf": 7.08203125,
      "calib/ece": 0.07134920634920633,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.09126984126984126,
      "calib/gap": 0.10139690252049816,
      "calib/mean_conf": 0.6749999999999999,
      "calib/mu_c": 0.7047752808988764,
      "calib/mu_w": 0.6033783783783783,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.019999999999999966,
      "calib/std_conf": 0.15763505169793146,
      "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": 2668.0,
      "completions/max_terminated_length": 2668.0,
      "completions/mean_length": 561.0078125,
      "completions/mean_terminated_length": 563.2078857421875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 227.0,
      "epoch": 0.1024,
      "grad_norm": 0.01089701522141695,
      "learning_rate": 2.888888888888889e-06,
      "loss": 0.0017,
      "num_tokens": 21017440.0,
      "reward": 1.9514626264572144,
      "reward_std": 0.3948931396007538,
      "rewards/accuracy_reward_step": 0.69921875,
      "rewards/final_brier_reward_step": 0.7932445406913757,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8563559055328369,
      "step": 96
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.626808347202663,
      "calib/avg_num_step_conf": 6.82421875,
      "calib/ece": 0.09111857707509882,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.08695652173913043,
      "calib/gap": 0.06799491102291633,
      "calib/mean_conf": 0.6514110671936758,
      "calib/mu_c": 0.6801678082191781,
      "calib/mu_w": 0.6121728971962618,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.08272727272727275,
      "calib/std_conf": 0.16670608854078667,
      "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": 2574.0,
      "completions/max_terminated_length": 2574.0,
      "completions/mean_length": 601.69140625,
      "completions/mean_terminated_length": 601.69140625,
      "completions/min_length": 242.0,
      "completions/min_terminated_length": 242.0,
      "epoch": 0.10346666666666667,
      "grad_norm": 0.009758230298757553,
      "learning_rate": 2.861111111111111e-06,
      "loss": 0.0657,
      "num_tokens": 21276545.0,
      "reward": 1.754123568534851,
      "reward_std": 0.42388081550598145,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.7469562292098999,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8711007833480835,
      "step": 97
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.45511760966306425,
      "calib/avg_num_step_conf": 6.828125,
      "calib/ece": 0.16434782608695653,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.09486166007905138,
      "calib/gap": -0.023517482517482535,
      "calib/mean_conf": 0.7031620553359685,
      "calib/mu_c": 0.6929370629370629,
      "calib/mu_w": 0.7164545454545455,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.15114624505928853,
      "calib/std_conf": 0.14104172504784265,
      "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": 2613.0,
      "completions/max_terminated_length": 2613.0,
      "completions/mean_length": 591.82421875,
      "completions/mean_terminated_length": 594.1451416015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 271.0,
      "epoch": 0.10453333333333334,
      "grad_norm": 0.0114395497366786,
      "learning_rate": 2.8333333333333335e-06,
      "loss": 0.0575,
      "num_tokens": 21534236.0,
      "reward": 1.7173734903335571,
      "reward_std": 0.49968981742858887,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.6920980215072632,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8570834398269653,
      "step": 98
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5435785864978904,
      "calib/avg_num_step_conf": 7.51171875,
      "calib/ece": 0.3524015748031496,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.17716535433070865,
      "calib/gap": 0.028568037974683547,
      "calib/mean_conf": 0.7303543307086614,
      "calib/mu_c": 0.748125,
      "calib/mu_w": 0.7195569620253165,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3524015748031496,
      "calib/std_conf": 0.16391216332046432,
      "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": 2838.0,
      "completions/max_terminated_length": 2838.0,
      "completions/mean_length": 648.99609375,
      "completions/mean_terminated_length": 651.5411987304688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 259.0,
      "epoch": 0.1056,
      "grad_norm": 0.009856244549155235,
      "learning_rate": 2.805555555555556e-06,
      "loss": 0.0082,
      "num_tokens": 21806179.0,
      "reward": 1.422696828842163,
      "reward_std": 0.4133927822113037,
      "rewards/accuracy_reward_step": 0.375,
      "rewards/final_brier_reward_step": 0.6222191452980042,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.849818229675293,
      "step": 99
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.619396551724138,
      "calib/avg_num_step_conf": 6.41015625,
      "calib/ece": 0.14656250000000004,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0546875,
      "calib/gap": 0.05310344827586211,
      "calib/mean_conf": 0.6934375,
      "calib/mu_c": 0.7175000000000001,
      "calib/mu_w": 0.664396551724138,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.14656250000000004,
      "calib/std_conf": 0.1285370660305812,
      "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": 1506.0,
      "completions/max_terminated_length": 1506.0,
      "completions/mean_length": 560.96484375,
      "completions/mean_terminated_length": 563.1647338867188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 269.0,
      "epoch": 0.10666666666666667,
      "grad_norm": 0.010851933620870113,
      "learning_rate": 2.7777777777777783e-06,
      "loss": 0.0178,
      "num_tokens": 22057194.0,
      "reward": 1.725217580795288,
      "reward_std": 0.3864947557449341,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.7405133247375488,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.8791074752807617,
      "step": 100
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5550820707070706,
      "calib/avg_num_step_conf": 6.8125,
      "calib/ece": 0.23219246031746035,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.12301587301587301,
      "calib/gap": 0.024342424242424276,
      "calib/mean_conf": 0.6928075396825396,
      "calib/mu_c": 0.7055583333333334,
      "calib/mu_w": 0.6812159090909091,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.22440476190476194,
      "calib/std_conf": 0.16549637237278048,
      "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": 2709.0,
      "completions/max_terminated_length": 2709.0,
      "completions/mean_length": 591.1171875,
      "completions/mean_terminated_length": 595.7716674804688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 224.0,
      "epoch": 0.10773333333333333,
      "grad_norm": 0.010762743651866913,
      "learning_rate": 2.7500000000000004e-06,
      "loss": -0.013,
      "num_tokens": 22315512.0,
      "reward": 1.5801358222961426,
      "reward_std": 0.4919942617416382,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.677642285823822,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8616510629653931,
      "step": 101
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.7328753215107621,
      "calib/avg_num_step_conf": 6.734375,
      "calib/ece": 0.06649999999999999,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.047058823529411764,
      "calib/gap": 0.10636895898199539,
      "calib/mean_conf": 0.6635980392156864,
      "calib/mu_c": 0.7007228915662651,
      "calib/mu_w": 0.5943539325842697,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.039558823529411806,
      "calib/std_conf": 0.13501931668656994,
      "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": 1378.0,
      "completions/max_terminated_length": 1378.0,
      "completions/mean_length": 488.171875,
      "completions/mean_terminated_length": 490.0863037109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 259.0,
      "epoch": 0.1088,
      "grad_norm": 0.012388897128403187,
      "learning_rate": 2.7222222222222224e-06,
      "loss": -0.0081,
      "num_tokens": 22547180.0,
      "reward": 1.8868448734283447,
      "reward_std": 0.3592468798160553,
      "rewards/accuracy_reward_step": 0.6484375,
      "rewards/final_brier_reward_step": 0.796177327632904,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8762019872665405,
      "step": 102
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5109670987038883,
      "calib/avg_num_step_conf": 6.59375,
      "calib/ece": 0.1766141732283465,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.09448818897637795,
      "calib/gap": -0.0017746759720835836,
      "calib/mean_conf": 0.6902362204724409,
      "calib/mu_c": 0.6894117647058825,
      "calib/mu_w": 0.6911864406779661,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.16570866141732285,
      "calib/std_conf": 0.14031160242364688,
      "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": 1306.0,
      "completions/max_terminated_length": 1306.0,
      "completions/mean_length": 581.57421875,
      "completions/mean_terminated_length": 583.8549194335938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 197.0,
      "epoch": 0.10986666666666667,
      "grad_norm": 0.0105710718780756,
      "learning_rate": 2.6944444444444444e-06,
      "loss": 0.0269,
      "num_tokens": 22800615.0,
      "reward": 1.6926425695419312,
      "reward_std": 0.408366858959198,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.701200008392334,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8740577697753906,
      "step": 103
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5903319707667534,
      "calib/avg_num_step_conf": 6.10546875,
      "calib/ece": 0.19290196078431376,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.023529411764705882,
      "calib/gap": 0.04008547008546992,
      "calib/mean_conf": 0.6517254901960784,
      "calib/mu_c": 0.6734188034188033,
      "calib/mu_w": 0.6333333333333334,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.19290196078431376,
      "calib/std_conf": 0.12792216972484732,
      "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": 1129.0,
      "completions/max_terminated_length": 1129.0,
      "completions/mean_length": 517.54296875,
      "completions/mean_terminated_length": 519.5725708007812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 196.0,
      "epoch": 0.11093333333333333,
      "grad_norm": 0.011246602050960064,
      "learning_rate": 2.666666666666667e-06,
      "loss": 0.0073,
      "num_tokens": 23039786.0,
      "reward": 1.584836721420288,
      "reward_std": 0.4208795428276062,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.7152222990989685,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8897498846054077,
      "step": 104
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.465679012345679,
      "calib/avg_num_step_conf": 6.61328125,
      "calib/ece": 0.17496470588235297,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.09019607843137255,
      "calib/gap": -0.034275925925925876,
      "calib/mean_conf": 0.6586039215686276,
      "calib/mu_c": 0.6424740740740741,
      "calib/mu_w": 0.67675,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.15207843137254906,
      "calib/std_conf": 0.15082412814825125,
      "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": 2774.0,
      "completions/max_terminated_length": 2774.0,
      "completions/mean_length": 569.0390625,
      "completions/mean_terminated_length": 569.0390625,
      "completions/min_length": 264.0,
      "completions/min_terminated_length": 264.0,
      "epoch": 0.112,
      "grad_norm": 0.01068261917680502,
      "learning_rate": 2.6388888888888893e-06,
      "loss": 0.0117,
      "num_tokens": 23291220.0,
      "reward": 1.679351806640625,
      "reward_std": 0.4076327979564667,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.6891355514526367,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8798341751098633,
      "step": 105
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.47643656945982527,
      "calib/avg_num_step_conf": 6.28125,
      "calib/ece": 0.1744313725490196,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.03529411764705882,
      "calib/gap": -0.006450719822812756,
      "calib/mean_conf": 0.6525490196078432,
      "calib/mu_c": 0.6492857142857142,
      "calib/mu_w": 0.655736434108527,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.16643137254901963,
      "calib/std_conf": 0.12793753281566905,
      "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": 1438.0,
      "completions/max_terminated_length": 1438.0,
      "completions/mean_length": 496.4296875,
      "completions/mean_terminated_length": 498.3764953613281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 250.0,
      "epoch": 0.11306666666666666,
      "grad_norm": 0.011055233888328075,
      "learning_rate": 2.6111111111111113e-06,
      "loss": 0.0335,
      "num_tokens": 23522890.0,
      "reward": 1.6351900100708008,
      "reward_std": 0.38616669178009033,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.7025859355926514,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8928613662719727,
      "step": 106
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5764588776906806,
      "calib/avg_num_step_conf": 6.27734375,
      "calib/ece": 0.0989411764705883,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.03529411764705882,
      "calib/gap": 0.03592571855169835,
      "calib/mean_conf": 0.6518823529411766,
      "calib/mu_c": 0.6679432624113476,
      "calib/mu_w": 0.6320175438596493,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0989411764705883,
      "calib/std_conf": 0.12448003967229689,
      "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": 1292.0,
      "completions/max_terminated_length": 1292.0,
      "completions/mean_length": 485.84375,
      "completions/mean_terminated_length": 487.7490539550781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.11413333333333334,
      "grad_norm": 0.01080196350812912,
      "learning_rate": 2.5833333333333337e-06,
      "loss": -0.02,
      "num_tokens": 23751882.0,
      "reward": 1.7295396327972412,
      "reward_std": 0.3548358678817749,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.7417683601379395,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8873279094696045,
      "step": 107
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5119723183391003,
      "calib/avg_num_step_conf": 6.49609375,
      "calib/ece": 0.09463529411764712,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.043137254901960784,
      "calib/gap": 4.705882352940005e-05,
      "calib/mean_conf": 0.6680313725490197,
      "calib/mu_c": 0.6680470588235294,
      "calib/mu_w": 0.668,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.04800000000000009,
      "calib/std_conf": 0.13613137036023448,
      "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": 1659.0,
      "completions/max_terminated_length": 1659.0,
      "completions/mean_length": 530.328125,
      "completions/mean_terminated_length": 532.4078979492188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 187.0,
      "epoch": 0.1152,
      "grad_norm": 0.01080621313303709,
      "learning_rate": 2.5555555555555557e-06,
      "loss": -0.0048,
      "num_tokens": 23990878.0,
      "reward": 1.9048094749450684,
      "reward_std": 0.45833563804626465,
      "rewards/accuracy_reward_step": 0.6640625,
      "rewards/final_brier_reward_step": 0.7562992572784424,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8863762617111206,
      "step": 108
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5950152847963067,
      "calib/avg_num_step_conf": 6.70703125,
      "calib/ece": 0.12139763779527565,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.03937007874015748,
      "calib/gap": 0.04088059142803657,
      "calib/mean_conf": 0.660767716535433,
      "calib/mu_c": 0.6795985401459853,
      "calib/mu_w": 0.6387179487179487,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.12139763779527565,
      "calib/std_conf": 0.1292966328747139,
      "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": 1938.0,
      "completions/max_terminated_length": 1938.0,
      "completions/mean_length": 520.4609375,
      "completions/mean_terminated_length": 524.55908203125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 209.0,
      "epoch": 0.11626666666666667,
      "grad_norm": 0.010906565003097057,
      "learning_rate": 2.5277777777777778e-06,
      "loss": 0.0007,
      "num_tokens": 24228716.0,
      "reward": 1.7045462131500244,
      "reward_std": 0.27251356840133667,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.7346241474151611,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8882478475570679,
      "step": 109
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.48100837232208815,
      "calib/avg_num_step_conf": 6.04296875,
      "calib/ece": 0.1586596078431373,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.00784313725490196,
      "calib/gap": -0.0097410859394238,
      "calib/mean_conf": 0.6281639215686274,
      "calib/mu_c": 0.6231596774193549,
      "calib/mu_w": 0.6329007633587787,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1502745098039216,
      "calib/std_conf": 0.11000198907432343,
      "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": 1159.0,
      "completions/max_terminated_length": 1159.0,
      "completions/mean_length": 506.15234375,
      "completions/mean_terminated_length": 508.1372985839844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 183.0,
      "epoch": 0.11733333333333333,
      "grad_norm": 0.011566469445824623,
      "learning_rate": 2.5e-06,
      "loss": 0.0107,
      "num_tokens": 24463211.0,
      "reward": 1.6272175312042236,
      "reward_std": 0.4574228525161743,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.7103029489517212,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9001297950744629,
      "step": 110
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5219117279319829,
      "calib/avg_num_step_conf": 6.13671875,
      "calib/ece": 0.1483003952569171,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.015810276679841896,
      "calib/gap": 0.014963115778944824,
      "calib/mean_conf": 0.6384189723320157,
      "calib/mu_c": 0.6460483870967743,
      "calib/mu_w": 0.6310852713178294,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1483003952569171,
      "calib/std_conf": 0.11171971543964937,
      "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": 2605.0,
      "completions/max_terminated_length": 2605.0,
      "completions/mean_length": 543.1953125,
      "completions/mean_terminated_length": 543.1953125,
      "completions/min_length": 183.0,
      "completions/min_terminated_length": 183.0,
      "epoch": 0.1184,
      "grad_norm": 0.010652108117938042,
      "learning_rate": 2.4722222222222226e-06,
      "loss": 0.0414,
      "num_tokens": 24709677.0,
      "reward": 1.6171307563781738,
      "reward_std": 0.41484975814819336,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.7135441303253174,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8799788951873779,
      "step": 111
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.651970756516211,
      "calib/avg_num_step_conf": 6.03125,
      "calib/ece": 0.11434262948207179,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0199203187250996,
      "calib/gap": 0.05719707565162102,
      "calib/mean_conf": 0.6278884462151394,
      "calib/mu_c": 0.6554615384615384,
      "calib/mu_w": 0.5982644628099174,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.11215139442231084,
      "calib/std_conf": 0.12040735766082009,
      "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": 2921.0,
      "completions/max_terminated_length": 2921.0,
      "completions/mean_length": 551.93359375,
      "completions/mean_terminated_length": 556.279541015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 227.0,
      "epoch": 0.11946666666666667,
      "grad_norm": 0.012287004850804806,
      "learning_rate": 2.4444444444444447e-06,
      "loss": 0.0056,
      "num_tokens": 24958892.0,
      "reward": 1.6569784879684448,
      "reward_std": 0.40331631898880005,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.7376008033752441,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8825008869171143,
      "step": 112
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5629016312407317,
      "calib/avg_num_step_conf": 5.44921875,
      "calib/ece": 0.07247058823529404,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.011764705882352941,
      "calib/gap": 0.023644957983193216,
      "calib/mean_conf": 0.6058039215686274,
      "calib/mu_c": 0.6168382352941176,
      "calib/mu_w": 0.5931932773109244,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.07247058823529404,
      "calib/std_conf": 0.105149943285085,
      "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": 2071.0,
      "completions/max_terminated_length": 2071.0,
      "completions/mean_length": 465.55859375,
      "completions/mean_terminated_length": 465.55859375,
      "completions/min_length": 244.0,
      "completions/min_terminated_length": 244.0,
      "epoch": 0.12053333333333334,
      "grad_norm": 0.012438077479600906,
      "learning_rate": 2.4166666666666667e-06,
      "loss": 0.0378,
      "num_tokens": 25183275.0,
      "reward": 1.7068159580230713,
      "reward_std": 0.45404136180877686,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.7436562776565552,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9039199948310852,
      "step": 113
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6661144977471509,
      "calib/avg_num_step_conf": 6.02734375,
      "calib/ece": 0.062289682539682506,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.01984126984126984,
      "calib/gap": 0.060761595547309866,
      "calib/mean_conf": 0.6243769841269841,
      "calib/mu_c": 0.6480064935064935,
      "calib/mu_w": 0.5872448979591837,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.03777777777777778,
      "calib/std_conf": 0.11961116212624232,
      "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": 1409.0,
      "completions/max_terminated_length": 1409.0,
      "completions/mean_length": 473.33203125,
      "completions/mean_terminated_length": 477.0590515136719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 206.0,
      "epoch": 0.1216,
      "grad_norm": 0.011878692544996738,
      "learning_rate": 2.388888888888889e-06,
      "loss": -0.0168,
      "num_tokens": 25409472.0,
      "reward": 1.810877799987793,
      "reward_std": 0.4106457233428955,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.7618821859359741,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8878784775733948,
      "step": 114
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6275689223057644,
      "calib/avg_num_step_conf": 5.671875,
      "calib/ece": 0.09525691699604741,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.007905138339920948,
      "calib/gap": 0.04418358395989985,
      "calib/mean_conf": 0.6158102766798419,
      "calib/mu_c": 0.6367669172932331,
      "calib/mu_w": 0.5925833333333332,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0926877470355731,
      "calib/std_conf": 0.09809970916301491,
      "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": 2786.0,
      "completions/max_terminated_length": 2786.0,
      "completions/mean_length": 480.609375,
      "completions/mean_terminated_length": 482.494140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 221.0,
      "epoch": 0.12266666666666666,
      "grad_norm": 0.011466645635664463,
      "learning_rate": 2.361111111111111e-06,
      "loss": 0.0137,
      "num_tokens": 25637772.0,
      "reward": 1.6898902654647827,
      "reward_std": 0.4999661445617676,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.7461015582084656,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8962720632553101,
      "step": 115
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5467625899280576,
      "calib/avg_num_step_conf": 6.3046875,
      "calib/ece": 0.068046875,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.01171875,
      "calib/gap": 0.015604746971653372,
      "calib/mean_conf": 0.603515625,
      "calib/mu_c": 0.6106474820143885,
      "calib/mu_w": 0.5950427350427351,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.06429687499999999,
      "calib/std_conf": 0.09694626285143422,
      "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": 1236.0,
      "completions/max_terminated_length": 1236.0,
      "completions/mean_length": 545.19140625,
      "completions/mean_terminated_length": 547.3294677734375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 160.0,
      "epoch": 0.12373333333333333,
      "grad_norm": 0.010517150163650513,
      "learning_rate": 2.3333333333333336e-06,
      "loss": 0.0064,
      "num_tokens": 25881861.0,
      "reward": 1.7288788557052612,
      "reward_std": 0.48391813039779663,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.7465265393257141,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.9111762642860413,
      "step": 116
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5255786752031519,
      "calib/avg_num_step_conf": 6.04296875,
      "calib/ece": 0.09160784313725492,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.00392156862745098,
      "calib/gap": -0.002237749322826943,
      "calib/mean_conf": 0.592156862745098,
      "calib/mu_c": 0.5910687022900764,
      "calib/mu_w": 0.5933064516129033,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.08501960784313728,
      "calib/std_conf": 0.0972305472351556,
      "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": 2147.0,
      "completions/max_terminated_length": 2147.0,
      "completions/mean_length": 508.953125,
      "completions/mean_terminated_length": 508.953125,
      "completions/min_length": 208.0,
      "completions/min_terminated_length": 208.0,
      "epoch": 0.1248,
      "grad_norm": 0.010903061367571354,
      "learning_rate": 2.305555555555556e-06,
      "loss": 0.0075,
      "num_tokens": 26118753.0,
      "reward": 1.6741626262664795,
      "reward_std": 0.32085680961608887,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.7305999994277954,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9035505652427673,
      "step": 117
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4932255244755245,
      "calib/avg_num_step_conf": 5.91015625,
      "calib/ece": 0.1524313725490196,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.00784313725490196,
      "calib/gap": 0.002771603396603428,
      "calib/mean_conf": 0.5936078431372549,
      "calib/mu_c": 0.5948251748251748,
      "calib/mu_w": 0.5920535714285714,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.09262745098039213,
      "calib/std_conf": 0.10274431970966981,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1289.0,
      "completions/max_terminated_length": 1289.0,
      "completions/mean_length": 492.77734375,
      "completions/mean_terminated_length": 494.7098388671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 229.0,
      "epoch": 0.12586666666666665,
      "grad_norm": 0.011059942655265331,
      "learning_rate": 2.277777777777778e-06,
      "loss": -0.0104,
      "num_tokens": 26348912.0,
      "reward": 1.7454500198364258,
      "reward_std": 0.23404794931411743,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.7382664084434509,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9075959920883179,
      "step": 118
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6213670267489713,
      "calib/avg_num_step_conf": 5.828125,
      "calib/ece": 0.0653571428571429,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.03456018518518511,
      "calib/mean_conf": 0.5662301587301587,
      "calib/mu_c": 0.5810416666666667,
      "calib/mu_w": 0.5464814814814816,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.030079365079365098,
      "calib/std_conf": 0.08284903680776745,
      "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": 2963.0,
      "completions/max_terminated_length": 2963.0,
      "completions/mean_length": 529.625,
      "completions/mean_terminated_length": 535.9051513671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 237.0,
      "epoch": 0.12693333333333334,
      "grad_norm": 0.01092610228806734,
      "learning_rate": 2.25e-06,
      "loss": -0.0133,
      "num_tokens": 26589560.0,
      "reward": 1.7466264963150024,
      "reward_std": 0.3065892457962036,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.7506831884384155,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8998852968215942,
      "step": 119
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6439732866498088,
      "calib/avg_num_step_conf": 6.0234375,
      "calib/ece": 0.059648437499999894,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.03892303702262845,
      "calib/mean_conf": 0.5687109374999999,
      "calib/mu_c": 0.5834591194968552,
      "calib/mu_w": 0.5445360824742268,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0036328125000000006,
      "calib/std_conf": 0.08146784760180603,
      "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": 1108.0,
      "completions/max_terminated_length": 1108.0,
      "completions/mean_length": 487.7265625,
      "completions/mean_terminated_length": 489.6392517089844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 215.0,
      "epoch": 0.128,
      "grad_norm": 0.012312943115830421,
      "learning_rate": 2.222222222222222e-06,
      "loss": 0.0096,
      "num_tokens": 26821106.0,
      "reward": 1.8510717153549194,
      "reward_std": 0.36764204502105713,
      "rewards/accuracy_reward_step": 0.62109375,
      "rewards/final_brier_reward_step": 0.7707937955856323,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9147434830665588,
      "step": 120
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5252698412698412,
      "calib/avg_num_step_conf": 6.2734375,
      "calib/ece": 0.09816733067729083,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.00796812749003984,
      "calib/gap": -0.0029422222222222905,
      "calib/mean_conf": 0.5907569721115538,
      "calib/mu_c": 0.58928,
      "calib/mu_w": 0.5922222222222223,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0954581673306773,
      "calib/std_conf": 0.10277341679828322,
      "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": 2747.0,
      "completions/max_terminated_length": 2747.0,
      "completions/mean_length": 559.203125,
      "completions/mean_terminated_length": 568.0794067382812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 229.0,
      "epoch": 0.12906666666666666,
      "grad_norm": 0.009793057106435299,
      "learning_rate": 2.1944444444444445e-06,
      "loss": -0.0068,
      "num_tokens": 27069318.0,
      "reward": 1.6255204677581787,
      "reward_std": 0.49173569679260254,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.715122640132904,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8963341116905212,
      "step": 121
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6879084967320261,
      "calib/avg_num_step_conf": 6.08203125,
      "calib/ece": 0.08913725490196082,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.05862745098039224,
      "calib/mean_conf": 0.5873333333333333,
      "calib/mu_c": 0.6107843137254902,
      "calib/mu_w": 0.552156862745098,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.038235294117647076,
      "calib/std_conf": 0.09221429999833543,
      "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": 2373.0,
      "completions/max_terminated_length": 2373.0,
      "completions/mean_length": 495.89453125,
      "completions/mean_terminated_length": 495.89453125,
      "completions/min_length": 239.0,
      "completions/min_terminated_length": 239.0,
      "epoch": 0.13013333333333332,
      "grad_norm": 0.012019730173051357,
      "learning_rate": 2.166666666666667e-06,
      "loss": 0.0493,
      "num_tokens": 27303611.0,
      "reward": 1.8140134811401367,
      "reward_std": 0.3575590252876282,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.7735027074813843,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9122384190559387,
      "step": 122
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5302514653053507,
      "calib/avg_num_step_conf": 6.2421875,
      "calib/ece": 0.09357142857142857,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.01984126984126984,
      "calib/gap": -0.0023255813953487747,
      "calib/mean_conf": 0.6054761904761905,
      "calib/mu_c": 0.6043410852713179,
      "calib/mu_w": 0.6066666666666667,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.09357142857142857,
      "calib/std_conf": 0.10294941848519548,
      "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": 2533.0,
      "completions/max_terminated_length": 2533.0,
      "completions/mean_length": 557.578125,
      "completions/mean_terminated_length": 564.1897583007812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 259.0,
      "epoch": 0.1312,
      "grad_norm": 0.01179084088653326,
      "learning_rate": 2.138888888888889e-06,
      "loss": 0.0182,
      "num_tokens": 27551639.0,
      "reward": 1.650080680847168,
      "reward_std": 0.5606234669685364,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.7182250022888184,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.889910101890564,
      "step": 123
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5170613700025465,
      "calib/avg_num_step_conf": 5.90234375,
      "calib/ece": 0.06238281249999999,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.00390625,
      "calib/gap": 0.0016768525592053551,
      "calib/mean_conf": 0.6077734375,
      "calib/mu_c": 0.6084415584415583,
      "calib/mu_w": 0.606764705882353,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.03429687499999999,
      "calib/std_conf": 0.10597816069565273,
      "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": 1461.0,
      "completions/max_terminated_length": 1461.0,
      "completions/mean_length": 508.66015625,
      "completions/mean_terminated_length": 510.6549377441406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 215.0,
      "epoch": 0.13226666666666667,
      "grad_norm": 0.011192714795470238,
      "learning_rate": 2.1111111111111114e-06,
      "loss": 0.0022,
      "num_tokens": 27788672.0,
      "reward": 1.8174453973770142,
      "reward_std": 0.3154680132865906,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.7498488426208496,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.9105578660964966,
      "step": 124
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4941504004004004,
      "calib/avg_num_step_conf": 6.1796875,
      "calib/ece": 0.10752941176470596,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.00784313725490196,
      "calib/gap": -0.0013475975975975052,
      "calib/mean_conf": 0.6323921568627451,
      "calib/mu_c": 0.6318055555555556,
      "calib/mu_w": 0.6331531531531531,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.087607843137255,
      "calib/std_conf": 0.104225261028814,
      "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": 2864.0,
      "completions/max_terminated_length": 2864.0,
      "completions/mean_length": 517.63671875,
      "completions/mean_terminated_length": 517.63671875,
      "completions/min_length": 246.0,
      "completions/min_terminated_length": 246.0,
      "epoch": 0.13333333333333333,
      "grad_norm": 0.010911756195127964,
      "learning_rate": 2.0833333333333334e-06,
      "loss": 0.0293,
      "num_tokens": 28025995.0,
      "reward": 1.7478008270263672,
      "reward_std": 0.4162144064903259,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.7334879040718079,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8983402252197266,
      "step": 125
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6463255813953488,
      "calib/avg_num_step_conf": 5.625,
      "calib/ece": 0.13480314960629927,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.015748031496062992,
      "calib/gap": 0.04857674418604663,
      "calib/mean_conf": 0.6269291338582678,
      "calib/mu_c": 0.6516000000000001,
      "calib/mu_w": 0.6030232558139534,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.13480314960629927,
      "calib/std_conf": 0.10181434246863584,
      "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": 2426.0,
      "completions/max_terminated_length": 2426.0,
      "completions/mean_length": 532.0390625,
      "completions/mean_terminated_length": 534.1255493164062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 199.0,
      "epoch": 0.1344,
      "grad_norm": 0.010709105059504509,
      "learning_rate": 2.0555555555555555e-06,
      "loss": -0.0286,
      "num_tokens": 28267661.0,
      "reward": 1.6336421966552734,
      "reward_std": 0.3790965676307678,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.7370499968528748,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8912684917449951,
      "step": 126
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.6661931818181818,
      "calib/avg_num_step_conf": 5.73046875,
      "calib/ece": 0.15080321285140563,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.024096385542168676,
      "calib/gap": 0.062597494834711,
      "calib/mean_conf": 0.6367469879518073,
      "calib/mu_c": 0.6689256198347109,
      "calib/mu_w": 0.6063281249999999,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.15080321285140563,
      "calib/std_conf": 0.10699369534856351,
      "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": 2885.0,
      "completions/max_terminated_length": 2885.0,
      "completions/mean_length": 502.5234375,
      "completions/mean_terminated_length": 510.5000305175781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 217.0,
      "epoch": 0.13546666666666668,
      "grad_norm": 0.0120786027982831,
      "learning_rate": 2.027777777777778e-06,
      "loss": 0.012,
      "num_tokens": 28499979.0,
      "reward": 1.597495436668396,
      "reward_std": 0.394074410200119,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.726848840713501,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.881882905960083,
      "step": 127
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.610711209548419,
      "calib/avg_num_step_conf": 5.61328125,
      "calib/ece": 0.14647058823529419,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0196078431372549,
      "calib/gap": 0.03744001476559622,
      "calib/mean_conf": 0.6523529411764707,
      "calib/mu_c": 0.6708527131782945,
      "calib/mu_w": 0.6334126984126983,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.14647058823529419,
      "calib/std_conf": 0.09856197881444989,
      "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": 2792.0,
      "completions/max_terminated_length": 2792.0,
      "completions/mean_length": 521.640625,
      "completions/mean_terminated_length": 521.640625,
      "completions/min_length": 230.0,
      "completions/min_terminated_length": 230.0,
      "epoch": 0.13653333333333334,
      "grad_norm": 0.012539071962237358,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0146,
      "num_tokens": 28740183.0,
      "reward": 1.661848545074463,
      "reward_std": 0.31563234329223633,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.7347027063369751,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8970666527748108,
      "step": 128
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5719571825029734,
      "calib/avg_num_step_conf": 5.90234375,
      "calib/ece": 0.05937254901960794,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0392156862745098,
      "calib/gap": 0.02695189639222928,
      "calib/mean_conf": 0.6818039215686275,
      "calib/mu_c": 0.6917391304347825,
      "calib/mu_w": 0.6647872340425532,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.054901960784313794,
      "calib/std_conf": 0.11230159316660897,
      "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": 2448.0,
      "completions/max_terminated_length": 2448.0,
      "completions/mean_length": 465.0625,
      "completions/mean_terminated_length": 465.0625,
      "completions/min_length": 218.0,
      "completions/min_terminated_length": 218.0,
      "epoch": 0.1376,
      "grad_norm": 0.012660709209740162,
      "learning_rate": 1.9722222222222224e-06,
      "loss": 0.0116,
      "num_tokens": 28961623.0,
      "reward": 1.8541687726974487,
      "reward_std": 0.4114261865615845,
      "rewards/accuracy_reward_step": 0.62890625,
      "rewards/final_brier_reward_step": 0.761662483215332,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8893877267837524,
      "step": 129
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6036775724275724,
      "calib/avg_num_step_conf": 5.6171875,
      "calib/ece": 0.12439215686274516,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.023529411764705882,
      "calib/gap": 0.046009615384615454,
      "calib/mean_conf": 0.6851764705882354,
      "calib/mu_c": 0.7053846153846154,
      "calib/mu_w": 0.6593749999999999,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.12439215686274516,
      "calib/std_conf": 0.115445718656738,
      "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": 2436.0,
      "completions/max_terminated_length": 2436.0,
      "completions/mean_length": 465.10546875,
      "completions/mean_terminated_length": 465.10546875,
      "completions/min_length": 183.0,
      "completions/min_terminated_length": 183.0,
      "epoch": 0.13866666666666666,
      "grad_norm": 0.011984642595052719,
      "learning_rate": 1.944444444444445e-06,
      "loss": 0.0308,
      "num_tokens": 29185978.0,
      "reward": 1.7440496683120728,
      "reward_std": 0.3274148404598236,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.7446382641792297,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8878102898597717,
      "step": 130
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5754952118588482,
      "calib/avg_num_step_conf": 5.4140625,
      "calib/ece": 0.30043478260869566,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.03162055335968379,
      "calib/gap": 0.02551948051948072,
      "calib/mean_conf": 0.6917391304347826,
      "calib/mu_c": 0.7072727272727274,
      "calib/mu_w": 0.6817532467532467,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.30043478260869566,
      "calib/std_conf": 0.11124936883257819,
      "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": 2219.0,
      "completions/max_terminated_length": 2219.0,
      "completions/mean_length": 462.59765625,
      "completions/mean_terminated_length": 466.2401428222656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 210.0,
      "epoch": 0.13973333333333332,
      "grad_norm": 0.012838898226618767,
      "learning_rate": 1.916666666666667e-06,
      "loss": -0.0023,
      "num_tokens": 29410611.0,
      "reward": 1.4626030921936035,
      "reward_std": 0.3501003384590149,
      "rewards/accuracy_reward_step": 0.390625,
      "rewards/final_brier_reward_step": 0.6607421636581421,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8771703243255615,
      "step": 131
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5402930402930404,
      "calib/avg_num_step_conf": 6.671875,
      "calib/ece": 0.14232283464566922,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.06692913385826772,
      "calib/gap": 0.011493982208267695,
      "calib/mean_conf": 0.7383858267716535,
      "calib/mu_c": 0.7428205128205128,
      "calib/mu_w": 0.7313265306122451,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.133267716535433,
      "calib/std_conf": 0.11589158300635134,
      "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": 2958.0,
      "completions/max_terminated_length": 2958.0,
      "completions/mean_length": 514.578125,
      "completions/mean_terminated_length": 514.578125,
      "completions/min_length": 197.0,
      "completions/min_terminated_length": 197.0,
      "epoch": 0.1408,
      "grad_norm": 0.012298737652599812,
      "learning_rate": 1.888888888888889e-06,
      "loss": 0.0328,
      "num_tokens": 29647935.0,
      "reward": 1.8090349435806274,
      "reward_std": 0.4609593152999878,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.7315890789031982,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8717383146286011,
      "step": 132
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6335258152173914,
      "calib/avg_num_step_conf": 6.1171875,
      "calib/ece": 0.3766269841269841,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.09126984126984126,
      "calib/gap": 0.05637499999999995,
      "calib/mean_conf": 0.7417063492063491,
      "calib/mu_c": 0.7775,
      "calib/mu_w": 0.721125,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3766269841269841,
      "calib/std_conf": 0.11357058971660076,
      "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": 1853.0,
      "completions/max_terminated_length": 1853.0,
      "completions/mean_length": 543.19140625,
      "completions/mean_terminated_length": 547.468505859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 258.0,
      "epoch": 0.14186666666666667,
      "grad_norm": 0.011609735898673534,
      "learning_rate": 1.8611111111111113e-06,
      "loss": 0.0158,
      "num_tokens": 29893336.0,
      "reward": 1.4053130149841309,
      "reward_std": 0.4048590064048767,
      "rewards/accuracy_reward_step": 0.359375,
      "rewards/final_brier_reward_step": 0.62959885597229,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8666534423828125,
      "step": 133
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5933084344849051,
      "calib/avg_num_step_conf": 5.859375,
      "calib/ece": 0.1909350393700788,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.08267716535433071,
      "calib/gap": 0.04812807345160297,
      "calib/mean_conf": 0.7146555118110235,
      "calib/mu_c": 0.7372037037037038,
      "calib/mu_w": 0.6890756302521008,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.18704724409448822,
      "calib/std_conf": 0.1408914406916585,
      "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": 2487.0,
      "completions/max_terminated_length": 2487.0,
      "completions/mean_length": 555.2734375,
      "completions/mean_terminated_length": 555.2734375,
      "completions/min_length": 207.0,
      "completions/min_terminated_length": 207.0,
      "epoch": 0.14293333333333333,
      "grad_norm": 0.012096771970391273,
      "learning_rate": 1.8333333333333333e-06,
      "loss": 0.0467,
      "num_tokens": 30144438.0,
      "reward": 1.6862773895263672,
      "reward_std": 0.514042854309082,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.7159255743026733,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8807463645935059,
      "step": 134
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5513161146811667,
      "calib/avg_num_step_conf": 6.140625,
      "calib/ece": 0.18443137254901962,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.06666666666666667,
      "calib/gap": 0.01707983193277285,
      "calib/mean_conf": 0.717764705882353,
      "calib/mu_c": 0.7257352941176469,
      "calib/mu_w": 0.7086554621848741,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.18443137254901962,
      "calib/std_conf": 0.11799012724323284,
      "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": 1626.0,
      "completions/max_terminated_length": 1626.0,
      "completions/mean_length": 519.21484375,
      "completions/mean_terminated_length": 521.2510375976562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 250.0,
      "epoch": 0.144,
      "grad_norm": 0.011714962311089039,
      "learning_rate": 1.8055555555555557e-06,
      "loss": -0.0191,
      "num_tokens": 30383237.0,
      "reward": 1.6944434642791748,
      "reward_std": 0.5428236722946167,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.7088964581489563,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8891900181770325,
      "step": 135
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5992043912668065,
      "calib/avg_num_step_conf": 6.1171875,
      "calib/ece": 0.219843137254902,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.043137254901960784,
      "calib/gap": 0.04109966695448375,
      "calib/mean_conf": 0.6943529411764706,
      "calib/mu_c": 0.7159504132231405,
      "calib/mu_w": 0.6748507462686567,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.219843137254902,
      "calib/std_conf": 0.11744115314545321,
      "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": 1315.0,
      "completions/max_terminated_length": 1315.0,
      "completions/mean_length": 476.47265625,
      "completions/mean_terminated_length": 478.3412170410156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 179.0,
      "epoch": 0.14506666666666668,
      "grad_norm": 0.011427158489823341,
      "learning_rate": 1.777777777777778e-06,
      "loss": 0.0088,
      "num_tokens": 30613702.0,
      "reward": 1.602003812789917,
      "reward_std": 0.36013883352279663,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.7039972543716431,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8837054371833801,
      "step": 136
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.562200956937799,
      "calib/avg_num_step_conf": 6.2578125,
      "calib/ece": 0.22330708661417323,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.051181102362204724,
      "calib/gap": 0.018671472068601314,
      "calib/mean_conf": 0.7428346456692912,
      "calib/mu_c": 0.7517293233082707,
      "calib/mu_w": 0.7330578512396694,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.22125984251968503,
      "calib/std_conf": 0.10044701378938628,
      "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": 1439.0,
      "completions/max_terminated_length": 1439.0,
      "completions/mean_length": 501.15234375,
      "completions/mean_terminated_length": 503.11767578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 258.0,
      "epoch": 0.14613333333333334,
      "grad_norm": 0.011493817903101444,
      "learning_rate": 1.75e-06,
      "loss": 0.0039,
      "num_tokens": 30848981.0,
      "reward": 1.6694798469543457,
      "reward_std": 0.41539496183395386,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.696246862411499,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8801102042198181,
      "step": 137
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6077588910493634,
      "calib/avg_num_step_conf": 6.10546875,
      "calib/ece": 0.13539062500000001,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.03515625,
      "calib/gap": 0.04363545129523916,
      "calib/mean_conf": 0.7125,
      "calib/mu_c": 0.730738255033557,
      "calib/mu_w": 0.6871028037383179,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1329296875,
      "calib/std_conf": 0.1132095678377053,
      "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": 1178.0,
      "completions/max_terminated_length": 1178.0,
      "completions/mean_length": 502.73828125,
      "completions/mean_terminated_length": 504.7098388671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 196.0,
      "epoch": 0.1472,
      "grad_norm": 0.012307239696383476,
      "learning_rate": 1.7222222222222224e-06,
      "loss": -0.0015,
      "num_tokens": 31082018.0,
      "reward": 1.7836527824401855,
      "reward_std": 0.44293880462646484,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7481210827827454,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.8943023681640625,
      "step": 138
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6541900868676545,
      "calib/avg_num_step_conf": 6.2421875,
      "calib/ece": 0.10200000000000006,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.03137254901960784,
      "calib/gap": 0.06101686254471117,
      "calib/mean_conf": 0.6921960784313725,
      "calib/mu_c": 0.7168421052631578,
      "calib/mu_w": 0.6558252427184467,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.09905882352941184,
      "calib/std_conf": 0.1281293800181558,
      "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": 2699.0,
      "completions/max_terminated_length": 2699.0,
      "completions/mean_length": 503.67578125,
      "completions/mean_terminated_length": 503.67578125,
      "completions/min_length": 190.0,
      "completions/min_terminated_length": 190.0,
      "epoch": 0.14826666666666666,
      "grad_norm": 0.011360560543835163,
      "learning_rate": 1.6944444444444446e-06,
      "loss": 0.0123,
      "num_tokens": 31314055.0,
      "reward": 1.8030741214752197,
      "reward_std": 0.4602932929992676,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.7599769830703735,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8976321816444397,
      "step": 139
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6284209085933223,
      "calib/avg_num_step_conf": 6.7421875,
      "calib/ece": 0.06776470588235292,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.07450980392156863,
      "calib/gap": 0.054493021346469606,
      "calib/mean_conf": 0.7167058823529413,
      "calib/mu_c": 0.735297619047619,
      "calib/mu_w": 0.6808045977011494,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.06282352941176467,
      "calib/std_conf": 0.12182972593854369,
      "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": 976.0,
      "completions/max_terminated_length": 976.0,
      "completions/mean_length": 501.578125,
      "completions/mean_terminated_length": 503.5451354980469,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 197.0,
      "epoch": 0.14933333333333335,
      "grad_norm": 0.012145115062594414,
      "learning_rate": 1.6666666666666667e-06,
      "loss": -0.022,
      "num_tokens": 31547475.0,
      "reward": 1.9001531600952148,
      "reward_std": 0.39148375391960144,
      "rewards/accuracy_reward_step": 0.65625,
      "rewards/final_brier_reward_step": 0.7784765362739563,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8924486637115479,
      "step": 140
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6302139891408496,
      "calib/avg_num_step_conf": 6.3515625,
      "calib/ece": 0.11343750000000007,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.03515625,
      "calib/gap": 0.04672628553177893,
      "calib/mean_conf": 0.678984375,
      "calib/mu_c": 0.6974193548387096,
      "calib/mu_w": 0.6506930693069307,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.09347656250000004,
      "calib/std_conf": 0.11790382629863787,
      "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": 1514.0,
      "completions/max_terminated_length": 1514.0,
      "completions/mean_length": 512.33203125,
      "completions/mean_terminated_length": 514.3411865234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 256.0,
      "epoch": 0.1504,
      "grad_norm": 0.011024919338524342,
      "learning_rate": 1.638888888888889e-06,
      "loss": -0.0074,
      "num_tokens": 31785728.0,
      "reward": 1.826422095298767,
      "reward_std": 0.45660847425460815,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.7641414403915405,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.9087342023849487,
      "step": 141
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6165994623655914,
      "calib/avg_num_step_conf": 6.546875,
      "calib/ece": 0.17617187500000006,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.04296875,
      "calib/gap": 0.0524389051808406,
      "calib/mean_conf": 0.691796875,
      "calib/mu_c": 0.7171969696969696,
      "calib/mu_w": 0.664758064516129,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.17617187500000006,
      "calib/std_conf": 0.12313885451080975,
      "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": 1168.0,
      "completions/max_terminated_length": 1168.0,
      "completions/mean_length": 525.8671875,
      "completions/mean_terminated_length": 527.929443359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 253.0,
      "epoch": 0.15146666666666667,
      "grad_norm": 0.010916810482740402,
      "learning_rate": 1.6111111111111113e-06,
      "loss": 0.0101,
      "num_tokens": 32025510.0,
      "reward": 1.6825264692306519,
      "reward_std": 0.3293830454349518,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.7302383184432983,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.9061173796653748,
      "step": 142
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.552674280721463,
      "calib/avg_num_step_conf": 6.80078125,
      "calib/ece": 0.08445312500000007,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.01171875,
      "calib/gap": 0.02032827810023108,
      "calib/mean_conf": 0.65671875,
      "calib/mu_c": 0.665374149659864,
      "calib/mu_w": 0.6450458715596329,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.08347656250000007,
      "calib/std_conf": 0.11319997084115128,
      "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": 1372.0,
      "completions/max_terminated_length": 1372.0,
      "completions/mean_length": 544.05078125,
      "completions/mean_terminated_length": 546.184326171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 258.0,
      "epoch": 0.15253333333333333,
      "grad_norm": 0.011365286074578762,
      "learning_rate": 1.5833333333333333e-06,
      "loss": 0.0016,
      "num_tokens": 32272123.0,
      "reward": 1.7716703414916992,
      "reward_std": 0.31434881687164307,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.7426109313964844,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.906570315361023,
      "step": 143
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5436744988716316,
      "calib/avg_num_step_conf": 6.4921875,
      "calib/ece": 0.09929411764705885,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.03137254901960784,
      "calib/gap": 0.01858622062923143,
      "calib/mean_conf": 0.6730980392156863,
      "calib/mu_c": 0.6798765432098766,
      "calib/mu_w": 0.6612903225806451,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0685490196078432,
      "calib/std_conf": 0.12247212646500918,
      "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": 1312.0,
      "completions/max_terminated_length": 1312.0,
      "completions/mean_length": 503.2265625,
      "completions/mean_terminated_length": 505.2000427246094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 236.0,
      "epoch": 0.1536,
      "grad_norm": 0.014862559735774994,
      "learning_rate": 1.5555555555555558e-06,
      "loss": 0.0214,
      "num_tokens": 32505077.0,
      "reward": 1.8636364936828613,
      "reward_std": 0.3292680084705353,
      "rewards/accuracy_reward_step": 0.6328125,
      "rewards/final_brier_reward_step": 0.75751793384552,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9079656004905701,
      "step": 144
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5333991339786042,
      "calib/avg_num_step_conf": 7.046875,
      "calib/ece": 0.12090196078431377,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0196078431372549,
      "calib/gap": 0.01643912379011714,
      "calib/mean_conf": 0.6881960784313725,
      "calib/mu_c": 0.6949006622516556,
      "calib/mu_w": 0.6784615384615384,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.10847058823529417,
      "calib/std_conf": 0.12198955590265578,
      "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": 1079.0,
      "completions/max_terminated_length": 1079.0,
      "completions/mean_length": 483.30078125,
      "completions/mean_terminated_length": 485.19610595703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 187.0,
      "epoch": 0.15466666666666667,
      "grad_norm": 0.01164257898926735,
      "learning_rate": 1.527777777777778e-06,
      "loss": -0.003,
      "num_tokens": 32731506.0,
      "reward": 1.7948551177978516,
      "reward_std": 0.42932164669036865,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.7394285202026367,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9087417125701904,
      "step": 145
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6459713755632122,
      "calib/avg_num_step_conf": 6.359375,
      "calib/ece": 0.26928571428571424,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.003968253968253968,
      "calib/gap": 0.05508348794063078,
      "calib/mean_conf": 0.6581746031746032,
      "calib/mu_c": 0.6918367346938775,
      "calib/mu_w": 0.6367532467532467,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.26928571428571424,
      "calib/std_conf": 0.11439211867658855,
      "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": 2741.0,
      "completions/max_terminated_length": 2741.0,
      "completions/mean_length": 526.6796875,
      "completions/mean_terminated_length": 528.7451171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 243.0,
      "epoch": 0.15573333333333333,
      "grad_norm": 0.011714852415025234,
      "learning_rate": 1.5e-06,
      "loss": 0.01,
      "num_tokens": 32973552.0,
      "reward": 1.4650468826293945,
      "reward_std": 0.42327678203582764,
      "rewards/accuracy_reward_step": 0.3828125,
      "rewards/final_brier_reward_step": 0.6919437646865845,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9026187658309937,
      "step": 146
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6474015453639083,
      "calib/avg_num_step_conf": 6.65234375,
      "calib/ece": 0.2271259842519685,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.027559055118110236,
      "calib/gap": 0.056592721834496706,
      "calib/mean_conf": 0.676732283464567,
      "calib/mu_c": 0.7070338983050849,
      "calib/mu_w": 0.6504411764705882,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2196456692913386,
      "calib/std_conf": 0.13867465742276497,
      "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": 1332.0,
      "completions/max_terminated_length": 1332.0,
      "completions/mean_length": 510.58203125,
      "completions/mean_terminated_length": 512.5843505859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 234.0,
      "epoch": 0.1568,
      "grad_norm": 0.011596626602113247,
      "learning_rate": 1.4722222222222225e-06,
      "loss": 0.0021,
      "num_tokens": 33207941.0,
      "reward": 1.5923233032226562,
      "reward_std": 0.3285621702671051,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.7095776796340942,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9097157716751099,
      "step": 147
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6376507936507938,
      "calib/avg_num_step_conf": 6.57421875,
      "calib/ece": 0.0972156862745098,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.01568627450980392,
      "calib/gap": 0.06232380952380956,
      "calib/mean_conf": 0.678470588235294,
      "calib/mu_c": 0.7041333333333334,
      "calib/mu_w": 0.6418095238095238,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.09372549019607843,
      "calib/std_conf": 0.12719451723002634,
      "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": 2408.0,
      "completions/max_terminated_length": 2408.0,
      "completions/mean_length": 486.48828125,
      "completions/mean_terminated_length": 486.48828125,
      "completions/min_length": 176.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.15786666666666666,
      "grad_norm": 0.012608281336724758,
      "learning_rate": 1.4444444444444445e-06,
      "loss": 0.0174,
      "num_tokens": 33437594.0,
      "reward": 1.7900137901306152,
      "reward_std": 0.3942890465259552,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.7568058371543884,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.903249204158783,
      "step": 148
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6321003134796238,
      "calib/avg_num_step_conf": 7.16796875,
      "calib/ece": 0.24623529411764708,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.047058823529411764,
      "calib/gap": 0.059203761755485806,
      "calib/mean_conf": 0.677607843137255,
      "calib/mu_c": 0.7112727272727273,
      "calib/mu_w": 0.6520689655172415,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.24623529411764708,
      "calib/std_conf": 0.13426109746097853,
      "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": 2843.0,
      "completions/max_terminated_length": 2843.0,
      "completions/mean_length": 588.4765625,
      "completions/mean_terminated_length": 588.4765625,
      "completions/min_length": 283.0,
      "completions/min_terminated_length": 283.0,
      "epoch": 0.15893333333333334,
      "grad_norm": 0.010876238346099854,
      "learning_rate": 1.4166666666666667e-06,
      "loss": 0.0435,
      "num_tokens": 33692700.0,
      "reward": 1.5448906421661377,
      "reward_std": 0.3771207332611084,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.7023417949676514,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9069079160690308,
      "step": 149
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.603125,
      "calib/avg_num_step_conf": 6.51953125,
      "calib/ece": 0.09409448818897644,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.015748031496062992,
      "calib/gap": 0.043016414141414105,
      "calib/mean_conf": 0.661023622047244,
      "calib/mu_c": 0.6796527777777778,
      "calib/mu_w": 0.6366363636363637,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.09409448818897644,
      "calib/std_conf": 0.11753595292250137,
      "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": 2878.0,
      "completions/max_terminated_length": 2878.0,
      "completions/mean_length": 517.515625,
      "completions/mean_terminated_length": 517.515625,
      "completions/min_length": 191.0,
      "completions/min_terminated_length": 191.0,
      "epoch": 0.16,
      "grad_norm": 0.01054420880973339,
      "learning_rate": 1.3888888888888892e-06,
      "loss": 0.0181,
      "num_tokens": 33930144.0,
      "reward": 1.7520647048950195,
      "reward_std": 0.4901063144207001,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.7453425526618958,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.9113534688949585,
      "step": 150
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6249682499364999,
      "calib/avg_num_step_conf": 6.59765625,
      "calib/ece": 0.1635458167330677,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.027888446215139442,
      "calib/gap": 0.058841757683515516,
      "calib/mean_conf": 0.6511952191235061,
      "calib/mu_c": 0.680967741935484,
      "calib/mu_w": 0.6221259842519685,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.16035856573705176,
      "calib/std_conf": 0.1367645697791074,
      "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": 2563.0,
      "completions/max_terminated_length": 2563.0,
      "completions/mean_length": 575.10546875,
      "completions/mean_terminated_length": 575.10546875,
      "completions/min_length": 261.0,
      "completions/min_terminated_length": 261.0,
      "epoch": 0.16106666666666666,
      "grad_norm": 0.011571124196052551,
      "learning_rate": 1.3611111111111112e-06,
      "loss": 0.0523,
      "num_tokens": 34184395.0,
      "reward": 1.6189985275268555,
      "reward_std": 0.4970731735229492,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.7191691398620605,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8974502086639404,
      "step": 151
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5705645161290323,
      "calib/avg_num_step_conf": 6.83984375,
      "calib/ece": 0.14840000000000003,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.008,
      "calib/gap": 0.031455453149001555,
      "calib/mean_conf": 0.6358400000000001,
      "calib/mu_c": 0.6516935483870969,
      "calib/mu_w": 0.6202380952380954,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.14412000000000003,
      "calib/std_conf": 0.1292358092790075,
      "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": 2007.0,
      "completions/max_terminated_length": 2007.0,
      "completions/mean_length": 501.0078125,
      "completions/mean_terminated_length": 513.0320434570312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 246.0,
      "epoch": 0.16213333333333332,
      "grad_norm": 0.01134491991251707,
      "learning_rate": 1.3333333333333334e-06,
      "loss": -0.0132,
      "num_tokens": 34418045.0,
      "reward": 1.6180343627929688,
      "reward_std": 0.458588182926178,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.712388277053833,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.9003739953041077,
      "step": 152
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.591959215281363,
      "calib/avg_num_step_conf": 6.7578125,
      "calib/ece": 0.08079051383399205,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.011857707509881422,
      "calib/gap": 0.03945211667527104,
      "calib/mean_conf": 0.6437154150197629,
      "calib/mu_c": 0.6599328859060403,
      "calib/mu_w": 0.6204807692307692,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.06778656126482208,
      "calib/std_conf": 0.12695751849145287,
      "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": 2129.0,
      "completions/max_terminated_length": 2129.0,
      "completions/mean_length": 523.40234375,
      "completions/mean_terminated_length": 527.5236206054688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 257.0,
      "epoch": 0.1632,
      "grad_norm": 0.011443248949944973,
      "learning_rate": 1.3055555555555556e-06,
      "loss": -0.0423,
      "num_tokens": 34659356.0,
      "reward": 1.7793679237365723,
      "reward_std": 0.4185265600681305,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7461199760437012,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9104143381118774,
      "step": 153
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5889194980104072,
      "calib/avg_num_step_conf": 5.8046875,
      "calib/ece": 0.161953125,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.015625,
      "calib/gap": 0.029592898683807545,
      "calib/mean_conf": 0.62546875,
      "calib/mu_c": 0.6410743801652891,
      "calib/mu_w": 0.6114814814814815,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.1573828125,
      "calib/std_conf": 0.1136734589666273,
      "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": 1215.0,
      "completions/max_terminated_length": 1215.0,
      "completions/mean_length": 478.8125,
      "completions/mean_terminated_length": 480.6902160644531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 233.0,
      "epoch": 0.16426666666666667,
      "grad_norm": 0.012007861398160458,
      "learning_rate": 1.2777777777777779e-06,
      "loss": 0.0362,
      "num_tokens": 34886372.0,
      "reward": 1.6205209493637085,
      "reward_std": 0.3650030791759491,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.7261112928390503,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9278475046157837,
      "step": 154
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6128,
      "calib/avg_num_step_conf": 5.8984375,
      "calib/ece": 0.1398431372549019,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.00392156862745098,
      "calib/gap": 0.057043076923077085,
      "calib/mean_conf": 0.6180392156862746,
      "calib/mu_c": 0.64712,
      "calib/mu_w": 0.5900769230769229,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.13384313725490193,
      "calib/std_conf": 0.14083810212847075,
      "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": 1113.0,
      "completions/max_terminated_length": 1113.0,
      "completions/mean_length": 467.5,
      "completions/mean_terminated_length": 469.3333740234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 241.0,
      "epoch": 0.16533333333333333,
      "grad_norm": 0.012868484482169151,
      "learning_rate": 1.25e-06,
      "loss": 0.0052,
      "num_tokens": 35113268.0,
      "reward": 1.648886799812317,
      "reward_std": 0.3540608584880829,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.7395272850990295,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9341446757316589,
      "step": 155
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6274998402657976,
      "calib/avg_num_step_conf": 6.26953125,
      "calib/ece": 0.05103174603174597,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.01984126984126984,
      "calib/gap": 0.07234809277362453,
      "calib/mean_conf": 0.6077777777777776,
      "calib/mu_c": 0.6396453900709219,
      "calib/mu_w": 0.5672972972972974,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.049642857142857086,
      "calib/std_conf": 0.14023443460516388,
      "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": 2975.0,
      "completions/max_terminated_length": 2975.0,
      "completions/mean_length": 542.72265625,
      "completions/mean_terminated_length": 544.8510131835938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 221.0,
      "epoch": 0.1664,
      "grad_norm": 0.010311256162822247,
      "learning_rate": 1.2222222222222223e-06,
      "loss": 0.058,
      "num_tokens": 35356965.0,
      "reward": 1.7402732372283936,
      "reward_std": 0.4105079770088196,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.7552226185798645,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9089950919151306,
      "step": 156
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6612486612486611,
      "calib/avg_num_step_conf": 6.85546875,
      "calib/ece": 0.08480314960629923,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.031496062992125984,
      "calib/gap": 0.08308196308196314,
      "calib/mean_conf": 0.6142519685039369,
      "calib/mu_c": 0.6505594405594406,
      "calib/mu_w": 0.5674774774774775,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.06803149606299214,
      "calib/std_conf": 0.15597051853990004,
      "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": 2884.0,
      "completions/max_terminated_length": 2884.0,
      "completions/mean_length": 515.26171875,
      "completions/mean_terminated_length": 515.26171875,
      "completions/min_length": 181.0,
      "completions/min_terminated_length": 181.0,
      "epoch": 0.16746666666666668,
      "grad_norm": 0.010643880814313889,
      "learning_rate": 1.1944444444444446e-06,
      "loss": 0.0583,
      "num_tokens": 35592600.0,
      "reward": 1.7511237859725952,
      "reward_std": 0.4374259114265442,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.7581789493560791,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.9181913733482361,
      "step": 157
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5130963213310802,
      "calib/avg_num_step_conf": 5.84375,
      "calib/ece": 0.1289019607843137,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.00784313725490196,
      "calib/gap": -0.0014753672169505183,
      "calib/mean_conf": 0.6472549019607844,
      "calib/mu_c": 0.6466878980891719,
      "calib/mu_w": 0.6481632653061224,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.08023529411764703,
      "calib/std_conf": 0.13824500035455026,
      "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": 2684.0,
      "completions/max_terminated_length": 2684.0,
      "completions/mean_length": 468.140625,
      "completions/mean_terminated_length": 468.140625,
      "completions/min_length": 232.0,
      "completions/min_terminated_length": 232.0,
      "epoch": 0.16853333333333334,
      "grad_norm": 0.012019583955407143,
      "learning_rate": 1.1666666666666668e-06,
      "loss": 0.039,
      "num_tokens": 35817684.0,
      "reward": 1.8365700244903564,
      "reward_std": 0.4772275984287262,
      "rewards/accuracy_reward_step": 0.61328125,
      "rewards/final_brier_reward_step": 0.7396761775016785,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9347293376922607,
      "step": 158
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5564952213279678,
      "calib/avg_num_step_conf": 5.64453125,
      "calib/ece": 0.12019685039370082,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.003937007874015748,
      "calib/gap": 0.02273390342052295,
      "calib/mean_conf": 0.6061023622047244,
      "calib/mu_c": 0.6161267605633801,
      "calib/mu_w": 0.5933928571428572,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.08362204724409446,
      "calib/std_conf": 0.1555387049405307,
      "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": 1125.0,
      "completions/max_terminated_length": 1125.0,
      "completions/mean_length": 447.15625,
      "completions/mean_terminated_length": 448.9098205566406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 217.0,
      "epoch": 0.1696,
      "grad_norm": 0.01172886323183775,
      "learning_rate": 1.138888888888889e-06,
      "loss": -0.0039,
      "num_tokens": 36036940.0,
      "reward": 1.7441353797912598,
      "reward_std": 0.3740137219429016,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.7325222492218018,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9315195679664612,
      "step": 159
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5577742946708463,
      "calib/avg_num_step_conf": 5.7734375,
      "calib/ece": 0.09682352941176474,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.023529411764705882,
      "calib/gap": 0.027965517241379412,
      "calib/mean_conf": 0.6369019607843138,
      "calib/mu_c": 0.6489655172413793,
      "calib/mu_w": 0.6209999999999999,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.08254901960784318,
      "calib/std_conf": 0.1524833973478541,
      "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": 2676.0,
      "completions/max_terminated_length": 2676.0,
      "completions/mean_length": 498.484375,
      "completions/mean_terminated_length": 498.484375,
      "completions/min_length": 198.0,
      "completions/min_terminated_length": 198.0,
      "epoch": 0.17066666666666666,
      "grad_norm": 0.011179703287780285,
      "learning_rate": 1.111111111111111e-06,
      "loss": 0.0211,
      "num_tokens": 36269392.0,
      "reward": 1.7648649215698242,
      "reward_std": 0.3944992423057556,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.7376238107681274,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9312103986740112,
      "step": 160
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6658963585434174,
      "calib/avg_num_step_conf": 5.90625,
      "calib/ece": 0.08980314960629919,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.015748031496062992,
      "calib/gap": 0.0897212885154064,
      "calib/mean_conf": 0.6137401574803149,
      "calib/mu_c": 0.6434117647058825,
      "calib/mu_w": 0.5536904761904761,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.017125984251968505,
      "calib/std_conf": 0.16111260367997945,
      "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": 2777.0,
      "completions/max_terminated_length": 2777.0,
      "completions/mean_length": 477.1328125,
      "completions/mean_terminated_length": 477.1328125,
      "completions/min_length": 167.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.17173333333333332,
      "grad_norm": 0.012054560706019402,
      "learning_rate": 1.0833333333333335e-06,
      "loss": 0.0817,
      "num_tokens": 36495458.0,
      "reward": 1.9207186698913574,
      "reward_std": 0.40740329027175903,
      "rewards/accuracy_reward_step": 0.6640625,
      "rewards/final_brier_reward_step": 0.7831676006317139,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9309569597244263,
      "step": 161
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5349786931818181,
      "calib/avg_num_step_conf": 5.734375,
      "calib/ece": 0.134921875,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.01171875,
      "calib/gap": 0.024159090909090852,
      "calib/mean_conf": 0.617734375,
      "calib/mu_c": 0.6252840909090909,
      "calib/mu_w": 0.601125,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.03257812500000002,
      "calib/std_conf": 0.16187241717896034,
      "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": 978.0,
      "completions/max_terminated_length": 978.0,
      "completions/mean_length": 450.125,
      "completions/mean_terminated_length": 451.8902282714844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 180.0,
      "epoch": 0.1728,
      "grad_norm": 0.011963211931288242,
      "learning_rate": 1.0555555555555557e-06,
      "loss": 0.0336,
      "num_tokens": 36714834.0,
      "reward": 1.9583287239074707,
      "reward_std": 0.3349304795265198,
      "rewards/accuracy_reward_step": 0.6875,
      "rewards/final_brier_reward_step": 0.764467179775238,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.9438478946685791,
      "step": 162
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6884183768367537,
      "calib/avg_num_step_conf": 6.05078125,
      "calib/ece": 0.1347637795275591,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.023622047244094488,
      "calib/gap": 0.09440944881889746,
      "calib/mean_conf": 0.6204330708661416,
      "calib/mu_c": 0.6676377952755905,
      "calib/mu_w": 0.5732283464566931,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.12759842519685038,
      "calib/std_conf": 0.15417359530410835,
      "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": 2032.0,
      "completions/max_terminated_length": 2032.0,
      "completions/mean_length": 519.28515625,
      "completions/mean_terminated_length": 521.3215942382812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 193.0,
      "epoch": 0.17386666666666667,
      "grad_norm": 0.010253848508000374,
      "learning_rate": 1.0277777777777777e-06,
      "loss": 0.0155,
      "num_tokens": 36952603.0,
      "reward": 1.65916907787323,
      "reward_std": 0.3765425980091095,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.7530019283294678,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9227369427680969,
      "step": 163
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6116176005064894,
      "calib/avg_num_step_conf": 6.2265625,
      "calib/ece": 0.12186507936507934,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.027777777777777776,
      "calib/gap": 0.06605698005698024,
      "calib/mean_conf": 0.6238492063492063,
      "calib/mu_c": 0.6545185185185186,
      "calib/mu_w": 0.5884615384615384,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.10499999999999998,
      "calib/std_conf": 0.16777448807424944,
      "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": 2938.0,
      "completions/max_terminated_length": 2938.0,
      "completions/mean_length": 537.68359375,
      "completions/mean_terminated_length": 539.7921752929688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 253.0,
      "epoch": 0.17493333333333333,
      "grad_norm": 0.010474850423634052,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0236,
      "num_tokens": 37196386.0,
      "reward": 1.6978435516357422,
      "reward_std": 0.41331610083580017,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.7365285158157349,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9220330715179443,
      "step": 164
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5495690991381983,
      "calib/avg_num_step_conf": 5.90234375,
      "calib/ece": 0.1635826771653543,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.027559055118110236,
      "calib/gap": 0.03362204724409468,
      "calib/mean_conf": 0.6525590551181102,
      "calib/mu_c": 0.6693700787401576,
      "calib/mu_w": 0.635748031496063,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.15807086614173227,
      "calib/std_conf": 0.1584426746749824,
      "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": 2493.0,
      "completions/max_terminated_length": 2493.0,
      "completions/mean_length": 505.2109375,
      "completions/mean_terminated_length": 505.2109375,
      "completions/min_length": 173.0,
      "completions/min_terminated_length": 173.0,
      "epoch": 0.176,
      "grad_norm": 0.010923215188086033,
      "learning_rate": 9.722222222222224e-07,
      "loss": 0.0323,
      "num_tokens": 37431296.0,
      "reward": 1.6498363018035889,
      "reward_std": 0.42734792828559875,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.7128199338912964,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.925588071346283,
      "step": 165
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6211840888066605,
      "calib/avg_num_step_conf": 6.22265625,
      "calib/ece": 0.10615686274509803,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.027450980392156862,
      "calib/gap": 0.06878485529271827,
      "calib/mean_conf": 0.6639607843137255,
      "calib/mu_c": 0.6893167701863353,
      "calib/mu_w": 0.620531914893617,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.06937254901960783,
      "calib/std_conf": 0.17315186135176522,
      "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": 3072.0,
      "completions/max_terminated_length": 3072.0,
      "completions/mean_length": 499.42578125,
      "completions/mean_terminated_length": 499.42578125,
      "completions/min_length": 216.0,
      "completions/min_terminated_length": 216.0,
      "epoch": 0.17706666666666668,
      "grad_norm": 0.011256815865635872,
      "learning_rate": 9.444444444444445e-07,
      "loss": 0.0294,
      "num_tokens": 37665333.0,
      "reward": 1.859373688697815,
      "reward_std": 0.33840852975845337,
      "rewards/accuracy_reward_step": 0.62890625,
      "rewards/final_brier_reward_step": 0.7575469017028809,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.9299476146697998,
      "step": 166
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5554705710955712,
      "calib/avg_num_step_conf": 6.4765625,
      "calib/ece": 0.12082677165354336,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.047244094488188976,
      "calib/gap": 0.04326194638694636,
      "calib/mean_conf": 0.6717716535433071,
      "calib/mu_c": 0.6850568181818182,
      "calib/mu_w": 0.6417948717948718,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0498425196850394,
      "calib/std_conf": 0.173406158741239,
      "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": 2718.0,
      "completions/max_terminated_length": 2718.0,
      "completions/mean_length": 500.79296875,
      "completions/mean_terminated_length": 502.75689697265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 237.0,
      "epoch": 0.17813333333333334,
      "grad_norm": 0.011857429519295692,
      "learning_rate": 9.166666666666666e-07,
      "loss": 0.0083,
      "num_tokens": 37899144.0,
      "reward": 1.94930100440979,
      "reward_std": 0.2797129452228546,
      "rewards/accuracy_reward_step": 0.6875,
      "rewards/final_brier_reward_step": 0.7690542936325073,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9187746644020081,
      "step": 167
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6973217677599088,
      "calib/avg_num_step_conf": 5.78515625,
      "calib/ece": 0.0985882352941177,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.09019607843137255,
      "calib/gap": 0.1266328985690769,
      "calib/mean_conf": 0.6668235294117648,
      "calib/mu_c": 0.7194630872483222,
      "calib/mu_w": 0.5928301886792453,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.09054901960784317,
      "calib/std_conf": 0.18775586003046227,
      "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": 2249.0,
      "completions/max_terminated_length": 2249.0,
      "completions/mean_length": 497.48828125,
      "completions/mean_terminated_length": 497.48828125,
      "completions/min_length": 214.0,
      "completions/min_terminated_length": 214.0,
      "epoch": 0.1792,
      "grad_norm": 0.012170032598078251,
      "learning_rate": 8.88888888888889e-07,
      "loss": 0.0252,
      "num_tokens": 38131173.0,
      "reward": 1.7952685356140137,
      "reward_std": 0.471716046333313,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7702500224113464,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9342615604400635,
      "step": 168
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6149819269599901,
      "calib/avg_num_step_conf": 5.953125,
      "calib/ece": 0.17509803921568623,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.08627450980392157,
      "calib/gap": 0.08630188208899436,
      "calib/mean_conf": 0.7182352941176472,
      "calib/mu_c": 0.7564788732394367,
      "calib/mu_w": 0.6701769911504424,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.168235294117647,
      "calib/std_conf": 0.18560030724231816,
      "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": 2180.0,
      "completions/max_terminated_length": 2180.0,
      "completions/mean_length": 493.14453125,
      "completions/mean_terminated_length": 495.0784606933594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.18026666666666666,
      "grad_norm": 0.01110941730439663,
      "learning_rate": 8.611111111111112e-07,
      "loss": -0.0118,
      "num_tokens": 38361602.0,
      "reward": 1.7430405616760254,
      "reward_std": 0.4263700544834137,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.7285687327384949,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9310938119888306,
      "step": 169
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6479359250348233,
      "calib/avg_num_step_conf": 6.6796875,
      "calib/ece": 0.15203921568627451,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.08627450980392157,
      "calib/gap": 0.09772825123464624,
      "calib/mean_conf": 0.7312549019607842,
      "calib/mu_c": 0.7718791946308726,
      "calib/mu_w": 0.6741509433962264,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1494901960784314,
      "calib/std_conf": 0.1699919078970543,
      "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": 2789.0,
      "completions/max_terminated_length": 2789.0,
      "completions/mean_length": 507.7734375,
      "completions/mean_terminated_length": 507.7734375,
      "completions/min_length": 231.0,
      "completions/min_terminated_length": 231.0,
      "epoch": 0.18133333333333335,
      "grad_norm": 0.010340871289372444,
      "learning_rate": 8.333333333333333e-07,
      "loss": 0.0158,
      "num_tokens": 38595744.0,
      "reward": 1.7924294471740723,
      "reward_std": 0.3177809715270996,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7511488199234009,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9341940879821777,
      "step": 170
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5383407243163341,
      "calib/avg_num_step_conf": 5.66796875,
      "calib/ece": 0.29325490196078435,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.13333333333333333,
      "calib/gap": 0.021783074648928258,
      "calib/mean_conf": 0.7370980392156863,
      "calib/mu_c": 0.7483739837398373,
      "calib/mu_w": 0.7265909090909091,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.274,
      "calib/std_conf": 0.19498066821238325,
      "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": 1677.0,
      "completions/max_terminated_length": 1677.0,
      "completions/mean_length": 497.26953125,
      "completions/mean_terminated_length": 499.2196350097656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 211.0,
      "epoch": 0.1824,
      "grad_norm": 0.010973098687827587,
      "learning_rate": 8.055555555555557e-07,
      "loss": 0.0116,
      "num_tokens": 38829941.0,
      "reward": 1.6121565103530884,
      "reward_std": 0.4386375844478607,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.6518382430076599,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9296000003814697,
      "step": 171
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6243540051679586,
      "calib/avg_num_step_conf": 5.921875,
      "calib/ece": 0.13403162055335982,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.17391304347826086,
      "calib/gap": 0.06261915015790998,
      "calib/mean_conf": 0.7865217391304348,
      "calib/mu_c": 0.8065697674418606,
      "calib/mu_w": 0.7439506172839506,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.12035573122529655,
      "calib/std_conf": 0.1586970057877189,
      "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": 2656.0,
      "completions/max_terminated_length": 2656.0,
      "completions/mean_length": 497.6328125,
      "completions/mean_terminated_length": 497.6328125,
      "completions/min_length": 204.0,
      "completions/min_terminated_length": 204.0,
      "epoch": 0.18346666666666667,
      "grad_norm": 0.011852969415485859,
      "learning_rate": 7.777777777777779e-07,
      "loss": 0.0177,
      "num_tokens": 39060687.0,
      "reward": 1.9252095222473145,
      "reward_std": 0.41335391998291016,
      "rewards/accuracy_reward_step": 0.671875,
      "rewards/final_brier_reward_step": 0.7639777660369873,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.9290477633476257,
      "step": 172
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5206237424547284,
      "calib/avg_num_step_conf": 6.6015625,
      "calib/ece": 0.24763779527559063,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.15748031496062992,
      "calib/gap": 0.05143234406438624,
      "calib/mean_conf": 0.7633070866141732,
      "calib/mu_c": 0.7859859154929577,
      "calib/mu_w": 0.7345535714285715,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.22594488188976383,
      "calib/std_conf": 0.20169187975530575,
      "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": 2978.0,
      "completions/max_terminated_length": 2978.0,
      "completions/mean_length": 534.05859375,
      "completions/mean_terminated_length": 534.05859375,
      "completions/min_length": 185.0,
      "completions/min_terminated_length": 185.0,
      "epoch": 0.18453333333333333,
      "grad_norm": 0.010401809588074684,
      "learning_rate": 7.5e-07,
      "loss": 0.002,
      "num_tokens": 39300566.0,
      "reward": 1.7318034172058105,
      "reward_std": 0.47870710492134094,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6910054683685303,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9237080812454224,
      "step": 173
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5059820538384845,
      "calib/avg_num_step_conf": 6.05078125,
      "calib/ece": 0.29574803149606305,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.21653543307086615,
      "calib/gap": 0.016704885343967857,
      "calib/mean_conf": 0.7960629921259843,
      "calib/mu_c": 0.8038235294117646,
      "calib/mu_w": 0.7871186440677967,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2781889763779528,
      "calib/std_conf": 0.17347834501594825,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2614.0,
      "completions/max_terminated_length": 2614.0,
      "completions/mean_length": 549.515625,
      "completions/mean_terminated_length": 549.515625,
      "completions/min_length": 249.0,
      "completions/min_terminated_length": 249.0,
      "epoch": 0.1856,
      "grad_norm": 0.011031479574739933,
      "learning_rate": 7.222222222222222e-07,
      "loss": 0.0039,
      "num_tokens": 39545474.0,
      "reward": 1.6889164447784424,
      "reward_std": 0.5492876768112183,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.656374990940094,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9274158477783203,
      "step": 174
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6095146871008942,
      "calib/avg_num_step_conf": 5.74609375,
      "calib/ece": 0.34766798418972344,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.16996047430830039,
      "calib/gap": 0.0715989782886336,
      "calib/mean_conf": 0.7741501976284585,
      "calib/mu_c": 0.8151851851851852,
      "calib/mu_w": 0.7435862068965516,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3474703557312254,
      "calib/std_conf": 0.1827365789396574,
      "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": 1173.0,
      "completions/max_terminated_length": 1173.0,
      "completions/mean_length": 506.0078125,
      "completions/mean_terminated_length": 507.9921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 233.0,
      "epoch": 0.18666666666666668,
      "grad_norm": 0.011073246598243713,
      "learning_rate": 6.944444444444446e-07,
      "loss": -0.0056,
      "num_tokens": 39780836.0,
      "reward": 1.515343189239502,
      "reward_std": 0.41250574588775635,
      "rewards/accuracy_reward_step": 0.421875,
      "rewards/final_brier_reward_step": 0.628931999206543,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.9246280193328857,
      "step": 175
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6658396946564885,
      "calib/avg_num_step_conf": 5.921875,
      "calib/ece": 0.2586454183266932,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.14741035856573706,
      "calib/gap": 0.10823982188295167,
      "calib/mean_conf": 0.7529083665338645,
      "calib/mu_c": 0.8046564885496182,
      "calib/mu_w": 0.6964166666666666,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.24482071713147407,
      "calib/std_conf": 0.21151866816381465,
      "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": 2594.0,
      "completions/max_terminated_length": 2594.0,
      "completions/mean_length": 515.3125,
      "completions/mean_terminated_length": 517.3333740234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 184.0,
      "epoch": 0.18773333333333334,
      "grad_norm": 0.01043481845408678,
      "learning_rate": 6.666666666666667e-07,
      "loss": 0.0522,
      "num_tokens": 40016820.0,
      "reward": 1.6600654125213623,
      "reward_std": 0.4616318941116333,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.6926000118255615,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.9164116382598877,
      "step": 176
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6783281339530146,
      "calib/avg_num_step_conf": 6.18359375,
      "calib/ece": 0.24646825396825398,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.2222222222222222,
      "calib/gap": 0.11471764181575084,
      "calib/mean_conf": 0.7980555555555556,
      "calib/mu_c": 0.8494964028776978,
      "calib/mu_w": 0.734778761061947,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.24646825396825398,
      "calib/std_conf": 0.18074962098164638,
      "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": 2641.0,
      "completions/max_terminated_length": 2641.0,
      "completions/mean_length": 528.4921875,
      "completions/mean_terminated_length": 530.5647583007812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 219.0,
      "epoch": 0.1888,
      "grad_norm": 0.011687429621815681,
      "learning_rate": 6.388888888888889e-07,
      "loss": 0.0367,
      "num_tokens": 40255946.0,
      "reward": 1.7122498750686646,
      "reward_std": 0.40932321548461914,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.7048050165176392,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9176322221755981,
      "step": 177
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5829978915085298,
      "calib/avg_num_step_conf": 5.95703125,
      "calib/ece": 0.28884920634920624,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.25,
      "calib/gap": 0.04688709986582351,
      "calib/mean_conf": 0.8301984126984128,
      "calib/mu_c": 0.8508510638297875,
      "calib/mu_w": 0.803963963963964,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.27976190476190466,
      "calib/std_conf": 0.1538125286370861,
      "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": 2468.0,
      "completions/max_terminated_length": 2468.0,
      "completions/mean_length": 478.14453125,
      "completions/mean_terminated_length": 481.9094543457031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 235.0,
      "epoch": 0.18986666666666666,
      "grad_norm": 0.011508328840136528,
      "learning_rate": 6.111111111111112e-07,
      "loss": 0.0095,
      "num_tokens": 40484423.0,
      "reward": 1.7157710790634155,
      "reward_std": 0.5118218660354614,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.6691105365753174,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9205360412597656,
      "step": 178
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6051773729626079,
      "calib/avg_num_step_conf": 5.796875,
      "calib/ece": 0.24692913385826748,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.25196850393700787,
      "calib/gap": 0.07034196228827116,
      "calib/mean_conf": 0.8307874015748031,
      "calib/mu_c": 0.8598657718120806,
      "calib/mu_w": 0.7895238095238094,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.245551181102362,
      "calib/std_conf": 0.14758212073609975,
      "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": 1484.0,
      "completions/max_terminated_length": 1484.0,
      "completions/mean_length": 496.484375,
      "completions/mean_terminated_length": 500.3937072753906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 217.0,
      "epoch": 0.19093333333333334,
      "grad_norm": 0.010500260628759861,
      "learning_rate": 5.833333333333334e-07,
      "loss": 0.0041,
      "num_tokens": 40717787.0,
      "reward": 1.777552843093872,
      "reward_std": 0.45863401889801025,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7046679258346558,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9289811849594116,
      "step": 179
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5870261437908497,
      "calib/avg_num_step_conf": 6.9765625,
      "calib/ece": 0.20739130434782602,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.2490118577075099,
      "calib/gap": 0.07275947712418296,
      "calib/mean_conf": 0.8066007905138339,
      "calib/mu_c": 0.835359477124183,
      "calib/mu_w": 0.7626000000000001,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.20462450592885367,
      "calib/std_conf": 0.1668929822049343,
      "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": 2883.0,
      "completions/max_terminated_length": 2883.0,
      "completions/mean_length": 528.54296875,
      "completions/mean_terminated_length": 530.61572265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 237.0,
      "epoch": 0.192,
      "grad_norm": 0.013314173556864262,
      "learning_rate": 5.555555555555555e-07,
      "loss": 0.0148,
      "num_tokens": 40956950.0,
      "reward": 1.8024948835372925,
      "reward_std": 0.40265530347824097,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.7186331748962402,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.9288461208343506,
      "step": 180
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.636540513399862,
      "calib/avg_num_step_conf": 5.78125,
      "calib/ece": 0.3806692913385827,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.2047244094488189,
      "calib/gap": 0.06348145358689505,
      "calib/mean_conf": 0.821220472440945,
      "calib/mu_c": 0.8564601769911504,
      "calib/mu_w": 0.7929787234042553,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.378503937007874,
      "calib/std_conf": 0.1561535895132963,
      "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": 2767.0,
      "completions/max_terminated_length": 2767.0,
      "completions/mean_length": 492.48046875,
      "completions/mean_terminated_length": 494.41180419921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 195.0,
      "epoch": 0.19306666666666666,
      "grad_norm": 0.01176389493048191,
      "learning_rate": 5.277777777777779e-07,
      "loss": 0.0108,
      "num_tokens": 41189289.0,
      "reward": 1.5452067852020264,
      "reward_std": 0.444354772567749,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.6135472655296326,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9344671964645386,
      "step": 181
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6405498505134538,
      "calib/avg_num_step_conf": 5.984375,
      "calib/ece": 0.22262745098039208,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.33725490196078434,
      "calib/gap": 0.0973293903548682,
      "calib/mean_conf": 0.8363529411764705,
      "calib/mu_c": 0.8737579617834396,
      "calib/mu_w": 0.7764285714285714,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.22164705882352934,
      "calib/std_conf": 0.15809786290793232,
      "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": 2870.0,
      "completions/max_terminated_length": 2870.0,
      "completions/mean_length": 515.68359375,
      "completions/mean_terminated_length": 515.68359375,
      "completions/min_length": 266.0,
      "completions/min_terminated_length": 266.0,
      "epoch": 0.19413333333333332,
      "grad_norm": 0.010333042591810226,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0207,
      "num_tokens": 41427464.0,
      "reward": 1.8302533626556396,
      "reward_std": 0.40433841943740845,
      "rewards/accuracy_reward_step": 0.61328125,
      "rewards/final_brier_reward_step": 0.729537844657898,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9274133443832397,
      "step": 182
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6521644753162986,
      "calib/avg_num_step_conf": 6.12890625,
      "calib/ece": 0.2826274509803923,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.3176470588235294,
      "calib/gap": 0.10953671545522214,
      "calib/mean_conf": 0.8277254901960784,
      "calib/mu_c": 0.8775539568345324,
      "calib/mu_w": 0.7680172413793103,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2826274509803923,
      "calib/std_conf": 0.17051572044499988,
      "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": 1567.0,
      "completions/max_terminated_length": 1567.0,
      "completions/mean_length": 526.44921875,
      "completions/mean_terminated_length": 528.5137329101562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 229.0,
      "epoch": 0.1952,
      "grad_norm": 0.010599082335829735,
      "learning_rate": 4.7222222222222226e-07,
      "loss": -0.0175,
      "num_tokens": 41668915.0,
      "reward": 1.7200071811676025,
      "reward_std": 0.47910067439079285,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.694678544998169,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9353501796722412,
      "step": 183
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5091120060064159,
      "calib/avg_num_step_conf": 6.234375,
      "calib/ece": 0.22690476190476203,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.3333333333333333,
      "calib/gap": 0.04325848064978499,
      "calib/mean_conf": 0.8535714285714286,
      "calib/mu_c": 0.8691925465838509,
      "calib/mu_w": 0.8259340659340659,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2207936507936509,
      "calib/std_conf": 0.13639906603573024,
      "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": 3048.0,
      "completions/max_terminated_length": 3048.0,
      "completions/mean_length": 523.75390625,
      "completions/mean_terminated_length": 527.8779296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 245.0,
      "epoch": 0.19626666666666667,
      "grad_norm": 0.011380488984286785,
      "learning_rate": 4.444444444444445e-07,
      "loss": 0.0052,
      "num_tokens": 41908276.0,
      "reward": 1.8424732685089111,
      "reward_std": 0.541731595993042,
      "rewards/accuracy_reward_step": 0.62890625,
      "rewards/final_brier_reward_step": 0.7132359743118286,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9144695997238159,
      "step": 184
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6891203703703704,
      "calib/avg_num_step_conf": 5.3828125,
      "calib/ece": 0.25826612903225804,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.2701612903225806,
      "calib/gap": 0.1236746031746031,
      "calib/mean_conf": 0.8159274193548386,
      "calib/mu_c": 0.8697857142857144,
      "calib/mu_w": 0.7461111111111113,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.25483870967741934,
      "calib/std_conf": 0.17083271733685884,
      "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": 2932.0,
      "completions/max_terminated_length": 2932.0,
      "completions/mean_length": 525.87890625,
      "completions/mean_terminated_length": 534.2261962890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 241.0,
      "epoch": 0.19733333333333333,
      "grad_norm": 0.009946313686668873,
      "learning_rate": 4.1666666666666667e-07,
      "loss": 0.0001,
      "num_tokens": 42149821.0,
      "reward": 1.7037546634674072,
      "reward_std": 0.407586008310318,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6962480545043945,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.9078329205513,
      "step": 185
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.4974448460675558,
      "calib/avg_num_step_conf": 5.62109375,
      "calib/ece": 0.30156862745098034,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.23529411764705882,
      "calib/gap": -0.0019986289417922576,
      "calib/mean_conf": 0.8072941176470587,
      "calib/mu_c": 0.8064084507042254,
      "calib/mu_w": 0.8084070796460177,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.27599999999999997,
      "calib/std_conf": 0.17284735350145441,
      "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": 1195.0,
      "completions/max_terminated_length": 1195.0,
      "completions/mean_length": 495.16015625,
      "completions/mean_terminated_length": 497.10198974609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 179.0,
      "epoch": 0.1984,
      "grad_norm": 0.01056612841784954,
      "learning_rate": 3.8888888888888895e-07,
      "loss": -0.0087,
      "num_tokens": 42381622.0,
      "reward": 1.7294514179229736,
      "reward_std": 0.42536190152168274,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6570781469345093,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9404152631759644,
      "step": 186
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.549240974614109,
      "calib/avg_num_step_conf": 6.76171875,
      "calib/ece": 0.2944223107569721,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.2749003984063745,
      "calib/gap": 0.032536037759918424,
      "calib/mean_conf": 0.8079681274900399,
      "calib/mu_c": 0.8231343283582089,
      "calib/mu_w": 0.7905982905982905,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.28426294820717135,
      "calib/std_conf": 0.17380054219194718,
      "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": 2903.0,
      "completions/max_terminated_length": 2903.0,
      "completions/mean_length": 550.0546875,
      "completions/mean_terminated_length": 554.3858032226562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 221.0,
      "epoch": 0.19946666666666665,
      "grad_norm": 0.009696499444544315,
      "learning_rate": 3.611111111111111e-07,
      "loss": 0.0437,
      "num_tokens": 42623980.0,
      "reward": 1.6667132377624512,
      "reward_std": 0.49259153008461,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.6490710973739624,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.9162194132804871,
      "step": 187
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5434809809809811,
      "calib/avg_num_step_conf": 5.87109375,
      "calib/ece": 0.263046875,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.31640625,
      "calib/gap": 0.04141641641641658,
      "calib/mean_conf": 0.823203125,
      "calib/mu_c": 0.8406756756756758,
      "calib/mu_w": 0.7992592592592592,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.25406249999999997,
      "calib/std_conf": 0.17621867449346668,
      "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": 1513.0,
      "completions/max_terminated_length": 1513.0,
      "completions/mean_length": 506.57421875,
      "completions/mean_terminated_length": 508.5608215332031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 172.0,
      "epoch": 0.20053333333333334,
      "grad_norm": 0.011942905373871326,
      "learning_rate": 3.3333333333333335e-07,
      "loss": 0.0329,
      "num_tokens": 42857735.0,
      "reward": 1.7743031978607178,
      "reward_std": 0.4323691129684448,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.6851898431777954,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.9432730674743652,
      "step": 188
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5701013513513514,
      "calib/avg_num_step_conf": 5.5234375,
      "calib/ece": 0.2772549019607843,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.2823529411764706,
      "calib/gap": 0.048393393393393325,
      "calib/mean_conf": 0.816156862745098,
      "calib/mu_c": 0.8372222222222222,
      "calib/mu_w": 0.7888288288288289,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.26435294117647057,
      "calib/std_conf": 0.18745806041268814,
      "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": 3064.0,
      "completions/max_terminated_length": 3064.0,
      "completions/mean_length": 493.03125,
      "completions/mean_terminated_length": 493.03125,
      "completions/min_length": 197.0,
      "completions/min_terminated_length": 197.0,
      "epoch": 0.2016,
      "grad_norm": 0.012970229610800743,
      "learning_rate": 3.055555555555556e-07,
      "loss": 0.0477,
      "num_tokens": 43091719.0,
      "reward": 1.7471036911010742,
      "reward_std": 0.34374135732650757,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.6769554615020752,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9442719221115112,
      "step": 189
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6959006211180124,
      "calib/avg_num_step_conf": 5.6328125,
      "calib/ece": 0.27568627450980404,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.23137254901960785,
      "calib/gap": 0.09738198757763972,
      "calib/mean_conf": 0.8097254901960784,
      "calib/mu_c": 0.8536428571428571,
      "calib/mu_w": 0.7562608695652174,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.26819607843137266,
      "calib/std_conf": 0.1693873711559178,
      "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": 1197.0,
      "completions/max_terminated_length": 1197.0,
      "completions/mean_length": 519.3828125,
      "completions/mean_terminated_length": 521.4196166992188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 264.0,
      "epoch": 0.20266666666666666,
      "grad_norm": 0.010747202672064304,
      "learning_rate": 2.7777777777777776e-07,
      "loss": 0.0033,
      "num_tokens": 43330289.0,
      "reward": 1.7282007932662964,
      "reward_std": 0.4465250074863434,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.7012163996696472,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9381492733955383,
      "step": 190
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5797524142847884,
      "calib/avg_num_step_conf": 6.37890625,
      "calib/ece": 0.38432,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.276,
      "calib/gap": 0.05565493551105072,
      "calib/mean_conf": 0.8132,
      "calib/mu_c": 0.8441441441441442,
      "calib/mu_w": 0.7884892086330935,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.37676,
      "calib/std_conf": 0.17352625161629,
      "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": 3028.0,
      "completions/max_terminated_length": 3028.0,
      "completions/mean_length": 523.70703125,
      "completions/mean_terminated_length": 523.70703125,
      "completions/min_length": 180.0,
      "completions/min_terminated_length": 180.0,
      "epoch": 0.20373333333333332,
      "grad_norm": 0.011537563987076283,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0868,
      "num_tokens": 43568526.0,
      "reward": 1.5195637941360474,
      "reward_std": 0.4578823149204254,
      "rewards/accuracy_reward_step": 0.43359375,
      "rewards/final_brier_reward_step": 0.5997992157936096,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.9237686395645142,
      "step": 191
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6606349206349206,
      "calib/avg_num_step_conf": 5.88671875,
      "calib/ece": 0.22733333333333322,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.24313725490196078,
      "calib/gap": 0.10824761904761904,
      "calib/mean_conf": 0.8106274509803921,
      "calib/mu_c": 0.8552,
      "calib/mu_w": 0.7469523809523809,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2248627450980391,
      "calib/std_conf": 0.18391280551053915,
      "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": 1478.0,
      "completions/max_terminated_length": 1478.0,
      "completions/mean_length": 497.00390625,
      "completions/mean_terminated_length": 498.9529724121094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.2048,
      "grad_norm": 0.01216538529843092,
      "learning_rate": 2.2222222222222224e-07,
      "loss": 0.0048,
      "num_tokens": 43800735.0,
      "reward": 1.7913602590560913,
      "reward_std": 0.41253167390823364,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.7241019606590271,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.9335265159606934,
      "step": 192
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6474137931034483,
      "calib/avg_num_step_conf": 5.59375,
      "calib/ece": 0.242806324110672,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.25296442687747034,
      "calib/gap": 0.08819476372924662,
      "calib/mean_conf": 0.7889723320158103,
      "calib/mu_c": 0.8266206896551725,
      "calib/mu_w": 0.7384259259259259,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.22932806324110677,
      "calib/std_conf": 0.19632526055027869,
      "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": 2200.0,
      "completions/max_terminated_length": 2200.0,
      "completions/mean_length": 498.4375,
      "completions/mean_terminated_length": 500.3921813964844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 199.0,
      "epoch": 0.20586666666666667,
      "grad_norm": 0.010590976104140282,
      "learning_rate": 1.9444444444444447e-07,
      "loss": 0.0293,
      "num_tokens": 44034047.0,
      "reward": 1.7486381530761719,
      "reward_std": 0.504930853843689,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.7042644023895264,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9231005907058716,
      "step": 193
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5860252148278242,
      "calib/avg_num_step_conf": 5.29296875,
      "calib/ece": 0.24011718750000005,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.328125,
      "calib/gap": 0.06934516715800043,
      "calib/mean_conf": 0.7983984375,
      "calib/mu_c": 0.8273825503355705,
      "calib/mu_w": 0.7580373831775701,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.22824218750000005,
      "calib/std_conf": 0.20291858126489695,
      "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": 1110.0,
      "completions/max_terminated_length": 1110.0,
      "completions/mean_length": 455.2109375,
      "completions/mean_terminated_length": 456.99609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 188.0,
      "epoch": 0.20693333333333333,
      "grad_norm": 0.012992671690881252,
      "learning_rate": 1.6666666666666668e-07,
      "loss": 0.0068,
      "num_tokens": 44256525.0,
      "reward": 1.7848504781723022,
      "reward_std": 0.37915557622909546,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7024776935577393,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.9447369575500488,
      "step": 194
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6206853890882097,
      "calib/avg_num_step_conf": 5.703125,
      "calib/ece": 0.30819607843137264,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.2980392156862745,
      "calib/gap": 0.04618644067796607,
      "calib/mean_conf": 0.8186274509803921,
      "calib/mu_c": 0.8399999999999999,
      "calib/mu_w": 0.7938135593220338,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2947843137254903,
      "calib/std_conf": 0.17406579337398334,
      "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": 1543.0,
      "completions/max_terminated_length": 1543.0,
      "completions/mean_length": 483.6640625,
      "completions/mean_terminated_length": 485.5608215332031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 204.0,
      "epoch": 0.208,
      "grad_norm": 0.011555323377251625,
      "learning_rate": 1.3888888888888888e-07,
      "loss": -0.0162,
      "num_tokens": 44486327.0,
      "reward": 1.7013813257217407,
      "reward_std": 0.37319985032081604,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.6622863411903381,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.94011390209198,
      "step": 195
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6548038229376258,
      "calib/avg_num_step_conf": 5.53125,
      "calib/ece": 0.2738188976377952,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.31496062992125984,
      "calib/gap": 0.09700704225352119,
      "calib/mean_conf": 0.8167322834645668,
      "calib/mu_c": 0.8595070422535213,
      "calib/mu_w": 0.7625000000000001,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.26574803149606296,
      "calib/std_conf": 0.17315947539308332,
      "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": 981.0,
      "completions/max_terminated_length": 981.0,
      "completions/mean_length": 454.4375,
      "completions/mean_terminated_length": 456.2196350097656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 252.0,
      "epoch": 0.20906666666666668,
      "grad_norm": 0.01108147669583559,
      "learning_rate": 1.1111111111111112e-07,
      "loss": 0.0098,
      "num_tokens": 44705207.0,
      "reward": 1.7390491962432861,
      "reward_std": 0.34051012992858887,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6994253993034363,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.944271445274353,
      "step": 196
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6603300539511265,
      "calib/avg_num_step_conf": 5.6953125,
      "calib/ece": 0.3408730158730157,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.19444444444444445,
      "calib/gap": 0.09362361155188825,
      "calib/mean_conf": 0.7966666666666667,
      "calib/mu_c": 0.8475652173913043,
      "calib/mu_w": 0.753941605839416,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.34059523809523795,
      "calib/std_conf": 0.16966119366231228,
      "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": 2876.0,
      "completions/max_terminated_length": 2876.0,
      "completions/mean_length": 518.34375,
      "completions/mean_terminated_length": 520.3765258789062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 214.0,
      "epoch": 0.21013333333333334,
      "grad_norm": 0.010996755212545395,
      "learning_rate": 8.333333333333334e-08,
      "loss": 0.0487,
      "num_tokens": 44942959.0,
      "reward": 1.5602569580078125,
      "reward_std": 0.5935218334197998,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.6435444951057434,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.9334204196929932,
      "step": 197
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6740728692257644,
      "calib/avg_num_step_conf": 5.43359375,
      "calib/ece": 0.19486055776892441,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.18326693227091634,
      "calib/gap": 0.1336389069616135,
      "calib/mean_conf": 0.7723904382470119,
      "calib/mu_c": 0.8288275862068966,
      "calib/mu_w": 0.695188679245283,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.19478087649402404,
      "calib/std_conf": 0.19325648932324366,
      "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": 2876.0,
      "completions/max_terminated_length": 2876.0,
      "completions/mean_length": 456.890625,
      "completions/mean_terminated_length": 462.3083190917969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 213.0,
      "epoch": 0.2112,
      "grad_norm": 0.0115453926846385,
      "learning_rate": 5.555555555555556e-08,
      "loss": 0.0046,
      "num_tokens": 45165307.0,
      "reward": 1.7558231353759766,
      "reward_std": 0.3816918134689331,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.7314152121543884,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.9325023293495178,
      "step": 198
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.6003032649374114,
      "calib/avg_num_step_conf": 5.578125,
      "calib/ece": 0.3161847389558233,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.22088353413654618,
      "calib/gap": 0.07058072009291538,
      "calib/mean_conf": 0.8013253012048194,
      "calib/mu_c": 0.8361904761904764,
      "calib/mu_w": 0.765609756097561,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.3057429718875502,
      "calib/std_conf": 0.18988542995302038,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2610.0,
      "completions/max_terminated_length": 2610.0,
      "completions/mean_length": 557.78515625,
      "completions/mean_terminated_length": 559.9725952148438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 204.0,
      "epoch": 0.21226666666666666,
      "grad_norm": 0.010695649310946465,
      "learning_rate": 2.777777777777778e-08,
      "loss": 0.0343,
      "num_tokens": 45412300.0,
      "reward": 1.615269422531128,
      "reward_std": 0.5218293070793152,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.6439590454101562,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.918681263923645,
      "step": 199
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6688959937361346,
      "calib/avg_num_step_conf": 5.4140625,
      "calib/ece": 0.2097647058823529,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.26666666666666666,
      "calib/gap": 0.10869307059898237,
      "calib/mean_conf": 0.8246666666666668,
      "calib/mu_c": 0.8660126582278482,
      "calib/mu_w": 0.7573195876288659,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.20741176470588232,
      "calib/std_conf": 0.1617835235638414,
      "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": 3055.0,
      "completions/max_terminated_length": 3055.0,
      "completions/mean_length": 504.14453125,
      "completions/mean_terminated_length": 504.14453125,
      "completions/min_length": 218.0,
      "completions/min_terminated_length": 218.0,
      "epoch": 0.21333333333333335,
      "grad_norm": 0.010254484601318836,
      "learning_rate": 0.0,
      "loss": 0.0446,
      "num_tokens": 45649409.0,
      "reward": 1.8421803712844849,
      "reward_std": 0.38195812702178955,
      "rewards/accuracy_reward_step": 0.6171875,
      "rewards/final_brier_reward_step": 0.7412852048873901,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.9399361610412598,
      "step": 200
    },
    {
      "epoch": 0.21333333333333335,
      "step": 200,
      "total_flos": 0.0,
      "train_loss": 0.02880778172169812,
      "train_runtime": 12568.5189,
      "train_samples_per_second": 4.074,
      "train_steps_per_second": 0.016
    }
  ],
  "logging_steps": 1,
  "max_steps": 200,
  "num_input_tokens_seen": 45649409,
  "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
}