{
  "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": 0.004028667230159044,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0324,
      "num_tokens": 264685.0,
      "reward": 0.05698804929852486,
      "reward_std": 0.11656394600868225,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.01655624993145466,
      "rewards/format_reward_step": 0.04296875,
      "rewards/stepwise_brier_reward": 0.031708456575870514,
      "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": 0.0062691220082342625,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.065,
      "num_tokens": 533467.0,
      "reward": 0.11318854987621307,
      "reward_std": 0.21675795316696167,
      "rewards/accuracy_reward_step": 0.03125,
      "rewards/final_brier_reward_step": 0.02965039201080799,
      "rewards/format_reward_step": 0.08984375,
      "rewards/stepwise_brier_reward": 0.05591631680727005,
      "step": 2
    },
    {
      "calib/answer_extract_rate": 0.0546875,
      "calib/auroc": 0.9444444444444445,
      "calib/avg_num_step_conf": 0.31640625,
      "calib/ece": 0.8260000000000001,
      "calib/final_conf_rate": 0.0390625,
      "calib/format_rate": 0.0390625,
      "calib/frac_conf_gt_0.9": 0.7,
      "calib/gap": 0.07111111111111112,
      "calib/mean_conf": 0.9260000000000002,
      "calib/mu_c": 0.99,
      "calib/mu_w": 0.9188888888888889,
      "calib/nonempty_final_conf_rate": 0.0390625,
      "calib/nonempty_reasoning_rate": 0.08203125,
      "calib/nonempty_step_conf_rate": 0.0703125,
      "calib/pce": 0.8260000000000001,
      "calib/std_conf": 0.08464041587799531,
      "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": 0.07421875,
      "completions/max_length": 3008.0,
      "completions/max_terminated_length": 3008.0,
      "completions/mean_length": 625.85546875,
      "completions/mean_terminated_length": 676.0294799804688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0032,
      "grad_norm": 0.003487977897748351,
      "learning_rate": 7.5e-07,
      "loss": 0.028,
      "num_tokens": 798942.0,
      "reward": 0.032370757311582565,
      "reward_std": 0.08481056988239288,
      "rewards/accuracy_reward_step": 0.00390625,
      "rewards/final_brier_reward_step": 0.00911562517285347,
      "rewards/format_reward_step": 0.0390625,
      "rewards/stepwise_brier_reward": 0.01880490593612194,
      "step": 3
    },
    {
      "calib/answer_extract_rate": 0.0546875,
      "calib/auroc": 0.7045454545454546,
      "calib/avg_num_step_conf": 0.1640625,
      "calib/ece": 0.7776923076923077,
      "calib/final_conf_rate": 0.05078125,
      "calib/format_rate": 0.0234375,
      "calib/frac_conf_gt_0.9": 0.8461538461538461,
      "calib/gap": 0.057272727272727364,
      "calib/mean_conf": 0.9315384615384615,
      "calib/mu_c": 0.98,
      "calib/mu_w": 0.9227272727272726,
      "calib/nonempty_final_conf_rate": 0.05078125,
      "calib/nonempty_reasoning_rate": 0.05859375,
      "calib/nonempty_step_conf_rate": 0.0390625,
      "calib/pce": 0.7776923076923077,
      "calib/std_conf": 0.09678255992339446,
      "calib/step_conf_rate": 0.0390625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2951.0,
      "completions/max_terminated_length": 2951.0,
      "completions/mean_length": 600.2421875,
      "completions/mean_terminated_length": 671.0131225585938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.004266666666666667,
      "grad_norm": 0.003240991849452257,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0088,
      "num_tokens": 1058772.0,
      "reward": 0.02906983345746994,
      "reward_std": 0.059641189873218536,
      "rewards/accuracy_reward_step": 0.0078125,
      "rewards/final_brier_reward_step": 0.006975390948355198,
      "rewards/format_reward_step": 0.0234375,
      "rewards/stepwise_brier_reward": 0.015553947538137436,
      "step": 4
    },
    {
      "calib/answer_extract_rate": 0.06640625,
      "calib/auroc": 0.6481481481481481,
      "calib/avg_num_step_conf": 0.25,
      "calib/ece": 0.5453333333333332,
      "calib/final_conf_rate": 0.046875,
      "calib/format_rate": 0.0234375,
      "calib/frac_conf_gt_0.9": 0.5833333333333334,
      "calib/gap": 0.20311111111111113,
      "calib/mean_conf": 0.753,
      "calib/mu_c": 0.9053333333333334,
      "calib/mu_w": 0.7022222222222223,
      "calib/nonempty_final_conf_rate": 0.046875,
      "calib/nonempty_reasoning_rate": 0.08203125,
      "calib/nonempty_step_conf_rate": 0.05078125,
      "calib/pce": 0.5241666666666666,
      "calib/std_conf": 0.358504765193807,
      "calib/step_conf_rate": 0.05078125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2984.0,
      "completions/max_terminated_length": 2984.0,
      "completions/mean_length": 682.78125,
      "completions/mean_terminated_length": 750.1802368164062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.005333333333333333,
      "grad_norm": 0.003930769860744476,
      "learning_rate": 1.25e-06,
      "loss": 0.0265,
      "num_tokens": 1340252.0,
      "reward": 0.03678886219859123,
      "reward_std": 0.10405462235212326,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.011540234088897705,
      "rewards/format_reward_step": 0.0234375,
      "rewards/stepwise_brier_reward": 0.018427718430757523,
      "step": 5
    },
    {
      "calib/answer_extract_rate": 0.11328125,
      "calib/auroc": 0.4166666666666667,
      "calib/avg_num_step_conf": 0.3359375,
      "calib/ece": 0.7371428571428571,
      "calib/final_conf_rate": 0.08203125,
      "calib/format_rate": 0.0625,
      "calib/frac_conf_gt_0.9": 0.6190476190476191,
      "calib/gap": 0.054444444444444406,
      "calib/mean_conf": 0.8800000000000002,
      "calib/mu_c": 0.9266666666666666,
      "calib/mu_w": 0.8722222222222222,
      "calib/nonempty_final_conf_rate": 0.08203125,
      "calib/nonempty_reasoning_rate": 0.125,
      "calib/nonempty_step_conf_rate": 0.078125,
      "calib/pce": 0.7371428571428571,
      "calib/std_conf": 0.1976769851960275,
      "calib/step_conf_rate": 0.078125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.109375,
      "completions/max_length": 3029.0,
      "completions/max_terminated_length": 3029.0,
      "completions/mean_length": 556.98046875,
      "completions/mean_terminated_length": 625.381591796875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.0064,
      "grad_norm": 0.007071846164762974,
      "learning_rate": 1.5e-06,
      "loss": -0.003,
      "num_tokens": 1588791.0,
      "reward": 0.060225263237953186,
      "reward_std": 0.1364026665687561,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.011326172389090061,
      "rewards/format_reward_step": 0.0625,
      "rewards/stepwise_brier_reward": 0.03426238149404526,
      "step": 6
    },
    {
      "calib/answer_extract_rate": 0.08984375,
      "calib/auroc": 0.4083333333333333,
      "calib/avg_num_step_conf": 0.29296875,
      "calib/ece": 0.7584210526315789,
      "calib/final_conf_rate": 0.07421875,
      "calib/format_rate": 0.0546875,
      "calib/frac_conf_gt_0.9": 0.8947368421052632,
      "calib/gap": -0.021166666666666556,
      "calib/mean_conf": 0.9542105263157894,
      "calib/mu_c": 0.9375,
      "calib/mu_w": 0.9586666666666666,
      "calib/nonempty_final_conf_rate": 0.07421875,
      "calib/nonempty_reasoning_rate": 0.09375,
      "calib/nonempty_step_conf_rate": 0.07421875,
      "calib/pce": 0.7510526315789473,
      "calib/std_conf": 0.02925672023092062,
      "calib/step_conf_rate": 0.07421875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08203125,
      "completions/max_length": 3040.0,
      "completions/max_terminated_length": 3040.0,
      "completions/mean_length": 784.6484375,
      "completions/mean_terminated_length": 854.7659301757812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.007466666666666667,
      "grad_norm": 0.004246986471116543,
      "learning_rate": 1.75e-06,
      "loss": 0.0156,
      "num_tokens": 1897085.0,
      "reward": 0.06449669599533081,
      "reward_std": 0.15668700635433197,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.018946874886751175,
      "rewards/format_reward_step": 0.0546875,
      "rewards/stepwise_brier_reward": 0.035914916545152664,
      "step": 7
    },
    {
      "calib/answer_extract_rate": 0.07421875,
      "calib/auroc": 0.6666666666666666,
      "calib/avg_num_step_conf": 0.203125,
      "calib/ece": 0.656923076923077,
      "calib/final_conf_rate": 0.05078125,
      "calib/format_rate": 0.0390625,
      "calib/frac_conf_gt_0.9": 0.7692307692307693,
      "calib/gap": 0.1026666666666668,
      "calib/mean_conf": 0.8876923076923077,
      "calib/mu_c": 0.9666666666666667,
      "calib/mu_w": 0.8639999999999999,
      "calib/nonempty_final_conf_rate": 0.05078125,
      "calib/nonempty_reasoning_rate": 0.078125,
      "calib/nonempty_step_conf_rate": 0.046875,
      "calib/pce": 0.656923076923077,
      "calib/std_conf": 0.2027940920909242,
      "calib/step_conf_rate": 0.046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0625,
      "completions/max_length": 3035.0,
      "completions/max_terminated_length": 3035.0,
      "completions/mean_length": 675.6015625,
      "completions/mean_terminated_length": 720.6417236328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.008533333333333334,
      "grad_norm": 0.004281523171812296,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0453,
      "num_tokens": 2176551.0,
      "reward": 0.04871947318315506,
      "reward_std": 0.10926631093025208,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.018850000575184822,
      "rewards/format_reward_step": 0.0390625,
      "rewards/stepwise_brier_reward": 0.027590394020080566,
      "step": 8
    },
    {
      "calib/answer_extract_rate": 0.08203125,
      "calib/auroc": 0.25757575757575757,
      "calib/avg_num_step_conf": 0.375,
      "calib/ece": 0.7464285714285711,
      "calib/final_conf_rate": 0.0546875,
      "calib/format_rate": 0.046875,
      "calib/frac_conf_gt_0.9": 0.9285714285714286,
      "calib/gap": -0.005151515151515129,
      "calib/mean_conf": 0.9607142857142857,
      "calib/mu_c": 0.9566666666666667,
      "calib/mu_w": 0.9618181818181818,
      "calib/nonempty_final_conf_rate": 0.0546875,
      "calib/nonempty_reasoning_rate": 0.1015625,
      "calib/nonempty_step_conf_rate": 0.0703125,
      "calib/pce": 0.7464285714285711,
      "calib/std_conf": 0.03127397039855662,
      "calib/step_conf_rate": 0.0703125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08203125,
      "completions/max_length": 3006.0,
      "completions/max_terminated_length": 3006.0,
      "completions/mean_length": 691.19140625,
      "completions/mean_terminated_length": 752.9573974609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.0096,
      "grad_norm": 0.003581782802939415,
      "learning_rate": 2.25e-06,
      "loss": 0.0397,
      "num_tokens": 2461032.0,
      "reward": 0.057361483573913574,
      "reward_std": 0.1387012004852295,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.014629686251282692,
      "rewards/format_reward_step": 0.046875,
      "rewards/stepwise_brier_reward": 0.027316249907016754,
      "step": 9
    },
    {
      "calib/answer_extract_rate": 0.1171875,
      "calib/auroc": 0.44565217391304346,
      "calib/avg_num_step_conf": 0.421875,
      "calib/ece": 0.8316,
      "calib/final_conf_rate": 0.09765625,
      "calib/format_rate": 0.078125,
      "calib/frac_conf_gt_0.9": 0.6,
      "calib/gap": -0.0691304347826085,
      "calib/mean_conf": 0.8835999999999999,
      "calib/mu_c": 0.8200000000000001,
      "calib/mu_w": 0.8891304347826086,
      "calib/nonempty_final_conf_rate": 0.09765625,
      "calib/nonempty_reasoning_rate": 0.140625,
      "calib/nonempty_step_conf_rate": 0.1015625,
      "calib/pce": 0.8176,
      "calib/std_conf": 0.15653446904755514,
      "calib/step_conf_rate": 0.1015625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.07421875,
      "completions/max_length": 2983.0,
      "completions/max_terminated_length": 2983.0,
      "completions/mean_length": 648.01953125,
      "completions/mean_terminated_length": 699.970458984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.010666666666666666,
      "grad_norm": 0.004958965349942446,
      "learning_rate": 2.5e-06,
      "loss": 0.0548,
      "num_tokens": 2733725.0,
      "reward": 0.06633052229881287,
      "reward_std": 0.14511729776859283,
      "rewards/accuracy_reward_step": 0.0078125,
      "rewards/final_brier_reward_step": 0.01895195245742798,
      "rewards/format_reward_step": 0.078125,
      "rewards/stepwise_brier_reward": 0.043245136737823486,
      "step": 10
    },
    {
      "calib/answer_extract_rate": 0.17578125,
      "calib/auroc": 0.4781818181818182,
      "calib/avg_num_step_conf": 0.8359375,
      "calib/ece": 0.6636666666666666,
      "calib/final_conf_rate": 0.140625,
      "calib/format_rate": 0.12109375,
      "calib/frac_conf_gt_0.9": 0.75,
      "calib/gap": -0.0985454545454546,
      "calib/mean_conf": 0.8858888888888888,
      "calib/mu_c": 0.8174545454545453,
      "calib/mu_w": 0.9159999999999999,
      "calib/nonempty_final_conf_rate": 0.140625,
      "calib/nonempty_reasoning_rate": 0.2109375,
      "calib/nonempty_step_conf_rate": 0.171875,
      "calib/pce": 0.6219999999999999,
      "calib/std_conf": 0.19406582184886795,
      "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": 0.078125,
      "completions/max_length": 3031.0,
      "completions/max_terminated_length": 3031.0,
      "completions/mean_length": 702.24609375,
      "completions/mean_terminated_length": 761.7584838867188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 7.0,
      "epoch": 0.011733333333333333,
      "grad_norm": 0.006088519003242254,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.0526,
      "num_tokens": 3017980.0,
      "reward": 0.1578739732503891,
      "reward_std": 0.238123819231987,
      "rewards/accuracy_reward_step": 0.04296875,
      "rewards/final_brier_reward_step": 0.04927154630422592,
      "rewards/format_reward_step": 0.12109375,
      "rewards/stepwise_brier_reward": 0.08222433924674988,
      "step": 11
    },
    {
      "calib/answer_extract_rate": 0.1796875,
      "calib/auroc": 0.561965811965812,
      "calib/avg_num_step_conf": 0.86328125,
      "calib/ece": 0.5247727272727273,
      "calib/final_conf_rate": 0.171875,
      "calib/format_rate": 0.1328125,
      "calib/frac_conf_gt_0.9": 0.75,
      "calib/gap": -0.020982905982905975,
      "calib/mean_conf": 0.8879545454545454,
      "calib/mu_c": 0.8755555555555556,
      "calib/mu_w": 0.8965384615384616,
      "calib/nonempty_final_conf_rate": 0.171875,
      "calib/nonempty_reasoning_rate": 0.2265625,
      "calib/nonempty_step_conf_rate": 0.18359375,
      "calib/pce": 0.5018181818181818,
      "calib/std_conf": 0.18735697024002246,
      "calib/step_conf_rate": 0.18359375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0625,
      "completions/max_length": 3035.0,
      "completions/max_terminated_length": 3035.0,
      "completions/mean_length": 537.953125,
      "completions/mean_terminated_length": 573.8167114257812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.0128,
      "grad_norm": 0.007944895885884762,
      "learning_rate": 3e-06,
      "loss": 0.053,
      "num_tokens": 3259872.0,
      "reward": 0.214485764503479,
      "reward_std": 0.3647361993789673,
      "rewards/accuracy_reward_step": 0.0703125,
      "rewards/final_brier_reward_step": 0.0732371062040329,
      "rewards/format_reward_step": 0.1328125,
      "rewards/stepwise_brier_reward": 0.09720595926046371,
      "step": 12
    },
    {
      "calib/answer_extract_rate": 0.21484375,
      "calib/auroc": 0.6019283746556474,
      "calib/avg_num_step_conf": 0.96875,
      "calib/ece": 0.6293181818840909,
      "calib/final_conf_rate": 0.171875,
      "calib/format_rate": 0.15234375,
      "calib/frac_conf_gt_0.9": 0.7272727272727273,
      "calib/gap": 0.03484848476060598,
      "calib/mean_conf": 0.8793181818840909,
      "calib/mu_c": 0.9054545454545454,
      "calib/mu_w": 0.8706060606939394,
      "calib/nonempty_final_conf_rate": 0.171875,
      "calib/nonempty_reasoning_rate": 0.26953125,
      "calib/nonempty_step_conf_rate": 0.21875,
      "calib/pce": 0.6293181818840909,
      "calib/std_conf": 0.18553850006954573,
      "calib/step_conf_rate": 0.21875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.06640625,
      "completions/max_length": 3011.0,
      "completions/max_terminated_length": 3011.0,
      "completions/mean_length": 621.42578125,
      "completions/mean_terminated_length": 665.6276245117188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.013866666666666666,
      "grad_norm": 0.00830151978880167,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.0941,
      "num_tokens": 3523549.0,
      "reward": 0.18398210406303406,
      "reward_std": 0.3511863350868225,
      "rewards/accuracy_reward_step": 0.04296875,
      "rewards/final_brier_reward_step": 0.06592421978712082,
      "rewards/format_reward_step": 0.15234375,
      "rewards/stepwise_brier_reward": 0.10750417411327362,
      "step": 13
    },
    {
      "calib/answer_extract_rate": 0.33203125,
      "calib/auroc": 0.4606643356643357,
      "calib/avg_num_step_conf": 1.28125,
      "calib/ece": 0.6291891891891892,
      "calib/final_conf_rate": 0.2890625,
      "calib/format_rate": 0.2265625,
      "calib/frac_conf_gt_0.9": 0.7567567567567568,
      "calib/gap": 0.018583916083916074,
      "calib/mean_conf": 0.9264864864864866,
      "calib/mu_c": 0.9395454545454546,
      "calib/mu_w": 0.9209615384615385,
      "calib/nonempty_final_conf_rate": 0.2890625,
      "calib/nonempty_reasoning_rate": 0.375,
      "calib/nonempty_step_conf_rate": 0.29296875,
      "calib/pce": 0.6291891891891892,
      "calib/std_conf": 0.1226336274198138,
      "calib/step_conf_rate": 0.29296875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0625,
      "completions/max_length": 2978.0,
      "completions/max_terminated_length": 2978.0,
      "completions/mean_length": 557.3046875,
      "completions/mean_terminated_length": 594.4583740234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.014933333333333333,
      "grad_norm": 0.011097606271505356,
      "learning_rate": 3.5e-06,
      "loss": 0.0855,
      "num_tokens": 3771619.0,
      "reward": 0.3060101270675659,
      "reward_std": 0.4477435350418091,
      "rewards/accuracy_reward_step": 0.0859375,
      "rewards/final_brier_reward_step": 0.09617148339748383,
      "rewards/format_reward_step": 0.2265625,
      "rewards/stepwise_brier_reward": 0.15911908447742462,
      "step": 14
    },
    {
      "calib/answer_extract_rate": 0.375,
      "calib/auroc": 0.5381578947368421,
      "calib/avg_num_step_conf": 1.6875,
      "calib/ece": 0.6374810126582278,
      "calib/final_conf_rate": 0.30859375,
      "calib/format_rate": 0.265625,
      "calib/frac_conf_gt_0.9": 0.810126582278481,
      "calib/gap": 0.08719385964912263,
      "calib/mean_conf": 0.8779873417721515,
      "calib/mu_c": 0.9442105263157894,
      "calib/mu_w": 0.8570166666666668,
      "calib/nonempty_final_conf_rate": 0.30859375,
      "calib/nonempty_reasoning_rate": 0.44140625,
      "calib/nonempty_step_conf_rate": 0.35546875,
      "calib/pce": 0.6374810126582278,
      "calib/std_conf": 0.21590378152719775,
      "calib/step_conf_rate": 0.35546875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2950.0,
      "completions/max_terminated_length": 2950.0,
      "completions/mean_length": 517.3828125,
      "completions/mean_terminated_length": 538.4146118164062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.016,
      "grad_norm": 0.011781086213886738,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.1582,
      "num_tokens": 4011949.0,
      "reward": 0.3139559328556061,
      "reward_std": 0.43584704399108887,
      "rewards/accuracy_reward_step": 0.07421875,
      "rewards/final_brier_reward_step": 0.10283312201499939,
      "rewards/format_reward_step": 0.265625,
      "rewards/stepwise_brier_reward": 0.17642810940742493,
      "step": 15
    },
    {
      "calib/answer_extract_rate": 0.48046875,
      "calib/auroc": 0.538180155464106,
      "calib/avg_num_step_conf": 2.01953125,
      "calib/ece": 0.6676388888888889,
      "calib/final_conf_rate": 0.421875,
      "calib/format_rate": 0.34375,
      "calib/frac_conf_gt_0.9": 0.75,
      "calib/gap": 0.019444444444444486,
      "calib/mean_conf": 0.9089351851851851,
      "calib/mu_c": 0.9235185185185186,
      "calib/mu_w": 0.9040740740740741,
      "calib/nonempty_final_conf_rate": 0.421875,
      "calib/nonempty_reasoning_rate": 0.546875,
      "calib/nonempty_step_conf_rate": 0.42578125,
      "calib/pce": 0.6632870370370371,
      "calib/std_conf": 0.15926923822095784,
      "calib/step_conf_rate": 0.42578125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.046875,
      "completions/max_length": 2987.0,
      "completions/max_terminated_length": 2987.0,
      "completions/mean_length": 520.3671875,
      "completions/mean_terminated_length": 545.958984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 7.0,
      "epoch": 0.017066666666666667,
      "grad_norm": 0.011466552503407001,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.1582,
      "num_tokens": 4254011.0,
      "reward": 0.4321386218070984,
      "reward_std": 0.5988593697547913,
      "rewards/accuracy_reward_step": 0.11328125,
      "rewards/final_brier_reward_step": 0.13069754838943481,
      "rewards/format_reward_step": 0.34375,
      "rewards/stepwise_brier_reward": 0.23066940903663635,
      "step": 16
    },
    {
      "calib/answer_extract_rate": 0.609375,
      "calib/auroc": 0.574002574002574,
      "calib/avg_num_step_conf": 2.94140625,
      "calib/ece": 0.641338028169014,
      "calib/final_conf_rate": 0.5546875,
      "calib/format_rate": 0.4765625,
      "calib/frac_conf_gt_0.9": 0.7746478873239436,
      "calib/gap": 0.05225482625482625,
      "calib/mean_conf": 0.9019014084507042,
      "calib/mu_c": 0.9405405405405405,
      "calib/mu_w": 0.8882857142857142,
      "calib/nonempty_final_conf_rate": 0.5546875,
      "calib/nonempty_reasoning_rate": 0.7109375,
      "calib/nonempty_step_conf_rate": 0.62109375,
      "calib/pce": 0.641338028169014,
      "calib/std_conf": 0.1767664692352698,
      "calib/step_conf_rate": 0.62109375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2369.0,
      "completions/max_terminated_length": 2369.0,
      "completions/mean_length": 397.80078125,
      "completions/mean_terminated_length": 405.72509765625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 24.0,
      "epoch": 0.018133333333333335,
      "grad_norm": 0.012450379319489002,
      "learning_rate": 4.25e-06,
      "loss": 0.1221,
      "num_tokens": 4459376.0,
      "reward": 0.5983771085739136,
      "reward_std": 0.715158224105835,
      "rewards/accuracy_reward_step": 0.15234375,
      "rewards/final_brier_reward_step": 0.19193553924560547,
      "rewards/format_reward_step": 0.4765625,
      "rewards/stepwise_brier_reward": 0.334385484457016,
      "step": 17
    },
    {
      "calib/answer_extract_rate": 0.625,
      "calib/auroc": 0.5469966629588432,
      "calib/avg_num_step_conf": 2.8203125,
      "calib/ece": 0.7067973856209149,
      "calib/final_conf_rate": 0.59765625,
      "calib/format_rate": 0.515625,
      "calib/frac_conf_gt_0.9": 0.7320261437908496,
      "calib/gap": 0.04706340378198015,
      "calib/mean_conf": 0.8963398692810457,
      "calib/mu_c": 0.9344827586206897,
      "calib/mu_w": 0.8874193548387096,
      "calib/nonempty_final_conf_rate": 0.59765625,
      "calib/nonempty_reasoning_rate": 0.7109375,
      "calib/nonempty_step_conf_rate": 0.6328125,
      "calib/pce": 0.7067973856209149,
      "calib/std_conf": 0.18081424995172207,
      "calib/step_conf_rate": 0.6328125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0234375,
      "completions/max_length": 2825.0,
      "completions/max_terminated_length": 2825.0,
      "completions/mean_length": 422.49609375,
      "completions/mean_terminated_length": 432.6360168457031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 23.0,
      "epoch": 0.0192,
      "grad_norm": 0.01302229706197977,
      "learning_rate": 4.5e-06,
      "loss": 0.1514,
      "num_tokens": 4678255.0,
      "reward": 0.5619417428970337,
      "reward_std": 0.5964739322662354,
      "rewards/accuracy_reward_step": 0.11328125,
      "rewards/final_brier_reward_step": 0.17842578887939453,
      "rewards/format_reward_step": 0.515625,
      "rewards/stepwise_brier_reward": 0.35840368270874023,
      "step": 18
    },
    {
      "calib/answer_extract_rate": 0.87890625,
      "calib/auroc": 0.5733901515151516,
      "calib/avg_num_step_conf": 4.49609375,
      "calib/ece": 0.7009908256880734,
      "calib/final_conf_rate": 0.8515625,
      "calib/format_rate": 0.7890625,
      "calib/frac_conf_gt_0.9": 0.7385321100917431,
      "calib/gap": 0.054607413419913486,
      "calib/mean_conf": 0.893651376146789,
      "calib/mu_c": 0.9377380952380954,
      "calib/mu_w": 0.8831306818181819,
      "calib/nonempty_final_conf_rate": 0.8515625,
      "calib/nonempty_reasoning_rate": 0.96484375,
      "calib/nonempty_step_conf_rate": 0.9140625,
      "calib/pce": 0.7009908256880734,
      "calib/std_conf": 0.17218055789331976,
      "calib/step_conf_rate": 0.9140625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 2349.0,
      "completions/max_terminated_length": 2349.0,
      "completions/mean_length": 319.83984375,
      "completions/mean_terminated_length": 322.3582763671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 21.0,
      "epoch": 0.020266666666666665,
      "grad_norm": 0.01097859162837267,
      "learning_rate": 4.75e-06,
      "loss": 0.1433,
      "num_tokens": 4864894.0,
      "reward": 0.8522599935531616,
      "reward_std": 0.5781415104866028,
      "rewards/accuracy_reward_step": 0.171875,
      "rewards/final_brier_reward_step": 0.2576978802680969,
      "rewards/format_reward_step": 0.7890625,
      "rewards/stepwise_brier_reward": 0.5419670343399048,
      "step": 19
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5334415584415584,
      "calib/avg_num_step_conf": 4.38671875,
      "calib/ece": 0.5914308860759493,
      "calib/final_conf_rate": 0.92578125,
      "calib/format_rate": 0.86328125,
      "calib/frac_conf_gt_0.9": 0.7383966244725738,
      "calib/gap": 0.019769626623376335,
      "calib/mean_conf": 0.9092367932489452,
      "calib/mu_c": 0.9225833766233763,
      "calib/mu_w": 0.90281375,
      "calib/nonempty_final_conf_rate": 0.92578125,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.9296875,
      "calib/pce": 0.587886582278481,
      "calib/std_conf": 0.1429276261568444,
      "calib/step_conf_rate": 0.9296875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2277.0,
      "completions/max_terminated_length": 2277.0,
      "completions/mean_length": 267.58203125,
      "completions/mean_terminated_length": 267.58203125,
      "completions/min_length": 6.0,
      "completions/min_terminated_length": 6.0,
      "epoch": 0.021333333333333333,
      "grad_norm": 0.010439296253025532,
      "learning_rate": 5e-06,
      "loss": 0.0379,
      "num_tokens": 5038267.0,
      "reward": 1.1433179378509521,
      "reward_std": 0.6862012147903442,
      "rewards/accuracy_reward_step": 0.30859375,
      "rewards/final_brier_reward_step": 0.3739124834537506,
      "rewards/format_reward_step": 0.86328125,
      "rewards/stepwise_brier_reward": 0.621234118938446,
      "step": 20
    },
    {
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.4708354777834963,
      "calib/avg_num_step_conf": 4.36328125,
      "calib/ece": 0.6171383125864454,
      "calib/final_conf_rate": 0.94140625,
      "calib/format_rate": 0.9140625,
      "calib/frac_conf_gt_0.9": 0.7219917012448133,
      "calib/gap": 0.025947646823354376,
      "calib/mean_conf": 0.895146611341632,
      "calib/mu_c": 0.9138805970149253,
      "calib/mu_w": 0.8879329501915709,
      "calib/nonempty_final_conf_rate": 0.94140625,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.6171383125864454,
      "calib/std_conf": 0.1630527939879922,
      "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": 2598.0,
      "completions/max_terminated_length": 2598.0,
      "completions/mean_length": 256.2734375,
      "completions/mean_terminated_length": 256.2734375,
      "completions/min_length": 50.0,
      "completions/min_terminated_length": 50.0,
      "epoch": 0.0224,
      "grad_norm": 0.008440430276095867,
      "learning_rate": 4.9722222222222224e-06,
      "loss": 0.0662,
      "num_tokens": 5206833.0,
      "reward": 1.1193917989730835,
      "reward_std": 0.6949143409729004,
      "rewards/accuracy_reward_step": 0.265625,
      "rewards/final_brier_reward_step": 0.37111151218414307,
      "rewards/format_reward_step": 0.9140625,
      "rewards/stepwise_brier_reward": 0.6845806241035461,
      "step": 21
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5182633916432521,
      "calib/avg_num_step_conf": 4.4609375,
      "calib/ece": 0.6695951219512196,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9375,
      "calib/frac_conf_gt_0.9": 0.7032520325203252,
      "calib/gap": 0.024381074956947413,
      "calib/mean_conf": 0.9094325203252033,
      "calib/mu_c": 0.9279661016949152,
      "calib/mu_w": 0.9035850267379678,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.6695951219512196,
      "calib/std_conf": 0.13767820587632146,
      "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": 933.0,
      "completions/max_terminated_length": 933.0,
      "completions/mean_length": 249.48828125,
      "completions/mean_terminated_length": 249.48828125,
      "completions/min_length": 66.0,
      "completions/min_terminated_length": 66.0,
      "epoch": 0.023466666666666667,
      "grad_norm": 0.008359194733202457,
      "learning_rate": 4.944444444444445e-06,
      "loss": 0.0134,
      "num_tokens": 5372518.0,
      "reward": 1.0780291557312012,
      "reward_std": 0.6363198161125183,
      "rewards/accuracy_reward_step": 0.23828125,
      "rewards/final_brier_reward_step": 0.3360249698162079,
      "rewards/format_reward_step": 0.9375,
      "rewards/stepwise_brier_reward": 0.6714040040969849,
      "step": 22
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5079443079443079,
      "calib/avg_num_step_conf": 4.3125,
      "calib/ece": 0.6484940239043825,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.6733067729083665,
      "calib/gap": -0.0009551187551186668,
      "calib/mean_conf": 0.907219123505976,
      "calib/mu_c": 0.9065151515151516,
      "calib/mu_w": 0.9074702702702703,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.6463824701195219,
      "calib/std_conf": 0.11917633906801327,
      "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": 2924.0,
      "completions/max_terminated_length": 2924.0,
      "completions/mean_length": 253.38671875,
      "completions/mean_terminated_length": 253.38671875,
      "completions/min_length": 59.0,
      "completions/min_terminated_length": 59.0,
      "epoch": 0.024533333333333334,
      "grad_norm": 0.008912438526749611,
      "learning_rate": 4.9166666666666665e-06,
      "loss": 0.0577,
      "num_tokens": 5541321.0,
      "reward": 1.1239862442016602,
      "reward_std": 0.635131299495697,
      "rewards/accuracy_reward_step": 0.2578125,
      "rewards/final_brier_reward_step": 0.3617061376571655,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.6733016967773438,
      "step": 23
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.41240196078431374,
      "calib/avg_num_step_conf": 4.4140625,
      "calib/ece": 0.7071474103585658,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.6852589641434262,
      "calib/gap": -0.056395980392156986,
      "calib/mean_conf": 0.8924860557768924,
      "calib/mu_c": 0.8475490196078431,
      "calib/mu_w": 0.9039450000000001,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.6982231075697212,
      "calib/std_conf": 0.15955684022857175,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 731.0,
      "completions/max_terminated_length": 731.0,
      "completions/mean_length": 220.98828125,
      "completions/mean_terminated_length": 222.7283477783203,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 33.0,
      "epoch": 0.0256,
      "grad_norm": 0.009140349924564362,
      "learning_rate": 4.888888888888889e-06,
      "loss": -0.0009,
      "num_tokens": 5702406.0,
      "reward": 1.009996771812439,
      "reward_std": 0.4800751805305481,
      "rewards/accuracy_reward_step": 0.19921875,
      "rewards/final_brier_reward_step": 0.29869914054870605,
      "rewards/format_reward_step": 0.94921875,
      "rewards/stepwise_brier_reward": 0.6475378274917603,
      "step": 24
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5576576576576576,
      "calib/avg_num_step_conf": 4.4765625,
      "calib/ece": 0.630725806451613,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.6088709677419355,
      "calib/gap": 0.03967910767910765,
      "calib/mean_conf": 0.880241935483871,
      "calib/mu_c": 0.9098412698412699,
      "calib/mu_w": 0.8701621621621622,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.628467741935484,
      "calib/std_conf": 0.17615443512897638,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2016.0,
      "completions/max_terminated_length": 2016.0,
      "completions/mean_length": 238.125,
      "completions/mean_terminated_length": 238.125,
      "completions/min_length": 52.0,
      "completions/min_terminated_length": 52.0,
      "epoch": 0.02666666666666667,
      "grad_norm": 0.007780835032463074,
      "learning_rate": 4.861111111111111e-06,
      "loss": -0.0058,
      "num_tokens": 5866590.0,
      "reward": 1.1299071311950684,
      "reward_std": 0.6295166015625,
      "rewards/accuracy_reward_step": 0.25,
      "rewards/final_brier_reward_step": 0.3832150101661682,
      "rewards/format_reward_step": 0.953125,
      "rewards/stepwise_brier_reward": 0.7301633358001709,
      "step": 25
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.4824561403508772,
      "calib/avg_num_step_conf": 4.44921875,
      "calib/ece": 0.6230434782608696,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.5177865612648221,
      "calib/gap": -0.0037889366272826352,
      "calib/mean_conf": 0.8408300395256917,
      "calib/mu_c": 0.8378947368421052,
      "calib/mu_w": 0.8416836734693879,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.6192885375494072,
      "calib/std_conf": 0.1898525087166518,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 771.0,
      "completions/max_terminated_length": 771.0,
      "completions/mean_length": 235.75390625,
      "completions/mean_terminated_length": 236.67845153808594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 83.0,
      "epoch": 0.027733333333333332,
      "grad_norm": 0.008240653201937675,
      "learning_rate": 4.833333333333333e-06,
      "loss": -0.0121,
      "num_tokens": 6032183.0,
      "reward": 1.1215285062789917,
      "reward_std": 0.5799918174743652,
      "rewards/accuracy_reward_step": 0.22265625,
      "rewards/final_brier_reward_step": 0.4034687578678131,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.7779577970504761,
      "step": 26
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.4696751249519415,
      "calib/avg_num_step_conf": 4.4296875,
      "calib/ece": 0.6228627450980393,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.4470588235294118,
      "calib/gap": -0.025539215686274486,
      "calib/mean_conf": 0.8169019607843138,
      "calib/mu_c": 0.7964705882352942,
      "calib/mu_w": 0.8220098039215686,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.6198823529411766,
      "calib/std_conf": 0.20629744895935023,
      "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": 709.0,
      "completions/max_terminated_length": 709.0,
      "completions/mean_length": 219.49609375,
      "completions/mean_terminated_length": 220.35687255859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 48.0,
      "epoch": 0.0288,
      "grad_norm": 0.009285224601626396,
      "learning_rate": 4.805555555555556e-06,
      "loss": -0.0067,
      "num_tokens": 6193590.0,
      "reward": 1.08359694480896,
      "reward_std": 0.5318938493728638,
      "rewards/accuracy_reward_step": 0.19921875,
      "rewards/final_brier_reward_step": 0.40126872062683105,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.7768691778182983,
      "step": 27
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.42311778992106863,
      "calib/avg_num_step_conf": 4.5625,
      "calib/ece": 0.5322352941176471,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.3607843137254902,
      "calib/gap": -0.05540300546448107,
      "calib/mean_conf": 0.7651764705882353,
      "calib/mu_c": 0.7254166666666667,
      "calib/mu_w": 0.7808196721311478,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.5075294117647059,
      "calib/std_conf": 0.25032972028532985,
      "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": 723.0,
      "completions/max_terminated_length": 723.0,
      "completions/mean_length": 238.32421875,
      "completions/mean_terminated_length": 239.2588348388672,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 66.0,
      "epoch": 0.029866666666666666,
      "grad_norm": 0.008911133743822575,
      "learning_rate": 4.777777777777778e-06,
      "loss": 0.0103,
      "num_tokens": 6361545.0,
      "reward": 1.2308874130249023,
      "reward_std": 0.47560015320777893,
      "rewards/accuracy_reward_step": 0.28125,
      "rewards/final_brier_reward_step": 0.4669402241706848,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8081717491149902,
      "step": 28
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.49419869470630895,
      "calib/avg_num_step_conf": 4.83203125,
      "calib/ece": 0.4927272727272727,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.2766798418972332,
      "calib/gap": 0.028398295866570056,
      "calib/mean_conf": 0.7073517786561265,
      "calib/mu_c": 0.7294642857142856,
      "calib/mu_w": 0.7010659898477155,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.48936758893280624,
      "calib/std_conf": 0.26508202691782046,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 731.0,
      "completions/max_terminated_length": 731.0,
      "completions/mean_length": 263.83203125,
      "completions/mean_terminated_length": 264.8666687011719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 46.0,
      "epoch": 0.030933333333333334,
      "grad_norm": 0.00778714194893837,
      "learning_rate": 4.75e-06,
      "loss": 0.0313,
      "num_tokens": 6536214.0,
      "reward": 1.1755021810531616,
      "reward_std": 0.5025407075881958,
      "rewards/accuracy_reward_step": 0.22265625,
      "rewards/final_brier_reward_step": 0.5246660113334656,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8726552128791809,
      "step": 29
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5102362914862915,
      "calib/avg_num_step_conf": 4.7734375,
      "calib/ece": 0.4123228346456692,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.16929133858267717,
      "calib/gap": 0.019550865800865735,
      "calib/mean_conf": 0.6327952755905513,
      "calib/mu_c": 0.6480357142857143,
      "calib/mu_w": 0.6284848484848485,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.4123228346456692,
      "calib/std_conf": 0.2693000211567069,
      "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": 1441.0,
      "completions/max_terminated_length": 1441.0,
      "completions/mean_length": 262.63671875,
      "completions/mean_terminated_length": 263.66668701171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 51.0,
      "epoch": 0.032,
      "grad_norm": 0.008011657744646072,
      "learning_rate": 4.722222222222222e-06,
      "loss": -0.0092,
      "num_tokens": 6710433.0,
      "reward": 1.1868528127670288,
      "reward_std": 0.4904547929763794,
      "rewards/accuracy_reward_step": 0.21875,
      "rewards/final_brier_reward_step": 0.5869464874267578,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8792147040367126,
      "step": 30
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.4901009316770187,
      "calib/avg_num_step_conf": 4.9296875,
      "calib/ece": 0.2816929133858268,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.1220472440944882,
      "calib/gap": -0.007312111801242249,
      "calib/mean_conf": 0.5435826771653544,
      "calib/mu_c": 0.5382857142857141,
      "calib/mu_w": 0.5455978260869564,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.27484251968503937,
      "calib/std_conf": 0.2716386535917086,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 827.0,
      "completions/max_terminated_length": 827.0,
      "completions/mean_length": 262.58984375,
      "completions/mean_terminated_length": 263.61962890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 65.0,
      "epoch": 0.03306666666666667,
      "grad_norm": 0.007873645052313805,
      "learning_rate": 4.694444444444445e-06,
      "loss": 0.0047,
      "num_tokens": 6883568.0,
      "reward": 1.2863229513168335,
      "reward_std": 0.4989652633666992,
      "rewards/accuracy_reward_step": 0.2734375,
      "rewards/final_brier_reward_step": 0.6459980010986328,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8821063041687012,
      "step": 31
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5644962672208113,
      "calib/avg_num_step_conf": 4.63671875,
      "calib/ece": 0.19787401574803148,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.03937007874015748,
      "calib/gap": 0.05162472100361748,
      "calib/mean_conf": 0.45322834645669285,
      "calib/mu_c": 0.4904225352112677,
      "calib/mu_w": 0.43879781420765024,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.1857874015748031,
      "calib/std_conf": 0.24942258322623884,
      "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": 586.0,
      "completions/max_terminated_length": 586.0,
      "completions/mean_length": 250.46875,
      "completions/mean_terminated_length": 251.45098876953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 54.0,
      "epoch": 0.034133333333333335,
      "grad_norm": 0.008328165858983994,
      "learning_rate": 4.666666666666667e-06,
      "loss": 0.0275,
      "num_tokens": 7054392.0,
      "reward": 1.3090533018112183,
      "reward_std": 0.4853959381580353,
      "rewards/accuracy_reward_step": 0.27734375,
      "rewards/final_brier_reward_step": 0.7131855487823486,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8980278968811035,
      "step": 32
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5306075833714025,
      "calib/avg_num_step_conf": 4.76171875,
      "calib/ece": 0.22879921259842517,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.047244094488188976,
      "calib/gap": 0.01698446779351298,
      "calib/mean_conf": 0.39860236220472445,
      "calib/mu_c": 0.4119090909090909,
      "calib/mu_w": 0.3949246231155779,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2054330708661417,
      "calib/std_conf": 0.23329570332792327,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2223.0,
      "completions/max_terminated_length": 2223.0,
      "completions/mean_length": 268.8828125,
      "completions/mean_terminated_length": 269.937255859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 68.0,
      "epoch": 0.0352,
      "grad_norm": 0.00786682404577732,
      "learning_rate": 4.638888888888889e-06,
      "loss": 0.0603,
      "num_tokens": 7230098.0,
      "reward": 1.226273775100708,
      "reward_std": 0.38770055770874023,
      "rewards/accuracy_reward_step": 0.21484375,
      "rewards/final_brier_reward_step": 0.742691695690155,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.888965904712677,
      "step": 33
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.528998865570051,
      "calib/avg_num_step_conf": 4.25390625,
      "calib/ece": 0.13543307086614173,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.011811023622047244,
      "calib/gap": 0.013845717526942747,
      "calib/mean_conf": 0.3133070866141732,
      "calib/mu_c": 0.3226829268292683,
      "calib/mu_w": 0.30883720930232555,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.06295275590551182,
      "calib/std_conf": 0.19673738488451353,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1764.0,
      "completions/max_terminated_length": 1764.0,
      "completions/mean_length": 226.1796875,
      "completions/mean_terminated_length": 226.1796875,
      "completions/min_length": 60.0,
      "completions/min_terminated_length": 60.0,
      "epoch": 0.03626666666666667,
      "grad_norm": 0.009125818498432636,
      "learning_rate": 4.611111111111112e-06,
      "loss": -0.0259,
      "num_tokens": 7393112.0,
      "reward": 1.3747636079788208,
      "reward_std": 0.46768075227737427,
      "rewards/accuracy_reward_step": 0.3203125,
      "rewards/final_brier_reward_step": 0.7389289140701294,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8616880774497986,
      "step": 34
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5766674673712868,
      "calib/avg_num_step_conf": 4.640625,
      "calib/ece": 0.10104000000000002,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.008,
      "calib/gap": 0.0504812234766594,
      "calib/mean_conf": 0.27664,
      "calib/mu_c": 0.3131884057971014,
      "calib/mu_w": 0.262707182320442,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.05084000000000001,
      "calib/std_conf": 0.16791399703419607,
      "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": 2303.0,
      "completions/max_terminated_length": 2303.0,
      "completions/mean_length": 276.375,
      "completions/mean_terminated_length": 277.4588317871094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 53.0,
      "epoch": 0.037333333333333336,
      "grad_norm": 0.009144150651991367,
      "learning_rate": 4.583333333333333e-06,
      "loss": 0.0558,
      "num_tokens": 7573120.0,
      "reward": 1.307545781135559,
      "reward_std": 0.4622858166694641,
      "rewards/accuracy_reward_step": 0.26953125,
      "rewards/final_brier_reward_step": 0.7735890746116638,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.886281430721283,
      "step": 35
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5699076923076922,
      "calib/avg_num_step_conf": 4.6484375,
      "calib/ece": 0.2809411764705882,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0196078431372549,
      "calib/gap": 0.05214769230769234,
      "calib/mean_conf": 0.2372549019607843,
      "calib/mu_c": 0.26384,
      "calib/mu_w": 0.21169230769230768,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.014000000000000002,
      "calib/std_conf": 0.18865133760949887,
      "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": 885.0,
      "completions/max_terminated_length": 885.0,
      "completions/mean_length": 254.85546875,
      "completions/mean_terminated_length": 255.85491943359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 47.0,
      "epoch": 0.0384,
      "grad_norm": 0.00828314758837223,
      "learning_rate": 4.555555555555556e-06,
      "loss": -0.034,
      "num_tokens": 7741075.0,
      "reward": 1.595306158065796,
      "reward_std": 0.5270878076553345,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.6725422143936157,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.7946195602416992,
      "step": 36
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5662118780096309,
      "calib/avg_num_step_conf": 4.6484375,
      "calib/ece": 0.17750980392156862,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.01742558003793962,
      "calib/mean_conf": 0.1497843137254902,
      "calib/mu_c": 0.16194805194805198,
      "calib/mu_w": 0.14452247191011236,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.01266666666666667,
      "calib/std_conf": 0.11605463786349544,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2781.0,
      "completions/max_terminated_length": 2781.0,
      "completions/mean_length": 273.34375,
      "completions/mean_terminated_length": 273.34375,
      "completions/min_length": 55.0,
      "completions/min_terminated_length": 55.0,
      "epoch": 0.039466666666666664,
      "grad_norm": 0.008583128452301025,
      "learning_rate": 4.527777777777778e-06,
      "loss": 0.0637,
      "num_tokens": 7918147.0,
      "reward": 1.349078893661499,
      "reward_std": 0.3283109664916992,
      "rewards/accuracy_reward_step": 0.30078125,
      "rewards/final_brier_reward_step": 0.7569705843925476,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8424696922302246,
      "step": 37
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.48310152086312236,
      "calib/avg_num_step_conf": 4.765625,
      "calib/ece": 0.2688039215686274,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.008212985831275166,
      "calib/mean_conf": 0.13786274509803922,
      "calib/mu_c": 0.1328061224489796,
      "calib/mu_w": 0.14101910828025477,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.011176470588235291,
      "calib/std_conf": 0.10334522757753582,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2274.0,
      "completions/max_terminated_length": 2274.0,
      "completions/mean_length": 277.33203125,
      "completions/mean_terminated_length": 277.33203125,
      "completions/min_length": 55.0,
      "completions/min_terminated_length": 55.0,
      "epoch": 0.04053333333333333,
      "grad_norm": 0.00880447681993246,
      "learning_rate": 4.5e-06,
      "loss": 0.0176,
      "num_tokens": 8096032.0,
      "reward": 1.447131872177124,
      "reward_std": 0.3899373412132263,
      "rewards/accuracy_reward_step": 0.3828125,
      "rewards/final_brier_reward_step": 0.6853905320167542,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8140745162963867,
      "step": 38
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5303177376561186,
      "calib/avg_num_step_conf": 4.8203125,
      "calib/ece": 0.2983858267716535,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0018785996246683534,
      "calib/mean_conf": 0.10381889763779528,
      "calib/mu_c": 0.10495049504950495,
      "calib/mu_w": 0.1030718954248366,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.002283464566929135,
      "calib/std_conf": 0.08422400926913581,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1940.0,
      "completions/max_terminated_length": 1940.0,
      "completions/mean_length": 292.91796875,
      "completions/mean_terminated_length": 292.91796875,
      "completions/min_length": 73.0,
      "completions/min_terminated_length": 73.0,
      "epoch": 0.0416,
      "grad_norm": 0.008755859918892384,
      "learning_rate": 4.472222222222223e-06,
      "loss": 0.0457,
      "num_tokens": 8277107.0,
      "reward": 1.4554944038391113,
      "reward_std": 0.3366001844406128,
      "rewards/accuracy_reward_step": 0.39453125,
      "rewards/final_brier_reward_step": 0.6588300466537476,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8193971514701843,
      "step": 39
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4824913689847108,
      "calib/avg_num_step_conf": 5.2109375,
      "calib/ece": 0.22895669291338586,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0004780525611216696,
      "calib/mean_conf": 0.10702755905511811,
      "calib/mu_c": 0.10734939759036145,
      "calib/mu_w": 0.10687134502923978,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.004606299212598427,
      "calib/std_conf": 0.07688781087383696,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2009.0,
      "completions/max_terminated_length": 2009.0,
      "completions/mean_length": 335.40234375,
      "completions/mean_terminated_length": 335.40234375,
      "completions/min_length": 79.0,
      "completions/min_terminated_length": 79.0,
      "epoch": 0.042666666666666665,
      "grad_norm": 0.00742807611823082,
      "learning_rate": 4.444444444444444e-06,
      "loss": 0.0491,
      "num_tokens": 8469730.0,
      "reward": 1.3729383945465088,
      "reward_std": 0.426654577255249,
      "rewards/accuracy_reward_step": 0.32421875,
      "rewards/final_brier_reward_step": 0.7164600491523743,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.853418231010437,
      "step": 40
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5167355905918842,
      "calib/avg_num_step_conf": 5.15234375,
      "calib/ece": 0.5011111111111112,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.006202636340139567,
      "calib/mean_conf": 0.07825396825396826,
      "calib/mu_c": 0.08086301369863014,
      "calib/mu_w": 0.07466037735849057,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05007707337598469,
      "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": 2171.0,
      "completions/max_terminated_length": 2171.0,
      "completions/mean_length": 297.5078125,
      "completions/mean_terminated_length": 298.6745300292969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 65.0,
      "epoch": 0.04373333333333333,
      "grad_norm": 0.008484176360070705,
      "learning_rate": 4.416666666666667e-06,
      "loss": 0.0113,
      "num_tokens": 8653140.0,
      "reward": 1.6562676429748535,
      "reward_std": 0.35466429591178894,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.497800350189209,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.7132081985473633,
      "step": 41
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.46812698412698417,
      "calib/avg_num_step_conf": 5.00390625,
      "calib/ece": 0.3507098039215686,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.006761904761904761,
      "calib/mean_conf": 0.06834901960784315,
      "calib/mu_c": 0.06437142857142858,
      "calib/mu_w": 0.07113333333333334,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0036470588235294117,
      "calib/std_conf": 0.045180506696066845,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 595.0,
      "completions/max_terminated_length": 595.0,
      "completions/mean_length": 255.63671875,
      "completions/mean_terminated_length": 256.63922119140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 69.0,
      "epoch": 0.0448,
      "grad_norm": 0.008709295652806759,
      "learning_rate": 4.388888888888889e-06,
      "loss": 0.0217,
      "num_tokens": 8822951.0,
      "reward": 1.4708423614501953,
      "reward_std": 0.35339561104774475,
      "rewards/accuracy_reward_step": 0.41015625,
      "rewards/final_brier_reward_step": 0.6320555210113525,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.7981888651847839,
      "step": 42
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5630917754811561,
      "calib/avg_num_step_conf": 4.66015625,
      "calib/ece": 0.37861328125,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.010050436289374347,
      "calib/mean_conf": 0.06279296875000001,
      "calib/mu_c": 0.06840707964601771,
      "calib/mu_w": 0.05835664335664336,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.036447630846346155,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 778.0,
      "completions/max_terminated_length": 778.0,
      "completions/mean_length": 287.140625,
      "completions/mean_terminated_length": 288.2666931152344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 96.0,
      "epoch": 0.04586666666666667,
      "grad_norm": 0.008184474892914295,
      "learning_rate": 4.361111111111112e-06,
      "loss": -0.0489,
      "num_tokens": 9001683.0,
      "reward": 1.5127179622650146,
      "reward_std": 0.3781050741672516,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.6137130260467529,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.7887214422225952,
      "step": 43
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5126731893098693,
      "calib/avg_num_step_conf": 5.18359375,
      "calib/ece": 0.31243307086614175,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0025468514019305194,
      "calib/mean_conf": 0.06945669291338583,
      "calib/mu_c": 0.07103092783505155,
      "calib/mu_w": 0.06848407643312103,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.035622640749179,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1984.0,
      "completions/max_terminated_length": 1984.0,
      "completions/mean_length": 322.21875,
      "completions/mean_terminated_length": 322.21875,
      "completions/min_length": 90.0,
      "completions/min_terminated_length": 90.0,
      "epoch": 0.046933333333333334,
      "grad_norm": 0.007914330810308456,
      "learning_rate": 4.333333333333334e-06,
      "loss": 0.0373,
      "num_tokens": 9190491.0,
      "reward": 1.438327431678772,
      "reward_std": 0.33558836579322815,
      "rewards/accuracy_reward_step": 0.37890625,
      "rewards/final_brier_reward_step": 0.6610637903213501,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8344335556030273,
      "step": 44
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5676056338028168,
      "calib/avg_num_step_conf": 5.03125,
      "calib/ece": 0.37748809523809523,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0075166453265044925,
      "calib/mean_conf": 0.06140079365079366,
      "calib/mu_c": 0.06563636363636365,
      "calib/mu_w": 0.05811971830985915,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0011904761904761904,
      "calib/std_conf": 0.039779876455296966,
      "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": 2649.0,
      "completions/max_terminated_length": 2649.0,
      "completions/mean_length": 334.42578125,
      "completions/mean_terminated_length": 334.42578125,
      "completions/min_length": 52.0,
      "completions/min_terminated_length": 52.0,
      "epoch": 0.048,
      "grad_norm": 0.007068061735481024,
      "learning_rate": 4.305555555555556e-06,
      "loss": 0.0819,
      "num_tokens": 9381152.0,
      "reward": 1.4855284690856934,
      "reward_std": 0.40476733446121216,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.605824887752533,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.7894140481948853,
      "step": 45
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5422896465436347,
      "calib/avg_num_step_conf": 4.9765625,
      "calib/ece": 0.3928503937007874,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0007300594307163011,
      "calib/mean_conf": 0.06407874015748032,
      "calib/mu_c": 0.06447826086956522,
      "calib/mu_w": 0.06374820143884892,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0020866141732283467,
      "calib/std_conf": 0.03898012946234255,
      "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": 2643.0,
      "completions/max_terminated_length": 2643.0,
      "completions/mean_length": 332.79296875,
      "completions/mean_terminated_length": 332.79296875,
      "completions/min_length": 72.0,
      "completions/min_terminated_length": 72.0,
      "epoch": 0.04906666666666667,
      "grad_norm": 0.008498773910105228,
      "learning_rate": 4.277777777777778e-06,
      "loss": 0.0054,
      "num_tokens": 9571115.0,
      "reward": 1.5049870014190674,
      "reward_std": 0.3561868667602539,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.5914247035980225,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.7566481828689575,
      "step": 46
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5271616541353383,
      "calib/avg_num_step_conf": 5.14453125,
      "calib/ece": 0.40583003952569163,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0026613408521303644,
      "calib/mean_conf": 0.07005928853754942,
      "calib/mu_c": 0.07145833333333336,
      "calib/mu_w": 0.068796992481203,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0007905138339920949,
      "calib/std_conf": 0.03180362861639707,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2442.0,
      "completions/max_terminated_length": 2442.0,
      "completions/mean_length": 335.45703125,
      "completions/mean_terminated_length": 335.45703125,
      "completions/min_length": 122.0,
      "completions/min_terminated_length": 122.0,
      "epoch": 0.050133333333333335,
      "grad_norm": 0.00726919062435627,
      "learning_rate": 4.25e-06,
      "loss": 0.0374,
      "num_tokens": 9762968.0,
      "reward": 1.5351732969284058,
      "reward_std": 0.3819875419139862,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5806730389595032,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.770957350730896,
      "step": 47
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4951826575672421,
      "calib/avg_num_step_conf": 4.25,
      "calib/ece": 0.301699604743083,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0027110932691021133,
      "calib/mean_conf": 0.07553359683794467,
      "calib/mu_c": 0.07382978723404256,
      "calib/mu_w": 0.07654088050314467,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0028458498023715413,
      "calib/std_conf": 0.054097736336986196,
      "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": 1145.0,
      "completions/max_terminated_length": 1145.0,
      "completions/mean_length": 292.6640625,
      "completions/mean_terminated_length": 292.6640625,
      "completions/min_length": 64.0,
      "completions/min_terminated_length": 64.0,
      "epoch": 0.0512,
      "grad_norm": 0.008678781799972057,
      "learning_rate": 4.222222222222223e-06,
      "loss": 0.0135,
      "num_tokens": 9941578.0,
      "reward": 1.407670497894287,
      "reward_std": 0.46994927525520325,
      "rewards/accuracy_reward_step": 0.3671875,
      "rewards/final_brier_reward_step": 0.6629145741462708,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.7958927154541016,
      "step": 48
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5761041258608707,
      "calib/avg_num_step_conf": 5.01953125,
      "calib/ece": 0.38901031746031745,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009242016806722694,
      "calib/mean_conf": 0.08487857142857144,
      "calib/mu_c": 0.08975630252100841,
      "calib/mu_w": 0.08051428571428572,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0008333333333333333,
      "calib/std_conf": 0.03656608739208411,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2783.0,
      "completions/max_terminated_length": 2783.0,
      "completions/mean_length": 340.65625,
      "completions/mean_terminated_length": 341.9921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 81.0,
      "epoch": 0.05226666666666667,
      "grad_norm": 0.007961131632328033,
      "learning_rate": 4.194444444444445e-06,
      "loss": 0.0394,
      "num_tokens": 10133322.0,
      "reward": 1.5388092994689941,
      "reward_std": 0.31433722376823425,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5945686101913452,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8028562068939209,
      "step": 49
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5149862911266201,
      "calib/avg_num_step_conf": 5.05078125,
      "calib/ece": 0.4544192913385827,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0020358300099700827,
      "calib/mean_conf": 0.08101377952755906,
      "calib/mu_c": 0.08195955882352941,
      "calib/mu_w": 0.07992372881355933,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.03843053871841524,
      "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": 2817.0,
      "completions/max_terminated_length": 2817.0,
      "completions/mean_length": 371.90234375,
      "completions/mean_terminated_length": 371.90234375,
      "completions/min_length": 91.0,
      "completions/min_terminated_length": 91.0,
      "epoch": 0.05333333333333334,
      "grad_norm": 0.0076715680770576,
      "learning_rate": 4.166666666666667e-06,
      "loss": 0.0295,
      "num_tokens": 10333889.0,
      "reward": 1.6202490329742432,
      "reward_std": 0.4135303497314453,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.5400422215461731,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.7690788507461548,
      "step": 50
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6238005050505051,
      "calib/avg_num_step_conf": 5.1796875,
      "calib/ece": 0.43642857142857144,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.014520454545454545,
      "calib/mean_conf": 0.08738095238095239,
      "calib/mu_c": 0.09429545454545454,
      "calib/mu_w": 0.079775,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04108138595652347,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2928.0,
      "completions/max_terminated_length": 2928.0,
      "completions/mean_length": 363.79296875,
      "completions/mean_terminated_length": 365.2196350097656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 80.0,
      "epoch": 0.0544,
      "grad_norm": 0.007592128124088049,
      "learning_rate": 4.138888888888889e-06,
      "loss": 0.0402,
      "num_tokens": 10536316.0,
      "reward": 1.5992858409881592,
      "reward_std": 0.318310022354126,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.5568147301673889,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.7543909549713135,
      "step": 51
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6427639305754926,
      "calib/avg_num_step_conf": 4.84765625,
      "calib/ece": 0.5221078431372549,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.020756394362521205,
      "calib/mean_conf": 0.0975,
      "calib/mu_c": 0.10539556962025319,
      "calib/mu_w": 0.08463917525773199,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04598433237016073,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2606.0,
      "completions/max_terminated_length": 2606.0,
      "completions/mean_length": 330.4140625,
      "completions/mean_terminated_length": 330.4140625,
      "completions/min_length": 94.0,
      "completions/min_terminated_length": 94.0,
      "epoch": 0.055466666666666664,
      "grad_norm": 0.008600694127380848,
      "learning_rate": 4.111111111111111e-06,
      "loss": 0.013,
      "num_tokens": 10728854.0,
      "reward": 1.7439744472503662,
      "reward_std": 0.3749593496322632,
      "rewards/accuracy_reward_step": 0.6171875,
      "rewards/final_brier_reward_step": 0.4974284768104553,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.7831570506095886,
      "step": 52
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5912162162162162,
      "calib/avg_num_step_conf": 5.17578125,
      "calib/ece": 0.46482352941176475,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.01812894670371311,
      "calib/mean_conf": 0.1155686274509804,
      "calib/mu_c": 0.12317567567567571,
      "calib/mu_w": 0.1050467289719626,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.0512030387050018,
      "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": 837.0,
      "completions/max_terminated_length": 837.0,
      "completions/mean_length": 351.0859375,
      "completions/mean_terminated_length": 352.4627685546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 97.0,
      "epoch": 0.05653333333333333,
      "grad_norm": 0.007529969327151775,
      "learning_rate": 4.083333333333334e-06,
      "loss": 0.0067,
      "num_tokens": 10924556.0,
      "reward": 1.68137526512146,
      "reward_std": 0.2900999188423157,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.5411939024925232,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.7077450752258301,
      "step": 53
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5425420168067226,
      "calib/avg_num_step_conf": 5.06640625,
      "calib/ece": 0.5470078740157479,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0034243697478991697,
      "calib/mean_conf": 0.1262204724409449,
      "calib/mu_c": 0.12735294117647059,
      "calib/mu_w": 0.12392857142857142,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.001968503937007874,
      "calib/std_conf": 0.048916273218669995,
      "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": 2553.0,
      "completions/max_terminated_length": 2553.0,
      "completions/mean_length": 323.78125,
      "completions/mean_terminated_length": 323.78125,
      "completions/min_length": 101.0,
      "completions/min_terminated_length": 101.0,
      "epoch": 0.0576,
      "grad_norm": 0.008591331541538239,
      "learning_rate": 4.055555555555556e-06,
      "loss": 0.0204,
      "num_tokens": 11113676.0,
      "reward": 1.7960519790649414,
      "reward_std": 0.33986392617225647,
      "rewards/accuracy_reward_step": 0.66796875,
      "rewards/final_brier_reward_step": 0.4761546850204468,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.7236782312393188,
      "step": 54
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5162655971479501,
      "calib/avg_num_step_conf": 5.171875,
      "calib/ece": 0.33547808764940246,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.012121212121212116,
      "calib/mean_conf": 0.13862549800796814,
      "calib/mu_c": 0.145,
      "calib/mu_w": 0.13287878787878787,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04937662385499565,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2628.0,
      "completions/max_terminated_length": 2628.0,
      "completions/mean_length": 346.98828125,
      "completions/mean_terminated_length": 346.98828125,
      "completions/min_length": 75.0,
      "completions/min_terminated_length": 75.0,
      "epoch": 0.058666666666666666,
      "grad_norm": 0.007367644924670458,
      "learning_rate": 4.027777777777779e-06,
      "loss": 0.0481,
      "num_tokens": 11310329.0,
      "reward": 1.5486137866973877,
      "reward_std": 0.3473459780216217,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6291975975036621,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.7918204069137573,
      "step": 55
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5328179334326875,
      "calib/avg_num_step_conf": 5.90234375,
      "calib/ece": 0.36231496062992125,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009450074515648305,
      "calib/mean_conf": 0.15737007874015752,
      "calib/mu_c": 0.16190909090909092,
      "calib/mu_w": 0.15245901639344261,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.059392833835551426,
      "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": 1279.0,
      "completions/max_terminated_length": 1279.0,
      "completions/mean_length": 384.47265625,
      "completions/mean_terminated_length": 385.98040771484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 132.0,
      "epoch": 0.05973333333333333,
      "grad_norm": 0.006942269392311573,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0049,
      "num_tokens": 11515594.0,
      "reward": 1.6287014484405518,
      "reward_std": 0.34182778000831604,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.6154594421386719,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8212213516235352,
      "step": 56
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5612567204301074,
      "calib/avg_num_step_conf": 5.98046875,
      "calib/ece": 0.43988418972332005,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.018039845430107504,
      "calib/mean_conf": 0.19252687747035574,
      "calib/mu_c": 0.19915812500000002,
      "calib/mu_w": 0.18111827956989252,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.0756791296263428,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2607.0,
      "completions/max_terminated_length": 2607.0,
      "completions/mean_length": 376.69921875,
      "completions/mean_terminated_length": 378.1764831542969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 130.0,
      "epoch": 0.0608,
      "grad_norm": 0.006421912461519241,
      "learning_rate": 3.972222222222223e-06,
      "loss": -0.0172,
      "num_tokens": 11718821.0,
      "reward": 1.767425298690796,
      "reward_std": 0.4283270239830017,
      "rewards/accuracy_reward_step": 0.625,
      "rewards/final_brier_reward_step": 0.5660962462425232,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.7848550081253052,
      "step": 57
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6133403621930043,
      "calib/avg_num_step_conf": 6.3984375,
      "calib/ece": 0.21723215686274513,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.03453378814190025,
      "calib/mean_conf": 0.23919294117647058,
      "calib/mu_c": 0.2580172413793103,
      "calib/mu_w": 0.22348345323741006,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0007615686274509803,
      "calib/std_conf": 0.0914077172546302,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2475.0,
      "completions/max_terminated_length": 2475.0,
      "completions/mean_length": 416.4296875,
      "completions/mean_terminated_length": 416.4296875,
      "completions/min_length": 133.0,
      "completions/min_terminated_length": 133.0,
      "epoch": 0.06186666666666667,
      "grad_norm": 0.00647732662037015,
      "learning_rate": 3.944444444444445e-06,
      "loss": 0.0426,
      "num_tokens": 11931747.0,
      "reward": 1.5718785524368286,
      "reward_std": 0.46967893838882446,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.7114843726158142,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8650922775268555,
      "step": 58
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5699435238627195,
      "calib/avg_num_step_conf": 6.62109375,
      "calib/ece": 0.23027559055118113,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.018956743002544463,
      "calib/mean_conf": 0.3031102362204724,
      "calib/mu_c": 0.31229007633587785,
      "calib/mu_w": 0.2933333333333334,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.008818897637795278,
      "calib/std_conf": 0.09863323682730286,
      "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": 2618.0,
      "completions/max_terminated_length": 2618.0,
      "completions/mean_length": 424.06640625,
      "completions/mean_terminated_length": 424.06640625,
      "completions/min_length": 135.0,
      "completions/min_terminated_length": 135.0,
      "epoch": 0.06293333333333333,
      "grad_norm": 0.005999838467687368,
      "learning_rate": 3.916666666666667e-06,
      "loss": 0.0858,
      "num_tokens": 12146556.0,
      "reward": 1.6408140659332275,
      "reward_std": 0.39688020944595337,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.6954003572463989,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8209809064865112,
      "step": 59
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.518240275036608,
      "calib/avg_num_step_conf": 6.1875,
      "calib/ece": 0.2580158730158729,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.008892850321512602,
      "calib/mean_conf": 0.29428571428571426,
      "calib/mu_c": 0.298273381294964,
      "calib/mu_w": 0.2893805309734514,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.00035714285714285655,
      "calib/std_conf": 0.09703074763390498,
      "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": 2464.0,
      "completions/max_terminated_length": 2464.0,
      "completions/mean_length": 400.32421875,
      "completions/mean_terminated_length": 400.32421875,
      "completions/min_length": 140.0,
      "completions/min_terminated_length": 140.0,
      "epoch": 0.064,
      "grad_norm": 0.006711548659950495,
      "learning_rate": 3.88888888888889e-06,
      "loss": 0.0433,
      "num_tokens": 12357895.0,
      "reward": 1.6873522996902466,
      "reward_std": 0.47089654207229614,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.6707937717437744,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8520530462265015,
      "step": 60
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5720458327998976,
      "calib/avg_num_step_conf": 6.28515625,
      "calib/ece": 0.21197628458498027,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.02130585072333885,
      "calib/mean_conf": 0.365098814229249,
      "calib/mu_c": 0.37410958904109587,
      "calib/mu_w": 0.352803738317757,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.08898022451267007,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2119.0,
      "completions/max_terminated_length": 2119.0,
      "completions/mean_length": 373.49609375,
      "completions/mean_terminated_length": 373.49609375,
      "completions/min_length": 115.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.06506666666666666,
      "grad_norm": 0.007061994168907404,
      "learning_rate": 3.861111111111112e-06,
      "loss": 0.0512,
      "num_tokens": 12557574.0,
      "reward": 1.7396328449249268,
      "reward_std": 0.373993456363678,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.7051277160644531,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8315281271934509,
      "step": 61
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.49228086040953434,
      "calib/avg_num_step_conf": 7.4921875,
      "calib/ece": 0.0792,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0037542794393126133,
      "calib/mean_conf": 0.39104,
      "calib/mu_c": 0.3930973451327433,
      "calib/mu_w": 0.3893430656934307,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.00912000000000002,
      "calib/std_conf": 0.0883748742573363,
      "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": 2122.0,
      "completions/max_terminated_length": 2122.0,
      "completions/mean_length": 453.0859375,
      "completions/mean_terminated_length": 454.8627624511719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 134.0,
      "epoch": 0.06613333333333334,
      "grad_norm": 0.006072958931326866,
      "learning_rate": 3.833333333333334e-06,
      "loss": 0.056,
      "num_tokens": 12780644.0,
      "reward": 1.5368237495422363,
      "reward_std": 0.5321619510650635,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.7232398390769958,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.8303054571151733,
      "step": 62
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.526360225140713,
      "calib/avg_num_step_conf": 7.25,
      "calib/ece": 0.11614150197628462,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.01916672920575363,
      "calib/mean_conf": 0.4190664031620553,
      "calib/mu_c": 0.42838461538461536,
      "calib/mu_w": 0.40921788617886173,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.01068695652173916,
      "calib/std_conf": 0.09813836172064906,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2181.0,
      "completions/max_terminated_length": 2181.0,
      "completions/mean_length": 455.98828125,
      "completions/mean_terminated_length": 455.98828125,
      "completions/min_length": 131.0,
      "completions/min_terminated_length": 131.0,
      "epoch": 0.0672,
      "grad_norm": 0.006307361647486687,
      "learning_rate": 3.8055555555555556e-06,
      "loss": 0.0697,
      "num_tokens": 13006017.0,
      "reward": 1.6535255908966064,
      "reward_std": 0.5181097984313965,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.7295402884483337,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8454996347427368,
      "step": 63
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5252003101576634,
      "calib/avg_num_step_conf": 7.1640625,
      "calib/ece": 0.12638888888888886,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.008075730162832806,
      "calib/mean_conf": 0.453452380952381,
      "calib/mu_c": 0.4568493150684932,
      "calib/mu_w": 0.4487735849056604,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00023809523809523476,
      "calib/std_conf": 0.09760823110806462,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2729.0,
      "completions/max_terminated_length": 2729.0,
      "completions/mean_length": 419.171875,
      "completions/mean_terminated_length": 420.8157043457031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 139.0,
      "epoch": 0.06826666666666667,
      "grad_norm": 0.0064748735167086124,
      "learning_rate": 3.777777777777778e-06,
      "loss": -0.005,
      "num_tokens": 13217101.0,
      "reward": 1.7371028661727905,
      "reward_std": 0.49576738476753235,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.7233715057373047,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8344149589538574,
      "step": 64
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6220110747545934,
      "calib/avg_num_step_conf": 6.3203125,
      "calib/ece": 0.0823715415019763,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.007905138339920948,
      "calib/gap": 0.04439655172413787,
      "calib/mean_conf": 0.4696442687747036,
      "calib/mu_c": 0.49,
      "calib/mu_w": 0.4456034482758621,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0052569169960474315,
      "calib/std_conf": 0.1100928100058271,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2481.0,
      "completions/max_terminated_length": 2481.0,
      "completions/mean_length": 385.69140625,
      "completions/mean_terminated_length": 387.2039489746094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 146.0,
      "epoch": 0.06933333333333333,
      "grad_norm": 0.00643972447142005,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.0229,
      "num_tokens": 13420862.0,
      "reward": 1.6929399967193604,
      "reward_std": 0.3625901937484741,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.7444347739219666,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8476375937461853,
      "step": 65
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4985714285714286,
      "calib/avg_num_step_conf": 7.9609375,
      "calib/ece": 0.10665338645418335,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": 0.007637460317460154,
      "calib/mean_conf": 0.5084860557768924,
      "calib/mu_c": 0.5123199999999999,
      "calib/mu_w": 0.5046825396825397,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.05856573705179288,
      "calib/std_conf": 0.10827927410165324,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2636.0,
      "completions/max_terminated_length": 2636.0,
      "completions/mean_length": 478.234375,
      "completions/mean_terminated_length": 482.0,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 139.0,
      "epoch": 0.0704,
      "grad_norm": 0.005820919293910265,
      "learning_rate": 3.7222222222222225e-06,
      "loss": 0.0561,
      "num_tokens": 13649642.0,
      "reward": 1.613396406173706,
      "reward_std": 0.41773903369903564,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.7274965047836304,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8354638814926147,
      "step": 66
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.46465773809523814,
      "calib/avg_num_step_conf": 6.98828125,
      "calib/ece": 0.11601562500000005,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.01953125,
      "calib/gap": -0.0027579365079364315,
      "calib/mean_conf": 0.533984375,
      "calib/mu_c": 0.5327777777777778,
      "calib/mu_w": 0.5355357142857142,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.04375000000000004,
      "calib/std_conf": 0.11980022331306137,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1499.0,
      "completions/max_terminated_length": 1499.0,
      "completions/mean_length": 430.859375,
      "completions/mean_terminated_length": 432.5490417480469,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 173.0,
      "epoch": 0.07146666666666666,
      "grad_norm": 0.0060951621271669865,
      "learning_rate": 3.694444444444445e-06,
      "loss": -0.0146,
      "num_tokens": 13864950.0,
      "reward": 1.7398114204406738,
      "reward_std": 0.3218585252761841,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.7373836040496826,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.8468617796897888,
      "step": 67
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4321046600458365,
      "calib/avg_num_step_conf": 7.1875,
      "calib/ece": 0.1264143426294821,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.01593625498007968,
      "calib/gap": -0.03311751973516652,
      "calib/mean_conf": 0.5542231075697212,
      "calib/mu_c": 0.5368067226890758,
      "calib/mu_w": 0.5699242424242423,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.10326693227091635,
      "calib/std_conf": 0.11455306371406776,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2537.0,
      "completions/max_terminated_length": 2537.0,
      "completions/mean_length": 421.73828125,
      "completions/mean_terminated_length": 421.73828125,
      "completions/min_length": 134.0,
      "completions/min_terminated_length": 134.0,
      "epoch": 0.07253333333333334,
      "grad_norm": 0.007131033577024937,
      "learning_rate": 3.6666666666666666e-06,
      "loss": 0.0741,
      "num_tokens": 14077003.0,
      "reward": 1.5687748193740845,
      "reward_std": 0.4624478816986084,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.7006574273109436,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8244417309761047,
      "step": 68
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.47014877516993714,
      "calib/avg_num_step_conf": 6.92578125,
      "calib/ece": 0.16438247011952184,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.01195219123505976,
      "calib/gap": -0.013323072976786032,
      "calib/mean_conf": 0.5745816733067729,
      "calib/mu_c": 0.5672566371681416,
      "calib/mu_w": 0.5805797101449276,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.14438247011952188,
      "calib/std_conf": 0.12311092196212077,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2356.0,
      "completions/max_terminated_length": 2356.0,
      "completions/mean_length": 481.75,
      "completions/mean_terminated_length": 481.75,
      "completions/min_length": 163.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.0736,
      "grad_norm": 0.006324872374534607,
      "learning_rate": 3.638888888888889e-06,
      "loss": 0.0946,
      "num_tokens": 14304827.0,
      "reward": 1.5332916975021362,
      "reward_std": 0.4691646993160248,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.7012875080108643,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8225042819976807,
      "step": 69
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.589452603471295,
      "calib/avg_num_step_conf": 7.07421875,
      "calib/ece": 0.15118110236220478,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.003937007874015748,
      "calib/gap": 0.04132112658147369,
      "calib/mean_conf": 0.5724409448818898,
      "calib/mu_c": 0.5963551401869159,
      "calib/mu_w": 0.5550340136054422,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.15118110236220478,
      "calib/std_conf": 0.11975510796025264,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2603.0,
      "completions/max_terminated_length": 2603.0,
      "completions/mean_length": 459.87890625,
      "completions/mean_terminated_length": 459.87890625,
      "completions/min_length": 127.0,
      "completions/min_terminated_length": 127.0,
      "epoch": 0.07466666666666667,
      "grad_norm": 0.006250609178096056,
      "learning_rate": 3.6111111111111115e-06,
      "loss": 0.0705,
      "num_tokens": 14529548.0,
      "reward": 1.5173765420913696,
      "reward_std": 0.3422887325286865,
      "rewards/accuracy_reward_step": 0.41796875,
      "rewards/final_brier_reward_step": 0.7333765625953674,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8439421653747559,
      "step": 70
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5020195115888896,
      "calib/avg_num_step_conf": 7.4765625,
      "calib/ece": 0.16303149606299214,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.015748031496062992,
      "calib/gap": 0.004131610016777421,
      "calib/mean_conf": 0.6127952755905512,
      "calib/mu_c": 0.6149586776859505,
      "calib/mu_w": 0.610827067669173,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1497244094488189,
      "calib/std_conf": 0.1260095508555356,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2569.0,
      "completions/max_terminated_length": 2569.0,
      "completions/mean_length": 475.30859375,
      "completions/mean_terminated_length": 475.30859375,
      "completions/min_length": 147.0,
      "completions/min_terminated_length": 147.0,
      "epoch": 0.07573333333333333,
      "grad_norm": 0.0063377502374351025,
      "learning_rate": 3.5833333333333335e-06,
      "loss": 0.0823,
      "num_tokens": 14755635.0,
      "reward": 1.589433193206787,
      "reward_std": 0.4913855195045471,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.7125207185745239,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8248997926712036,
      "step": 71
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.615488795863088,
      "calib/avg_num_step_conf": 6.90625,
      "calib/ece": 0.14168627450980392,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.00392156862745098,
      "calib/gap": 0.04447057375030805,
      "calib/mean_conf": 0.6279607843137255,
      "calib/mu_c": 0.6508064516129034,
      "calib/mu_w": 0.6063358778625954,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.14168627450980392,
      "calib/std_conf": 0.11352704796211648,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2058.0,
      "completions/max_terminated_length": 2058.0,
      "completions/mean_length": 430.4296875,
      "completions/mean_terminated_length": 430.4296875,
      "completions/min_length": 151.0,
      "completions/min_terminated_length": 151.0,
      "epoch": 0.0768,
      "grad_norm": 0.007610781118273735,
      "learning_rate": 3.555555555555556e-06,
      "loss": 0.0198,
      "num_tokens": 14970233.0,
      "reward": 1.6153515577316284,
      "reward_std": 0.4659672975540161,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.7365550994873047,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8264135718345642,
      "step": 72
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5020294266869609,
      "calib/avg_num_step_conf": 6.4609375,
      "calib/ece": 0.07590551181102356,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.011811023622047244,
      "calib/gap": -0.0003018772196855535,
      "calib/mean_conf": 0.6051968503937007,
      "calib/mu_c": 0.6050684931506849,
      "calib/mu_w": 0.6053703703703704,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.053149606299212546,
      "calib/std_conf": 0.12075778055629845,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2512.0,
      "completions/max_terminated_length": 2512.0,
      "completions/mean_length": 431.52734375,
      "completions/mean_terminated_length": 431.52734375,
      "completions/min_length": 137.0,
      "completions/min_terminated_length": 137.0,
      "epoch": 0.07786666666666667,
      "grad_norm": 0.006429709959775209,
      "learning_rate": 3.5277777777777784e-06,
      "loss": 0.0205,
      "num_tokens": 15187736.0,
      "reward": 1.7411704063415527,
      "reward_std": 0.5212789177894592,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.7341609597206116,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8242703676223755,
      "step": 73
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4444444444444444,
      "calib/avg_num_step_conf": 7.12109375,
      "calib/ece": 0.1380161943319838,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.022857798766889648,
      "calib/mean_conf": 0.6063562753036437,
      "calib/mu_c": 0.5951587301587302,
      "calib/mu_w": 0.6180165289256199,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.11712550607287452,
      "calib/std_conf": 0.11307481396729323,
      "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": 2816.0,
      "completions/max_terminated_length": 2816.0,
      "completions/mean_length": 466.00390625,
      "completions/mean_terminated_length": 467.8313903808594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.07893333333333333,
      "grad_norm": 0.007456135004758835,
      "learning_rate": 3.5e-06,
      "loss": 0.0813,
      "num_tokens": 15410961.0,
      "reward": 1.5929173231124878,
      "reward_std": 0.5065418481826782,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.6914371252059937,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.7974199652671814,
      "step": 74
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6350518352479686,
      "calib/avg_num_step_conf": 6.3359375,
      "calib/ece": 0.07325490196078434,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.00392156862745098,
      "calib/gap": 0.060122583356682635,
      "calib/mean_conf": 0.6153725490196078,
      "calib/mu_c": 0.6349418604651162,
      "calib/mu_w": 0.5748192771084336,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.007058823529411764,
      "calib/std_conf": 0.12348754970732145,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1018.0,
      "completions/max_terminated_length": 1018.0,
      "completions/mean_length": 385.21875,
      "completions/mean_terminated_length": 386.72943115234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 148.0,
      "epoch": 0.08,
      "grad_norm": 0.007878283970057964,
      "learning_rate": 3.4722222222222224e-06,
      "loss": -0.0011,
      "num_tokens": 15614329.0,
      "reward": 1.9157955646514893,
      "reward_std": 0.33899492025375366,
      "rewards/accuracy_reward_step": 0.671875,
      "rewards/final_brier_reward_step": 0.7850281000137329,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.854716420173645,
      "step": 75
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5301724137931034,
      "calib/avg_num_step_conf": 6.2734375,
      "calib/ece": 0.10820312500000007,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0078125,
      "calib/gap": 0.009285113540790535,
      "calib/mean_conf": 0.6437499999999999,
      "calib/mu_c": 0.6467241379310344,
      "calib/mu_w": 0.6374390243902439,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.036132812500000014,
      "calib/std_conf": 0.11879604370516722,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1288.0,
      "completions/max_terminated_length": 1288.0,
      "completions/mean_length": 444.5234375,
      "completions/mean_terminated_length": 446.2666931152344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.08106666666666666,
      "grad_norm": 0.0073378118686378,
      "learning_rate": 3.444444444444445e-06,
      "loss": 0.0161,
      "num_tokens": 15831183.0,
      "reward": 1.9173394441604614,
      "reward_std": 0.38418424129486084,
      "rewards/accuracy_reward_step": 0.6796875,
      "rewards/final_brier_reward_step": 0.7709265947341919,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.8203064203262329,
      "step": 76
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.43308749580958766,
      "calib/avg_num_step_conf": 6.2421875,
      "calib/ece": 0.16432539682539682,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.007936507936507936,
      "calib/gap": -0.02743479718404296,
      "calib/mean_conf": 0.6399603174603173,
      "calib/mu_c": 0.6296178343949044,
      "calib/mu_w": 0.6570526315789473,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.09063492063492065,
      "calib/std_conf": 0.1143528999496005,
      "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": 2538.0,
      "completions/max_terminated_length": 2538.0,
      "completions/mean_length": 400.80078125,
      "completions/mean_terminated_length": 403.9566955566406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 158.0,
      "epoch": 0.08213333333333334,
      "grad_norm": 0.00810338743031025,
      "learning_rate": 3.416666666666667e-06,
      "loss": 0.0061,
      "num_tokens": 16038452.0,
      "reward": 1.7949559688568115,
      "reward_std": 0.48776593804359436,
      "rewards/accuracy_reward_step": 0.61328125,
      "rewards/final_brier_reward_step": 0.727337121963501,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8040491342544556,
      "step": 77
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6494444444444444,
      "calib/avg_num_step_conf": 6.62109375,
      "calib/ece": 0.1627450980392157,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.027450980392156862,
      "calib/gap": 0.05827777777777787,
      "calib/mean_conf": 0.6863529411764706,
      "calib/mu_c": 0.7137777777777778,
      "calib/mu_w": 0.6555,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.15984313725490196,
      "calib/std_conf": 0.11694658132457889,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1131.0,
      "completions/max_terminated_length": 1131.0,
      "completions/mean_length": 488.796875,
      "completions/mean_terminated_length": 490.7137451171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.0832,
      "grad_norm": 0.007442068308591843,
      "learning_rate": 3.3888888888888893e-06,
      "loss": 0.0194,
      "num_tokens": 16271608.0,
      "reward": 1.6781136989593506,
      "reward_std": 0.4367024898529053,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.738699197769165,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8175055980682373,
      "step": 78
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5434313725490196,
      "calib/avg_num_step_conf": 6.5,
      "calib/ece": 0.15086956521739123,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.05138339920948617,
      "calib/gap": 0.013113071895424988,
      "calib/mean_conf": 0.7208300395256917,
      "calib/mu_c": 0.7260130718954249,
      "calib/mu_w": 0.7128999999999999,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.13347826086956513,
      "calib/std_conf": 0.12622092576739985,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2597.0,
      "completions/max_terminated_length": 2597.0,
      "completions/mean_length": 465.40234375,
      "completions/mean_terminated_length": 465.40234375,
      "completions/min_length": 163.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.08426666666666667,
      "grad_norm": 0.0075131370685994625,
      "learning_rate": 3.3611111111111117e-06,
      "loss": 0.0361,
      "num_tokens": 16497127.0,
      "reward": 1.7761178016662598,
      "reward_std": 0.5020011067390442,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.7291855812072754,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8127859234809875,
      "step": 79
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5484515484515485,
      "calib/avg_num_step_conf": 6.3671875,
      "calib/ece": 0.0939843749999999,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0390625,
      "calib/gap": 0.01381684981684983,
      "calib/mean_conf": 0.722421875,
      "calib/mu_c": 0.7273333333333334,
      "calib/mu_w": 0.7135164835164836,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0859374999999999,
      "calib/std_conf": 0.09966542540662923,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1187.0,
      "completions/max_terminated_length": 1187.0,
      "completions/mean_length": 405.95703125,
      "completions/mean_terminated_length": 407.5490417480469,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.08533333333333333,
      "grad_norm": 0.00772659620270133,
      "learning_rate": 3.3333333333333333e-06,
      "loss": 0.028,
      "num_tokens": 16703212.0,
      "reward": 1.8596093654632568,
      "reward_std": 0.49589797854423523,
      "rewards/accuracy_reward_step": 0.64453125,
      "rewards/final_brier_reward_step": 0.7612203359603882,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.8100296854972839,
      "step": 80
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5213280846670107,
      "calib/avg_num_step_conf": 6.48046875,
      "calib/ece": 0.14794466403162046,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.04743083003952569,
      "calib/gap": 0.009204310789881376,
      "calib/mean_conf": 0.7320553359683795,
      "calib/mu_c": 0.7358389261744966,
      "calib/mu_w": 0.7266346153846153,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1455335968379446,
      "calib/std_conf": 0.10786338939567354,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2871.0,
      "completions/max_terminated_length": 2871.0,
      "completions/mean_length": 477.1953125,
      "completions/mean_terminated_length": 477.1953125,
      "completions/min_length": 169.0,
      "completions/min_terminated_length": 169.0,
      "epoch": 0.0864,
      "grad_norm": 0.007469775155186653,
      "learning_rate": 3.3055555555555558e-06,
      "loss": 0.0522,
      "num_tokens": 16931622.0,
      "reward": 1.7458605766296387,
      "reward_std": 0.3981347680091858,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7216894626617432,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.7930033206939697,
      "step": 81
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.625764742983286,
      "calib/avg_num_step_conf": 5.8359375,
      "calib/ece": 0.14945312499999994,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.04296875,
      "calib/gap": 0.04805424156417537,
      "calib/mean_conf": 0.739296875,
      "calib/mu_c": 0.7590066225165564,
      "calib/mu_w": 0.710952380952381,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.14945312499999994,
      "calib/std_conf": 0.10130926964120497,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1469.0,
      "completions/max_terminated_length": 1469.0,
      "completions/mean_length": 406.875,
      "completions/mean_terminated_length": 408.4706115722656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.08746666666666666,
      "grad_norm": 0.008878878317773342,
      "learning_rate": 3.277777777777778e-06,
      "loss": 0.0258,
      "num_tokens": 17141334.0,
      "reward": 1.7747294902801514,
      "reward_std": 0.38562148809432983,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.7487233877182007,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.8111320734024048,
      "step": 82
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.524426188385802,
      "calib/avg_num_step_conf": 5.88671875,
      "calib/ece": 0.22079051383399217,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.05138339920948617,
      "calib/gap": 0.005280321083657169,
      "calib/mean_conf": 0.7422924901185771,
      "calib/mu_c": 0.7447761194029849,
      "calib/mu_w": 0.7394957983193278,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.21671936758893287,
      "calib/std_conf": 0.1067128383304947,
      "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": 2266.0,
      "completions/max_terminated_length": 2266.0,
      "completions/mean_length": 502.19140625,
      "completions/mean_terminated_length": 504.16082763671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 126.0,
      "epoch": 0.08853333333333334,
      "grad_norm": 0.007681162096560001,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.0135,
      "num_tokens": 17377159.0,
      "reward": 1.6420714855194092,
      "reward_std": 0.35006090998649597,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.6831597685813904,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.783563494682312,
      "step": 83
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6484036881389952,
      "calib/avg_num_step_conf": 5.3671875,
      "calib/ece": 0.17785156250000003,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.05859375,
      "calib/gap": 0.051823997992849424,
      "calib/mean_conf": 0.7598828125000001,
      "calib/mu_c": 0.7815436241610737,
      "calib/mu_w": 0.7297196261682243,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.17785156250000003,
      "calib/std_conf": 0.09597152255794342,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1124.0,
      "completions/max_terminated_length": 1124.0,
      "completions/mean_length": 405.1328125,
      "completions/mean_terminated_length": 406.7215881347656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 121.0,
      "epoch": 0.0896,
      "grad_norm": 0.009634244255721569,
      "learning_rate": 3.2222222222222227e-06,
      "loss": -0.0104,
      "num_tokens": 17586793.0,
      "reward": 1.761276364326477,
      "reward_std": 0.4688519835472107,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7411019802093506,
      "rewards/format_reward_step": 1.0,
      "rewards/stepwise_brier_reward": 0.8118159174919128,
      "step": 84
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5388013830195929,
      "calib/avg_num_step_conf": 6.2265625,
      "calib/ece": 0.23410358565737058,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.09561752988047809,
      "calib/gap": 0.014091432961967021,
      "calib/mean_conf": 0.7750597609561752,
      "calib/mu_c": 0.7814598540145986,
      "calib/mu_w": 0.7673684210526316,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.23167330677290843,
      "calib/std_conf": 0.1057259877061306,
      "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": 2192.0,
      "completions/max_terminated_length": 2192.0,
      "completions/mean_length": 495.96484375,
      "completions/mean_terminated_length": 497.9098205566406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.09066666666666667,
      "grad_norm": 0.0075246118940413,
      "learning_rate": 3.1944444444444443e-06,
      "loss": 0.0339,
      "num_tokens": 17821584.0,
      "reward": 1.6571217775344849,
      "reward_std": 0.470680832862854,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.6796742081642151,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.7847504615783691,
      "step": 85
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6254012345679012,
      "calib/avg_num_step_conf": 5.5,
      "calib/ece": 0.23999999999999994,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.09019607843137255,
      "calib/gap": 0.041703703703703465,
      "calib/mean_conf": 0.7694117647058824,
      "calib/mu_c": 0.7890370370370369,
      "calib/mu_w": 0.7473333333333334,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.23999999999999994,
      "calib/std_conf": 0.09825560433461454,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1524.0,
      "completions/max_terminated_length": 1524.0,
      "completions/mean_length": 461.7578125,
      "completions/mean_terminated_length": 463.56866455078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 119.0,
      "epoch": 0.09173333333333333,
      "grad_norm": 0.008501407690346241,
      "learning_rate": 3.1666666666666667e-06,
      "loss": 0.0133,
      "num_tokens": 18045306.0,
      "reward": 1.6635324954986572,
      "reward_std": 0.41351452469825745,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.7016390562057495,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.7962414026260376,
      "step": 86
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5752205540937936,
      "calib/avg_num_step_conf": 5.13671875,
      "calib/ece": 0.08336363636363635,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.07905138339920949,
      "calib/gap": 0.028101532270546348,
      "calib/mean_conf": 0.7875533596837945,
      "calib/mu_c": 0.7954395604395603,
      "calib/mu_w": 0.767338028169014,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.07577470355731225,
      "calib/std_conf": 0.10003411531993953,
      "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": 2898.0,
      "completions/max_terminated_length": 2898.0,
      "completions/mean_length": 458.23046875,
      "completions/mean_terminated_length": 458.23046875,
      "completions/min_length": 135.0,
      "completions/min_terminated_length": 135.0,
      "epoch": 0.0928,
      "grad_norm": 0.00852898322045803,
      "learning_rate": 3.138888888888889e-06,
      "loss": 0.0834,
      "num_tokens": 18268109.0,
      "reward": 1.9659233093261719,
      "reward_std": 0.4677523672580719,
      "rewards/accuracy_reward_step": 0.7109375,
      "rewards/final_brier_reward_step": 0.7854979038238525,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8360080122947693,
      "step": 87
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5800427565392354,
      "calib/avg_num_step_conf": 5.5859375,
      "calib/ece": 0.24023385826771654,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.12992125984251968,
      "calib/gap": 0.024400603621730355,
      "calib/mean_conf": 0.7914984251968503,
      "calib/mu_c": 0.8022577464788733,
      "calib/mu_w": 0.7778571428571429,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.23633858267716534,
      "calib/std_conf": 0.10777079168692731,
      "calib/step_conf_rate": 0.98046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1592.0,
      "completions/max_terminated_length": 1592.0,
      "completions/mean_length": 485.9375,
      "completions/mean_terminated_length": 485.9375,
      "completions/min_length": 64.0,
      "completions/min_terminated_length": 64.0,
      "epoch": 0.09386666666666667,
      "grad_norm": 0.008167689666152,
      "learning_rate": 3.1111111111111116e-06,
      "loss": 0.0287,
      "num_tokens": 18502357.0,
      "reward": 1.689322829246521,
      "reward_std": 0.3582633137702942,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6878530979156494,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.7803753614425659,
      "step": 88
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.6251151770435699,
      "calib/avg_num_step_conf": 5.28125,
      "calib/ece": 0.2280722891566265,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.17670682730923695,
      "calib/gap": 0.0585678557325261,
      "calib/mean_conf": 0.7983534136546184,
      "calib/mu_c": 0.8235211267605633,
      "calib/mu_w": 0.7649532710280372,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.9609375,
      "calib/pce": 0.2280722891566265,
      "calib/std_conf": 0.12099362556806864,
      "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": 0.0,
      "completions/max_length": 2545.0,
      "completions/max_terminated_length": 2545.0,
      "completions/mean_length": 507.54296875,
      "completions/mean_terminated_length": 507.54296875,
      "completions/min_length": 179.0,
      "completions/min_terminated_length": 179.0,
      "epoch": 0.09493333333333333,
      "grad_norm": 0.008763005957007408,
      "learning_rate": 3.0833333333333336e-06,
      "loss": 0.0242,
      "num_tokens": 18741176.0,
      "reward": 1.668036937713623,
      "reward_std": 0.5595136880874634,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6811015605926514,
      "rewards/format_reward_step": 0.953125,
      "rewards/stepwise_brier_reward": 0.7566711902618408,
      "step": 89
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5663843691651108,
      "calib/avg_num_step_conf": 5.859375,
      "calib/ece": 0.155921568627451,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.16470588235294117,
      "calib/gap": 0.02716791144305164,
      "calib/mean_conf": 0.7944313725490195,
      "calib/mu_c": 0.804233128834356,
      "calib/mu_w": 0.7770652173913044,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1555686274509804,
      "calib/std_conf": 0.11551916145307665,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2360.0,
      "completions/max_terminated_length": 2360.0,
      "completions/mean_length": 486.13671875,
      "completions/mean_terminated_length": 486.13671875,
      "completions/min_length": 167.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.096,
      "grad_norm": 0.00820145569741726,
      "learning_rate": 3.055555555555556e-06,
      "loss": 0.0478,
      "num_tokens": 18968947.0,
      "reward": 1.8456428050994873,
      "reward_std": 0.41628795862197876,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.7415671348571777,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8285040855407715,
      "step": 90
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5138343864316595,
      "calib/avg_num_step_conf": 5.88671875,
      "calib/ece": 0.18202500000000005,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.19047619047619047,
      "calib/gap": 0.016582873295643452,
      "calib/mean_conf": 0.7943265873015873,
      "calib/mu_c": 0.8007096774193547,
      "calib/mu_w": 0.7841268041237113,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.18063611111111116,
      "calib/std_conf": 0.12162123763516405,
      "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": 2692.0,
      "completions/max_terminated_length": 2692.0,
      "completions/mean_length": 532.609375,
      "completions/mean_terminated_length": 532.609375,
      "completions/min_length": 259.0,
      "completions/min_terminated_length": 259.0,
      "epoch": 0.09706666666666666,
      "grad_norm": 0.008124892599880695,
      "learning_rate": 3.0277777777777776e-06,
      "loss": 0.0222,
      "num_tokens": 19213007.0,
      "reward": 1.7842680215835571,
      "reward_std": 0.43133682012557983,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.7128589749336243,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8226507902145386,
      "step": 91
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6274210321071103,
      "calib/avg_num_step_conf": 5.828125,
      "calib/ece": 0.17494117647058827,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.18823529411764706,
      "calib/gap": 0.058860652541271485,
      "calib/mean_conf": 0.7906274509803922,
      "calib/mu_c": 0.8132484076433122,
      "calib/mu_w": 0.7543877551020407,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.17494117647058827,
      "calib/std_conf": 0.12230626309204984,
      "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": 1252.0,
      "completions/max_terminated_length": 1252.0,
      "completions/mean_length": 478.87109375,
      "completions/mean_terminated_length": 480.7490539550781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.09813333333333334,
      "grad_norm": 0.008572989143431187,
      "learning_rate": 3e-06,
      "loss": 0.0123,
      "num_tokens": 19442318.0,
      "reward": 1.8070766925811768,
      "reward_std": 0.4033738374710083,
      "rewards/accuracy_reward_step": 0.61328125,
      "rewards/final_brier_reward_step": 0.7377781271934509,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8342782855033875,
      "step": 92
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.4495913697286695,
      "calib/avg_num_step_conf": 5.89453125,
      "calib/ece": 0.2464112903225807,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.16532258064516128,
      "calib/gap": -0.011000326904216928,
      "calib/mean_conf": 0.781491935483871,
      "calib/mu_c": 0.776390977443609,
      "calib/mu_w": 0.7873913043478259,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.24580645161290327,
      "calib/std_conf": 0.11617124428824763,
      "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": 2576.0,
      "completions/max_terminated_length": 2576.0,
      "completions/mean_length": 532.98828125,
      "completions/mean_terminated_length": 537.18505859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 169.0,
      "epoch": 0.0992,
      "grad_norm": 0.007579736411571503,
      "learning_rate": 2.9722222222222225e-06,
      "loss": 0.0318,
      "num_tokens": 19684539.0,
      "reward": 1.6276981830596924,
      "reward_std": 0.49937906861305237,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.6468539237976074,
      "rewards/format_reward_step": 0.9609375,
      "rewards/stepwise_brier_reward": 0.7780015468597412,
      "step": 93
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6616434653331532,
      "calib/avg_num_step_conf": 6.11328125,
      "calib/ece": 0.18915662650602408,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.2891566265060241,
      "calib/gap": 0.07415056088660643,
      "calib/mean_conf": 0.7857831325301203,
      "calib/mu_c": 0.8149668874172186,
      "calib/mu_w": 0.7408163265306121,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.1842570281124498,
      "calib/std_conf": 0.14246358358266636,
      "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": 3025.0,
      "completions/max_terminated_length": 3025.0,
      "completions/mean_length": 536.671875,
      "completions/mean_terminated_length": 536.671875,
      "completions/min_length": 161.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.10026666666666667,
      "grad_norm": 0.007810818962752819,
      "learning_rate": 2.944444444444445e-06,
      "loss": 0.1054,
      "num_tokens": 19930607.0,
      "reward": 1.7506338357925415,
      "reward_std": 0.4305832087993622,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.7228222489356995,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.8031507730484009,
      "step": 94
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6220264883631219,
      "calib/avg_num_step_conf": 6.23828125,
      "calib/ece": 0.18133333333333326,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.2,
      "calib/gap": 0.06304423299472794,
      "calib/mean_conf": 0.7805490196078431,
      "calib/mu_c": 0.8055194805194804,
      "calib/mu_w": 0.7424752475247525,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.17898039215686268,
      "calib/std_conf": 0.13394355172257616,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1896.0,
      "completions/max_terminated_length": 1896.0,
      "completions/mean_length": 518.734375,
      "completions/mean_terminated_length": 520.7686767578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 158.0,
      "epoch": 0.10133333333333333,
      "grad_norm": 0.00817008689045906,
      "learning_rate": 2.916666666666667e-06,
      "loss": 0.0044,
      "num_tokens": 20169531.0,
      "reward": 1.7938333749771118,
      "reward_std": 0.47976094484329224,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.7389242649078369,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.834846556186676,
      "step": 95
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6551065975223279,
      "calib/avg_num_step_conf": 5.90234375,
      "calib/ece": 0.08769609374999998,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.15625,
      "calib/gap": 0.07325730337078662,
      "calib/mean_conf": 0.76183671875,
      "calib/mu_c": 0.7841573033707866,
      "calib/mu_w": 0.7109,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.07711015624999996,
      "calib/std_conf": 0.12444150883239642,
      "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": 1541.0,
      "completions/max_terminated_length": 1541.0,
      "completions/mean_length": 473.859375,
      "completions/mean_terminated_length": 475.7176818847656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.1024,
      "grad_norm": 0.00906490907073021,
      "learning_rate": 2.888888888888889e-06,
      "loss": 0.0075,
      "num_tokens": 20396655.0,
      "reward": 1.9543486833572388,
      "reward_std": 0.4145694971084595,
      "rewards/accuracy_reward_step": 0.6953125,
      "rewards/final_brier_reward_step": 0.7965941429138184,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8567379713058472,
      "step": 96
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6316791430046463,
      "calib/avg_num_step_conf": 6.06640625,
      "calib/ece": 0.1992490118577074,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.2134387351778656,
      "calib/gap": 0.056542333505420705,
      "calib/mean_conf": 0.7854150197628459,
      "calib/mu_c": 0.8086577181208053,
      "calib/mu_w": 0.7521153846153846,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.19786561264822125,
      "calib/std_conf": 0.12686056306888188,
      "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": 2338.0,
      "completions/max_terminated_length": 2338.0,
      "completions/mean_length": 514.7265625,
      "completions/mean_terminated_length": 516.7451171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 183.0,
      "epoch": 0.10346666666666667,
      "grad_norm": 0.009137057699263096,
      "learning_rate": 2.861111111111111e-06,
      "loss": 0.0054,
      "num_tokens": 20633497.0,
      "reward": 1.743384599685669,
      "reward_std": 0.48850369453430176,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7179961204528809,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8024172782897949,
      "step": 97
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5226375272331154,
      "calib/avg_num_step_conf": 5.8359375,
      "calib/ece": 0.20200803212851404,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.28112449799196787,
      "calib/gap": 0.009822303921568687,
      "calib/mean_conf": 0.8073895582329317,
      "calib/mu_c": 0.8111764705882353,
      "calib/mu_w": 0.8013541666666666,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.19746987951807227,
      "calib/std_conf": 0.12097998793738965,
      "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": 2740.0,
      "completions/max_terminated_length": 2740.0,
      "completions/mean_length": 535.359375,
      "completions/mean_terminated_length": 539.5748291015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 160.0,
      "epoch": 0.10453333333333334,
      "grad_norm": 0.008485455065965652,
      "learning_rate": 2.8333333333333335e-06,
      "loss": -0.0083,
      "num_tokens": 20876733.0,
      "reward": 1.7485198974609375,
      "reward_std": 0.48161444067955017,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.6922656297683716,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.7861887216567993,
      "step": 98
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.555098845175181,
      "calib/avg_num_step_conf": 6.38671875,
      "calib/ece": 0.3326612903225806,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.29435483870967744,
      "calib/gap": 0.02752397729496936,
      "calib/mean_conf": 0.8044354838709677,
      "calib/mu_c": 0.8189743589743588,
      "calib/mu_w": 0.7914503816793894,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.3326612903225806,
      "calib/std_conf": 0.12465308258907766,
      "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": 2532.0,
      "completions/max_terminated_length": 2532.0,
      "completions/mean_length": 600.7421875,
      "completions/mean_terminated_length": 607.8656616210938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 177.0,
      "epoch": 0.1056,
      "grad_norm": 0.007896266877651215,
      "learning_rate": 2.805555555555556e-06,
      "loss": 0.0384,
      "num_tokens": 21136323.0,
      "reward": 1.5078518390655518,
      "reward_std": 0.5043842196464539,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.6141847372055054,
      "rewards/format_reward_step": 0.9609375,
      "rewards/stepwise_brier_reward": 0.7531599402427673,
      "step": 99
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6357911220043573,
      "calib/avg_num_step_conf": 6.3203125,
      "calib/ece": 0.20755020080321285,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.3654618473895582,
      "calib/gap": 0.07124795751633994,
      "calib/mean_conf": 0.822008032128514,
      "calib/mu_c": 0.8494771241830067,
      "calib/mu_w": 0.7782291666666667,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.20755020080321285,
      "calib/std_conf": 0.1281000484989945,
      "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": 1633.0,
      "completions/max_terminated_length": 1633.0,
      "completions/mean_length": 576.546875,
      "completions/mean_terminated_length": 578.807861328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 234.0,
      "epoch": 0.10666666666666667,
      "grad_norm": 0.00841725617647171,
      "learning_rate": 2.7777777777777783e-06,
      "loss": 0.0162,
      "num_tokens": 21391327.0,
      "reward": 1.7648289203643799,
      "reward_std": 0.35282963514328003,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.7172086238861084,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.8108574151992798,
      "step": 100
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.536111111111111,
      "calib/avg_num_step_conf": 6.703125,
      "calib/ece": 0.31154523809523804,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.38095238095238093,
      "calib/gap": 0.005383333333333518,
      "calib/mean_conf": 0.8178198412698412,
      "calib/mu_c": 0.8203833333333335,
      "calib/mu_w": 0.815,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.3027777777777777,
      "calib/std_conf": 0.1456421260900784,
      "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": 3056.0,
      "completions/max_terminated_length": 3056.0,
      "completions/mean_length": 631.36328125,
      "completions/mean_terminated_length": 633.8392333984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 178.0,
      "epoch": 0.10773333333333333,
      "grad_norm": 0.008130279369652271,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.0191,
      "num_tokens": 21659948.0,
      "reward": 1.6108900308609009,
      "reward_std": 0.5622965693473816,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.6314835548400879,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.7652013301849365,
      "step": 101
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5404174838890445,
      "calib/avg_num_step_conf": 6.58203125,
      "calib/ece": 0.17404761904761898,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.26587301587301587,
      "calib/gap": 0.014502661810030903,
      "calib/mean_conf": 0.8024603174603174,
      "calib/mu_c": 0.8074096385542169,
      "calib/mu_w": 0.792906976744186,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.1588888888888888,
      "calib/std_conf": 0.13047645599169677,
      "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": 2219.0,
      "completions/max_terminated_length": 2219.0,
      "completions/mean_length": 518.046875,
      "completions/mean_terminated_length": 518.046875,
      "completions/min_length": 192.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.1088,
      "grad_norm": 0.009473362937569618,
      "learning_rate": 2.7222222222222224e-06,
      "loss": 0.0185,
      "num_tokens": 21899264.0,
      "reward": 1.8465378284454346,
      "reward_std": 0.3376423120498657,
      "rewards/accuracy_reward_step": 0.6484375,
      "rewards/final_brier_reward_step": 0.7265894412994385,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8158115148544312,
      "step": 102
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5657860657860659,
      "calib/avg_num_step_conf": 6.046875,
      "calib/ece": 0.26681314741035844,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.3784860557768924,
      "calib/gap": 0.03490389147889139,
      "calib/mean_conf": 0.8312752988047809,
      "calib/mu_c": 0.8462937062937064,
      "calib/mu_w": 0.811389814814815,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.2641836653386453,
      "calib/std_conf": 0.13386125579940442,
      "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": 2054.0,
      "completions/max_terminated_length": 2054.0,
      "completions/mean_length": 629.0390625,
      "completions/mean_terminated_length": 631.5059204101562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 185.0,
      "epoch": 0.10986666666666667,
      "grad_norm": 0.00732939038425684,
      "learning_rate": 2.6944444444444444e-06,
      "loss": 0.0255,
      "num_tokens": 22164850.0,
      "reward": 1.6884933710098267,
      "reward_std": 0.4924623966217041,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.6660183668136597,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.7988923788070679,
      "step": 103
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5902724726254138,
      "calib/avg_num_step_conf": 6.7265625,
      "calib/ece": 0.2969322709163345,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.36254980079681276,
      "calib/gap": 0.03865991851285999,
      "calib/mean_conf": 0.8193227091633465,
      "calib/mu_c": 0.8376515151515151,
      "calib/mu_w": 0.7989915966386552,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.29517928286852574,
      "calib/std_conf": 0.1326797789689224,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 2485.0,
      "completions/max_terminated_length": 2485.0,
      "completions/mean_length": 562.4453125,
      "completions/mean_terminated_length": 566.8740234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.11093333333333333,
      "grad_norm": 0.00816579069942236,
      "learning_rate": 2.666666666666667e-06,
      "loss": -0.0043,
      "num_tokens": 22415516.0,
      "reward": 1.6142315864562988,
      "reward_std": 0.43704116344451904,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.6459335684776306,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.7797425985336304,
      "step": 104
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.4416758951642672,
      "calib/avg_num_step_conf": 6.02734375,
      "calib/ece": 0.32403921568627453,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.3686274509803922,
      "calib/gap": -0.02618493909191577,
      "calib/mean_conf": 0.8160392156862747,
      "calib/mu_c": 0.8031007751937985,
      "calib/mu_w": 0.8292857142857143,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.3170980392156863,
      "calib/std_conf": 0.12947404324378942,
      "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": 1756.0,
      "completions/max_terminated_length": 1756.0,
      "completions/mean_length": 571.13671875,
      "completions/mean_terminated_length": 573.3765258789062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 205.0,
      "epoch": 0.112,
      "grad_norm": 0.008314027450978756,
      "learning_rate": 2.6388888888888893e-06,
      "loss": 0.0259,
      "num_tokens": 22667487.0,
      "reward": 1.595670461654663,
      "reward_std": 0.527245283126831,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.6148367524147034,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.7678452730178833,
      "step": 105
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.528975219411461,
      "calib/avg_num_step_conf": 5.75,
      "calib/ece": 0.24782490118577077,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.3359683794466403,
      "calib/gap": 0.0043418430562724986,
      "calib/mean_conf": 0.8170762845849803,
      "calib/mu_c": 0.8188610738255033,
      "calib/mu_w": 0.8145192307692308,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2379841897233202,
      "calib/std_conf": 0.13104552836618566,
      "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": 3059.0,
      "completions/max_terminated_length": 3059.0,
      "completions/mean_length": 534.875,
      "completions/mean_terminated_length": 536.9725952148438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 191.0,
      "epoch": 0.11306666666666666,
      "grad_norm": 0.008834516629576683,
      "learning_rate": 2.6111111111111113e-06,
      "loss": 0.0164,
      "num_tokens": 22908999.0,
      "reward": 1.7355353832244873,
      "reward_std": 0.4472808539867401,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.6821659803390503,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.7990379333496094,
      "step": 106
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6033805031446541,
      "calib/avg_num_step_conf": 6.2578125,
      "calib/ece": 0.20644705882352948,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.3411764705882353,
      "calib/gap": 0.04999056603773566,
      "calib/mean_conf": 0.8223372549019606,
      "calib/mu_c": 0.8411572327044023,
      "calib/mu_w": 0.7911666666666667,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.20262745098039223,
      "calib/std_conf": 0.14195010164550892,
      "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": 1654.0,
      "completions/max_terminated_length": 1654.0,
      "completions/mean_length": 527.75,
      "completions/mean_terminated_length": 529.8196411132812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.11413333333333334,
      "grad_norm": 0.008740349672734737,
      "learning_rate": 2.5833333333333337e-06,
      "loss": 0.0026,
      "num_tokens": 23148719.0,
      "reward": 1.8117653131484985,
      "reward_std": 0.46492135524749756,
      "rewards/accuracy_reward_step": 0.62109375,
      "rewards/final_brier_reward_step": 0.72620689868927,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8021043539047241,
      "step": 107
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5571535580524344,
      "calib/avg_num_step_conf": 5.765625,
      "calib/ece": 0.17939960474308297,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.5335968379446641,
      "calib/gap": 0.03685292883895119,
      "calib/mean_conf": 0.8673988142292491,
      "calib/mu_c": 0.878323595505618,
      "calib/mu_w": 0.8414706666666668,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.17162055335968376,
      "calib/std_conf": 0.1360048466011017,
      "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": 2941.0,
      "completions/max_terminated_length": 2941.0,
      "completions/mean_length": 555.4375,
      "completions/mean_terminated_length": 557.61572265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 180.0,
      "epoch": 0.1152,
      "grad_norm": 0.007499896455556154,
      "learning_rate": 2.5555555555555557e-06,
      "loss": 0.0035,
      "num_tokens": 23394143.0,
      "reward": 1.924324631690979,
      "reward_std": 0.4233069121837616,
      "rewards/accuracy_reward_step": 0.6953125,
      "rewards/final_brier_reward_step": 0.7488811016082764,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8077921867370605,
      "step": 108
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.648555439875065,
      "calib/avg_num_step_conf": 6.53125,
      "calib/ece": 0.3128469879518073,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.4939759036144578,
      "calib/gap": 0.0742291970327954,
      "calib/mean_conf": 0.8503658634538153,
      "calib/mu_c": 0.884052205882353,
      "calib/mu_w": 0.8098230088495576,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3085140562248997,
      "calib/std_conf": 0.1461503308109222,
      "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": 2683.0,
      "completions/max_terminated_length": 2683.0,
      "completions/mean_length": 570.71484375,
      "completions/mean_terminated_length": 575.2086791992188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.11626666666666667,
      "grad_norm": 0.007912173867225647,
      "learning_rate": 2.5277777777777778e-06,
      "loss": 0.0096,
      "num_tokens": 23644846.0,
      "reward": 1.6374064683914185,
      "reward_std": 0.31125980615615845,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6559866070747375,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.7686392664909363,
      "step": 109
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5742711558307534,
      "calib/avg_num_step_conf": 5.578125,
      "calib/ece": 0.291632,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.388,
      "calib/gap": 0.04077992776057782,
      "calib/mean_conf": 0.834272,
      "calib/mu_c": 0.8528676470588236,
      "calib/mu_w": 0.8120877192982457,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.290952,
      "calib/std_conf": 0.13710069298147257,
      "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": 2516.0,
      "completions/max_terminated_length": 2516.0,
      "completions/mean_length": 514.8828125,
      "completions/mean_terminated_length": 520.9881591796875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 182.0,
      "epoch": 0.11733333333333333,
      "grad_norm": 0.007297968491911888,
      "learning_rate": 2.5e-06,
      "loss": 0.0201,
      "num_tokens": 23881576.0,
      "reward": 1.638474941253662,
      "reward_std": 0.5752736330032349,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6495566368103027,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.7715306282043457,
      "step": 110
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5686034658511723,
      "calib/avg_num_step_conf": 5.1953125,
      "calib/ece": 0.28118852459016397,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.3975409836065574,
      "calib/gap": 0.02853075093442048,
      "calib/mean_conf": 0.8311065573770493,
      "calib/mu_c": 0.8438518518518517,
      "calib/mu_w": 0.8153211009174313,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.27950819672131155,
      "calib/std_conf": 0.13193082560147212,
      "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": 0.015625,
      "completions/max_length": 1838.0,
      "completions/max_terminated_length": 1838.0,
      "completions/mean_length": 550.53515625,
      "completions/mean_terminated_length": 559.2738647460938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.1184,
      "grad_norm": 0.009061508812010288,
      "learning_rate": 2.4722222222222226e-06,
      "loss": -0.0144,
      "num_tokens": 24129921.0,
      "reward": 1.6133426427841187,
      "reward_std": 0.4934231638908386,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.6369316577911377,
      "rewards/format_reward_step": 0.9453125,
      "rewards/stepwise_brier_reward": 0.7617512941360474,
      "step": 111
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.7267038007863695,
      "calib/avg_num_step_conf": 5.80078125,
      "calib/ece": 0.27678714859437753,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.43775100401606426,
      "calib/gap": 0.10810026212319801,
      "calib/mean_conf": 0.8390361445783133,
      "calib/mu_c": 0.8863571428571428,
      "calib/mu_w": 0.7782568807339448,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.27678714859437753,
      "calib/std_conf": 0.13733505878793048,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 1572.0,
      "completions/max_terminated_length": 1572.0,
      "completions/mean_length": 578.35546875,
      "completions/mean_terminated_length": 585.2134399414062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 213.0,
      "epoch": 0.11946666666666667,
      "grad_norm": 0.007538562640547752,
      "learning_rate": 2.4444444444444447e-06,
      "loss": -0.0037,
      "num_tokens": 24385900.0,
      "reward": 1.6774992942810059,
      "reward_std": 0.47562873363494873,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6921570301055908,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.7912775278091431,
      "step": 112
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6633072041806645,
      "calib/avg_num_step_conf": 5.4296875,
      "calib/ece": 0.2940392156862745,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.43529411764705883,
      "calib/gap": 0.06418253079507286,
      "calib/mean_conf": 0.8469803921568628,
      "calib/mu_c": 0.8756737588652482,
      "calib/mu_w": 0.8114912280701754,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2940392156862745,
      "calib/std_conf": 0.11789950692509084,
      "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": 1673.0,
      "completions/max_terminated_length": 1673.0,
      "completions/mean_length": 481.45703125,
      "completions/mean_terminated_length": 483.3451232910156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.12053333333333334,
      "grad_norm": 0.00867400411516428,
      "learning_rate": 2.4166666666666667e-06,
      "loss": 0.0009,
      "num_tokens": 24614353.0,
      "reward": 1.6906211376190186,
      "reward_std": 0.4227238893508911,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.676214873790741,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.805019736289978,
      "step": 113
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5278862242052426,
      "calib/avg_num_step_conf": 5.97265625,
      "calib/ece": 0.24046294820717135,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.43824701195219123,
      "calib/gap": 0.01336261851645315,
      "calib/mean_conf": 0.855393625498008,
      "calib/mu_c": 0.8600785276073621,
      "calib/mu_w": 0.846715909090909,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.22322709163346616,
      "calib/std_conf": 0.13699127621919044,
      "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": 2281.0,
      "completions/max_terminated_length": 2281.0,
      "completions/mean_length": 511.11328125,
      "completions/mean_terminated_length": 513.11767578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 119.0,
      "epoch": 0.1216,
      "grad_norm": 0.008046331815421581,
      "learning_rate": 2.388888888888889e-06,
      "loss": 0.0096,
      "num_tokens": 24850222.0,
      "reward": 1.8118925094604492,
      "reward_std": 0.4035493731498718,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.698589563369751,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.7833551168441772,
      "step": 114
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.49947247564078695,
      "calib/avg_num_step_conf": 5.66015625,
      "calib/ece": 0.34688976377952757,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.4645669291338583,
      "calib/gap": 0.007636690870725604,
      "calib/mean_conf": 0.8626377952755905,
      "calib/mu_c": 0.8663358778625955,
      "calib/mu_w": 0.8586991869918699,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.34688976377952757,
      "calib/std_conf": 0.10887637787018503,
      "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": 1380.0,
      "completions/max_terminated_length": 1380.0,
      "completions/mean_length": 473.11328125,
      "completions/mean_terminated_length": 474.9686584472656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 199.0,
      "epoch": 0.12266666666666666,
      "grad_norm": 0.00931251235306263,
      "learning_rate": 2.361111111111111e-06,
      "loss": 0.0275,
      "num_tokens": 25076603.0,
      "reward": 1.6056057214736938,
      "reward_std": 0.46782374382019043,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.6131113767623901,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.76243656873703,
      "step": 115
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5341447368421053,
      "calib/avg_num_step_conf": 6.046875,
      "calib/ece": 0.25729880952380957,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.373015873015873,
      "calib/gap": 0.016615131578947562,
      "calib/mean_conf": 0.8526218253968254,
      "calib/mu_c": 0.8592151315789475,
      "calib/mu_w": 0.8425999999999999,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.2533730158730159,
      "calib/std_conf": 0.12191742342570715,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 1682.0,
      "completions/max_terminated_length": 1682.0,
      "completions/mean_length": 544.50390625,
      "completions/mean_terminated_length": 548.7913208007812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.12373333333333333,
      "grad_norm": 0.00802499521523714,
      "learning_rate": 2.3333333333333336e-06,
      "loss": 0.0243,
      "num_tokens": 25320516.0,
      "reward": 1.7363612651824951,
      "reward_std": 0.47284185886383057,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.6692519187927246,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.7683802843093872,
      "step": 116
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.543340987370838,
      "calib/avg_num_step_conf": 5.86328125,
      "calib/ece": 0.2934298804780878,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.30677290836653387,
      "calib/gap": 0.025156869498660583,
      "calib/mean_conf": 0.8262585657370517,
      "calib/mu_c": 0.8379850746268658,
      "calib/mu_w": 0.8128282051282052,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2929119521912352,
      "calib/std_conf": 0.12798486011501525,
      "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": 1656.0,
      "completions/max_terminated_length": 1656.0,
      "completions/mean_length": 510.26953125,
      "completions/mean_terminated_length": 512.2706298828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 30.0,
      "epoch": 0.1248,
      "grad_norm": 0.008473314344882965,
      "learning_rate": 2.305555555555556e-06,
      "loss": -0.0105,
      "num_tokens": 25557745.0,
      "reward": 1.6333932876586914,
      "reward_std": 0.5400822162628174,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.6414456367492676,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.7827528715133667,
      "step": 117
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.47040350209364296,
      "calib/avg_num_step_conf": 6.09375,
      "calib/ece": 0.3058339920948617,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.5019762845849802,
      "calib/gap": -0.011115391447785727,
      "calib/mean_conf": 0.8395415019762845,
      "calib/mu_c": 0.8346647887323944,
      "calib/mu_w": 0.8457801801801801,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.29205533596837946,
      "calib/std_conf": 0.16702360101898417,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2584.0,
      "completions/max_terminated_length": 2584.0,
      "completions/mean_length": 517.484375,
      "completions/mean_terminated_length": 519.5137329101562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.12586666666666665,
      "grad_norm": 0.009097181260585785,
      "learning_rate": 2.277777777777778e-06,
      "loss": 0.0236,
      "num_tokens": 25794229.0,
      "reward": 1.6698601245880127,
      "reward_std": 0.3848813772201538,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6331539750099182,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.749411940574646,
      "step": 118
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.6242059737802407,
      "calib/avg_num_step_conf": 5.37890625,
      "calib/ece": 0.2405891566265059,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.3373493975903614,
      "calib/gap": 0.05021801594810105,
      "calib/mean_conf": 0.8312983935742972,
      "calib/mu_c": 0.8510629139072848,
      "calib/mu_w": 0.8008448979591838,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.232730923694779,
      "calib/std_conf": 0.14048016223422843,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 3059.0,
      "completions/max_terminated_length": 3059.0,
      "completions/mean_length": 584.92578125,
      "completions/mean_terminated_length": 587.2196655273438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 175.0,
      "epoch": 0.12693333333333334,
      "grad_norm": 0.0076350923627614975,
      "learning_rate": 2.25e-06,
      "loss": 0.0301,
      "num_tokens": 26049034.0,
      "reward": 1.7369259595870972,
      "reward_std": 0.5274437665939331,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.6887651085853577,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.7901886701583862,
      "step": 119
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6534184073572397,
      "calib/avg_num_step_conf": 5.578125,
      "calib/ece": 0.17479126984126983,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.29365079365079366,
      "calib/gap": 0.061183845440935025,
      "calib/mean_conf": 0.8218753968253968,
      "calib/mu_c": 0.8420272189349112,
      "calib/mu_w": 0.7808433734939761,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.163015873015873,
      "calib/std_conf": 0.1506151941082675,
      "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": 1653.0,
      "completions/max_terminated_length": 1653.0,
      "completions/mean_length": 471.40234375,
      "completions/mean_terminated_length": 473.2510070800781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 38.0,
      "epoch": 0.128,
      "grad_norm": 0.008784173056483269,
      "learning_rate": 2.222222222222222e-06,
      "loss": -0.0124,
      "num_tokens": 26276401.0,
      "reward": 1.8625600337982178,
      "reward_std": 0.32405993342399597,
      "rewards/accuracy_reward_step": 0.66015625,
      "rewards/final_brier_reward_step": 0.7397279143333435,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.8042622804641724,
      "step": 120
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5129757785467127,
      "calib/avg_num_step_conf": 5.65234375,
      "calib/ece": 0.2959215686274509,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.35294117647058826,
      "calib/gap": 0.01045168067226887,
      "calib/mean_conf": 0.8243137254901961,
      "calib/mu_c": 0.8291911764705883,
      "calib/mu_w": 0.8187394957983194,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2934509803921568,
      "calib/std_conf": 0.13685049667086305,
      "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": 1348.0,
      "completions/max_terminated_length": 1348.0,
      "completions/mean_length": 521.18359375,
      "completions/mean_terminated_length": 523.2274780273438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 175.0,
      "epoch": 0.12906666666666666,
      "grad_norm": 0.008949114009737968,
      "learning_rate": 2.1944444444444445e-06,
      "loss": -0.0196,
      "num_tokens": 26514880.0,
      "reward": 1.659590721130371,
      "reward_std": 0.5529522895812988,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6503655910491943,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8083097338676453,
      "step": 121
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6025606917193216,
      "calib/avg_num_step_conf": 5.30859375,
      "calib/ece": 0.2067460317460318,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.3373015873015873,
      "calib/gap": 0.051185899567675386,
      "calib/mean_conf": 0.8205555555555556,
      "calib/mu_c": 0.8402580645161289,
      "calib/mu_w": 0.7890721649484536,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.20611111111111116,
      "calib/std_conf": 0.13267579267221016,
      "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": 2802.0,
      "completions/max_terminated_length": 2802.0,
      "completions/mean_length": 485.28515625,
      "completions/mean_terminated_length": 487.1882629394531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 44.0,
      "epoch": 0.13013333333333332,
      "grad_norm": 0.008620054461061954,
      "learning_rate": 2.166666666666667e-06,
      "loss": -0.007,
      "num_tokens": 26746457.0,
      "reward": 1.779456377029419,
      "reward_std": 0.36303144693374634,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.7124691009521484,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8116064071655273,
      "step": 122
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5338949843260188,
      "calib/avg_num_step_conf": 5.734375,
      "calib/ece": 0.30592903225806467,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.3387096774193548,
      "calib/gap": 0.020396238244514198,
      "calib/mean_conf": 0.8256870967741935,
      "calib/mu_c": 0.8352272727272728,
      "calib/mu_w": 0.8148310344827586,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.2996790322580646,
      "calib/std_conf": 0.1362077261840034,
      "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": 2918.0,
      "completions/max_terminated_length": 2918.0,
      "completions/mean_length": 560.546875,
      "completions/mean_terminated_length": 567.1937255859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.1312,
      "grad_norm": 0.008393882773816586,
      "learning_rate": 2.138888888888889e-06,
      "loss": 0.0023,
      "num_tokens": 26995245.0,
      "reward": 1.6040575504302979,
      "reward_std": 0.5801326632499695,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.6302156448364258,
      "rewards/format_reward_step": 0.9609375,
      "rewards/stepwise_brier_reward": 0.7703891396522522,
      "step": 123
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5579488226487002,
      "calib/avg_num_step_conf": 5.5546875,
      "calib/ece": 0.2084689243027889,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.3545816733067729,
      "calib/gap": 0.055939015924867386,
      "calib/mean_conf": 0.8098577689243027,
      "calib/mu_c": 0.83058417721519,
      "calib/mu_w": 0.7746451612903226,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.19442231075697214,
      "calib/std_conf": 0.18619617607636454,
      "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": 2009.0,
      "completions/max_terminated_length": 2009.0,
      "completions/mean_length": 512.8046875,
      "completions/mean_terminated_length": 516.842529296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 138.0,
      "epoch": 0.13226666666666667,
      "grad_norm": 0.008557788096368313,
      "learning_rate": 2.1111111111111114e-06,
      "loss": 0.0473,
      "num_tokens": 27233339.0,
      "reward": 1.790103554725647,
      "reward_std": 0.4638766944408417,
      "rewards/accuracy_reward_step": 0.6171875,
      "rewards/final_brier_reward_step": 0.7087574601173401,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.795406699180603,
      "step": 124
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4478877679697352,
      "calib/avg_num_step_conf": 5.48046875,
      "calib/ece": 0.2940079365079366,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.26587301587301587,
      "calib/gap": -0.019524590163934374,
      "calib/mean_conf": 0.8084523809523809,
      "calib/mu_c": 0.799,
      "calib/mu_w": 0.8185245901639344,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.2932936507936509,
      "calib/std_conf": 0.1268591336716374,
      "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": 2813.0,
      "completions/max_terminated_length": 2813.0,
      "completions/mean_length": 548.40625,
      "completions/mean_terminated_length": 550.556884765625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 171.0,
      "epoch": 0.13333333333333333,
      "grad_norm": 0.0076705822721123695,
      "learning_rate": 2.0833333333333334e-06,
      "loss": 0.0209,
      "num_tokens": 27478539.0,
      "reward": 1.6000088453292847,
      "reward_std": 0.5619467496871948,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.6191773414611816,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.788670539855957,
      "step": 125
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6466991869918699,
      "calib/avg_num_step_conf": 6.12109375,
      "calib/ece": 0.30657016129032255,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.29838709677419356,
      "calib/gap": 0.07166073495934966,
      "calib/mean_conf": 0.8026233870967743,
      "calib/mu_c": 0.8381647999999999,
      "calib/mu_w": 0.7665040650406503,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3025806451612903,
      "calib/std_conf": 0.1565124645190213,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2905.0,
      "completions/max_terminated_length": 2905.0,
      "completions/mean_length": 523.421875,
      "completions/mean_terminated_length": 529.6284790039062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 142.0,
      "epoch": 0.1344,
      "grad_norm": 0.007916949689388275,
      "learning_rate": 2.0555555555555555e-06,
      "loss": 0.0079,
      "num_tokens": 27717999.0,
      "reward": 1.579500675201416,
      "reward_std": 0.43629828095436096,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.6511855125427246,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.7996293306350708,
      "step": 126
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5780618686868686,
      "calib/avg_num_step_conf": 5.50390625,
      "calib/ece": 0.2651666666666667,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.23412698412698413,
      "calib/gap": 0.04630909090909119,
      "calib/mean_conf": 0.7713571428571429,
      "calib/mu_c": 0.7934090909090911,
      "calib/mu_w": 0.7470999999999999,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2563571428571429,
      "calib/std_conf": 0.15733015353353075,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2976.0,
      "completions/max_terminated_length": 2976.0,
      "completions/mean_length": 495.203125,
      "completions/mean_terminated_length": 499.10235595703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 56.0,
      "epoch": 0.13546666666666668,
      "grad_norm": 0.009778374806046486,
      "learning_rate": 2.027777777777778e-06,
      "loss": 0.0039,
      "num_tokens": 27948443.0,
      "reward": 1.6397393941879272,
      "reward_std": 0.42718207836151123,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.6768920421600342,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8195655345916748,
      "step": 127
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5970234113712375,
      "calib/avg_num_step_conf": 4.609375,
      "calib/ece": 0.2467755102040817,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.22040816326530613,
      "calib/gap": 0.04780936454849505,
      "calib/mean_conf": 0.7750204081632653,
      "calib/mu_c": 0.7974615384615384,
      "calib/mu_w": 0.7496521739130434,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.24559183673469395,
      "calib/std_conf": 0.14539916764154398,
      "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": 0.01953125,
      "completions/max_length": 2841.0,
      "completions/max_terminated_length": 2841.0,
      "completions/mean_length": 505.72265625,
      "completions/mean_terminated_length": 515.7968139648438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.13653333333333334,
      "grad_norm": 0.008472949266433716,
      "learning_rate": 2.0000000000000003e-06,
      "loss": -0.007,
      "num_tokens": 28184572.0,
      "reward": 1.6027201414108276,
      "reward_std": 0.5276203751564026,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.6623015403747559,
      "rewards/format_reward_step": 0.94921875,
      "rewards/stepwise_brier_reward": 0.8032665252685547,
      "step": 128
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5278201219512195,
      "calib/avg_num_step_conf": 5.40234375,
      "calib/ece": 0.17888888888888882,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.36507936507936506,
      "calib/gap": 0.013054323725055661,
      "calib/mean_conf": 0.8219047619047619,
      "calib/mu_c": 0.8264634146341465,
      "calib/mu_w": 0.8134090909090909,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.17499999999999993,
      "calib/std_conf": 0.13017213442399592,
      "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": 2501.0,
      "completions/max_terminated_length": 2501.0,
      "completions/mean_length": 467.1953125,
      "completions/mean_terminated_length": 467.1953125,
      "completions/min_length": 138.0,
      "completions/min_terminated_length": 138.0,
      "epoch": 0.1376,
      "grad_norm": 0.009524206630885601,
      "learning_rate": 1.9722222222222224e-06,
      "loss": 0.0212,
      "num_tokens": 28406558.0,
      "reward": 1.8328227996826172,
      "reward_std": 0.3777114748954773,
      "rewards/accuracy_reward_step": 0.640625,
      "rewards/final_brier_reward_step": 0.7171855568885803,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8094183206558228,
      "step": 129
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5046078159645232,
      "calib/avg_num_step_conf": 5.12109375,
      "calib/ece": 0.192063492063492,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.24603174603174602,
      "calib/gap": -0.0006735033259424039,
      "calib/mean_conf": 0.8078571428571429,
      "calib/mu_c": 0.8076219512195121,
      "calib/mu_w": 0.8082954545454545,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.17456349206349198,
      "calib/std_conf": 0.1238253195052082,
      "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": 1851.0,
      "completions/max_terminated_length": 1851.0,
      "completions/mean_length": 447.87890625,
      "completions/mean_terminated_length": 449.63531494140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.13866666666666666,
      "grad_norm": 0.008591176941990852,
      "learning_rate": 1.944444444444445e-06,
      "loss": -0.0049,
      "num_tokens": 28626503.0,
      "reward": 1.8362488746643066,
      "reward_std": 0.30961671471595764,
      "rewards/accuracy_reward_step": 0.640625,
      "rewards/final_brier_reward_step": 0.7199695110321045,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8203378915786743,
      "step": 130
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6358108108108108,
      "calib/avg_num_step_conf": 5.5859375,
      "calib/ece": 0.39023715415019755,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.25691699604743085,
      "calib/gap": 0.05418211068211065,
      "calib/mean_conf": 0.8052569169960475,
      "calib/mu_c": 0.8369523809523809,
      "calib/mu_w": 0.7827702702702702,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.39023715415019755,
      "calib/std_conf": 0.1291521194639739,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3063.0,
      "completions/max_terminated_length": 3063.0,
      "completions/mean_length": 477.203125,
      "completions/mean_terminated_length": 477.203125,
      "completions/min_length": 200.0,
      "completions/min_terminated_length": 200.0,
      "epoch": 0.13973333333333332,
      "grad_norm": 0.008615897037088871,
      "learning_rate": 1.916666666666667e-06,
      "loss": 0.0548,
      "num_tokens": 28854875.0,
      "reward": 1.4560118913650513,
      "reward_std": 0.2913942337036133,
      "rewards/accuracy_reward_step": 0.41015625,
      "rewards/final_brier_reward_step": 0.6073628664016724,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.7791846990585327,
      "step": 131
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6040815022994379,
      "calib/avg_num_step_conf": 5.65234375,
      "calib/ece": 0.23422352941176475,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.34509803921568627,
      "calib/gap": 0.05555850792028616,
      "calib/mean_conf": 0.8148745098039216,
      "calib/mu_c": 0.8373157894736842,
      "calib/mu_w": 0.7817572815533981,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.22650980392156866,
      "calib/std_conf": 0.16289951172080705,
      "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": 2655.0,
      "completions/max_terminated_length": 2655.0,
      "completions/mean_length": 519.4765625,
      "completions/mean_terminated_length": 519.4765625,
      "completions/min_length": 152.0,
      "completions/min_terminated_length": 152.0,
      "epoch": 0.1408,
      "grad_norm": 0.008835425600409508,
      "learning_rate": 1.888888888888889e-06,
      "loss": 0.0067,
      "num_tokens": 29093453.0,
      "reward": 1.7685256004333496,
      "reward_std": 0.47421741485595703,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.7053423523902893,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8218851685523987,
      "step": 132
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6028761061946903,
      "calib/avg_num_step_conf": 5.640625,
      "calib/ece": 0.3680660079051383,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.3241106719367589,
      "calib/gap": 0.05708475979772443,
      "calib/mean_conf": 0.8103584980237154,
      "calib/mu_c": 0.8419469026548673,
      "calib/mu_w": 0.7848621428571428,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.3658920948616601,
      "calib/std_conf": 0.15401539851957424,
      "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": 1438.0,
      "completions/max_terminated_length": 1438.0,
      "completions/mean_length": 547.41015625,
      "completions/mean_terminated_length": 551.720458984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 86.0,
      "epoch": 0.14186666666666667,
      "grad_norm": 0.008469844236969948,
      "learning_rate": 1.8611111111111113e-06,
      "loss": 0.0036,
      "num_tokens": 29339934.0,
      "reward": 1.5014642477035522,
      "reward_std": 0.5480656027793884,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.6135154962539673,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.7829661965370178,
      "step": 133
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.7087638608870968,
      "calib/avg_num_step_conf": 5.6328125,
      "calib/ece": 0.3018650793650794,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.25793650793650796,
      "calib/gap": 0.10467489919354833,
      "calib/mean_conf": 0.7939285714285714,
      "calib/mu_c": 0.8470967741935483,
      "calib/mu_w": 0.742421875,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.3018650793650794,
      "calib/std_conf": 0.15239894797013406,
      "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": 2743.0,
      "completions/max_terminated_length": 2743.0,
      "completions/mean_length": 570.57421875,
      "completions/mean_terminated_length": 572.8118286132812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.14293333333333333,
      "grad_norm": 0.009082007221877575,
      "learning_rate": 1.8333333333333333e-06,
      "loss": 0.0301,
      "num_tokens": 29594953.0,
      "reward": 1.5818350315093994,
      "reward_std": 0.5352371335029602,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.6704590320587158,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.79750657081604,
      "step": 134
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.607094457243529,
      "calib/avg_num_step_conf": 5.546875,
      "calib/ece": 0.20537051792828687,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.3346613545816733,
      "calib/gap": 0.056958056647242516,
      "calib/mean_conf": 0.8163824701195219,
      "calib/mu_c": 0.8377133757961784,
      "calib/mu_w": 0.7807553191489359,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.19812749003984065,
      "calib/std_conf": 0.14917805723507854,
      "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": 2341.0,
      "completions/max_terminated_length": 2341.0,
      "completions/mean_length": 534.6171875,
      "completions/mean_terminated_length": 538.8267822265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.144,
      "grad_norm": 0.008807032369077206,
      "learning_rate": 1.8055555555555557e-06,
      "loss": -0.0138,
      "num_tokens": 29837695.0,
      "reward": 1.7926582098007202,
      "reward_std": 0.4692609906196594,
      "rewards/accuracy_reward_step": 0.61328125,
      "rewards/final_brier_reward_step": 0.715583086013794,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8222374320030212,
      "step": 135
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.595045632333768,
      "calib/avg_num_step_conf": 5.4609375,
      "calib/ece": 0.26666532258064524,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.23387096774193547,
      "calib/gap": 0.06105541069100384,
      "calib/mean_conf": 0.7894879032258065,
      "calib/mu_c": 0.8185384615384615,
      "calib/mu_w": 0.7574830508474577,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.26597983870967745,
      "calib/std_conf": 0.15299119682282689,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 1838.0,
      "completions/max_terminated_length": 1838.0,
      "completions/mean_length": 500.43359375,
      "completions/mean_terminated_length": 504.3740234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 31.0,
      "epoch": 0.14506666666666668,
      "grad_norm": 0.008869992569088936,
      "learning_rate": 1.777777777777778e-06,
      "loss": 0.015,
      "num_tokens": 30074294.0,
      "reward": 1.6056060791015625,
      "reward_std": 0.4623328447341919,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.6618715524673462,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.7839899659156799,
      "step": 136
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5775341914722446,
      "calib/avg_num_step_conf": 5.55078125,
      "calib/ece": 0.2938632653061225,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.3183673469387755,
      "calib/gap": 0.02456526548672555,
      "calib/mean_conf": 0.8185448979591837,
      "calib/mu_c": 0.8298749999999999,
      "calib/mu_w": 0.8053097345132744,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.9765625,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.2868163265306123,
      "calib/std_conf": 0.1390525301727321,
      "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": 528.796875,
      "completions/mean_terminated_length": 530.87060546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 183.0,
      "epoch": 0.14613333333333334,
      "grad_norm": 0.008724290877580643,
      "learning_rate": 1.75e-06,
      "loss": 0.0638,
      "num_tokens": 30316650.0,
      "reward": 1.6062359809875488,
      "reward_std": 0.51213139295578,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.637484073638916,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.7796474099159241,
      "step": 137
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5347516365036581,
      "calib/avg_num_step_conf": 5.05859375,
      "calib/ece": 0.2533992094861661,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.31225296442687744,
      "calib/gap": 0.015051341291233422,
      "calib/mean_conf": 0.807707509881423,
      "calib/mu_c": 0.8140136054421768,
      "calib/mu_w": 0.7989622641509434,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.24003952569169967,
      "calib/std_conf": 0.14094080213496563,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2911.0,
      "completions/max_terminated_length": 2911.0,
      "completions/mean_length": 495.41796875,
      "completions/mean_terminated_length": 495.41796875,
      "completions/min_length": 148.0,
      "completions/min_terminated_length": 148.0,
      "epoch": 0.1472,
      "grad_norm": 0.00900115817785263,
      "learning_rate": 1.7222222222222224e-06,
      "loss": 0.0096,
      "num_tokens": 30547813.0,
      "reward": 1.728182077407837,
      "reward_std": 0.42940008640289307,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.6845284700393677,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8063244223594666,
      "step": 138
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5773356401384082,
      "calib/avg_num_step_conf": 5.1875,
      "calib/ece": 0.12678431372549015,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.21568627450980393,
      "calib/gap": 0.03629411764705892,
      "calib/mean_conf": 0.7891372549019608,
      "calib/mu_c": 0.8012352941176472,
      "calib/mu_w": 0.7649411764705882,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.12462745098039213,
      "calib/std_conf": 0.12557960050522804,
      "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": 1942.0,
      "completions/max_terminated_length": 1942.0,
      "completions/mean_length": 484.1015625,
      "completions/mean_terminated_length": 486.0000305175781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 142.0,
      "epoch": 0.14826666666666666,
      "grad_norm": 0.010187380015850067,
      "learning_rate": 1.6944444444444446e-06,
      "loss": 0.0422,
      "num_tokens": 30774839.0,
      "reward": 1.896140456199646,
      "reward_std": 0.46494248509407043,
      "rewards/accuracy_reward_step": 0.6640625,
      "rewards/final_brier_reward_step": 0.7563285231590271,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8594832420349121,
      "step": 139
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6018867924528303,
      "calib/avg_num_step_conf": 4.6640625,
      "calib/ece": 0.15055118110236215,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.18110236220472442,
      "calib/gap": 0.04990268123138031,
      "calib/mean_conf": 0.7694488188976378,
      "calib/mu_c": 0.7881132075471698,
      "calib/mu_w": 0.7382105263157895,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.147007874015748,
      "calib/std_conf": 0.14089034707949227,
      "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": 996.0,
      "completions/max_terminated_length": 996.0,
      "completions/mean_length": 466.1328125,
      "completions/mean_terminated_length": 467.9608154296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.14933333333333335,
      "grad_norm": 0.009409532882273197,
      "learning_rate": 1.6666666666666667e-06,
      "loss": 0.0132,
      "num_tokens": 30999185.0,
      "reward": 1.8150107860565186,
      "reward_std": 0.4789872169494629,
      "rewards/accuracy_reward_step": 0.62109375,
      "rewards/final_brier_reward_step": 0.7314562797546387,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8410866260528564,
      "step": 140
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6037648893930799,
      "calib/avg_num_step_conf": 4.6796875,
      "calib/ece": 0.14019999999999996,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.176,
      "calib/gap": 0.045292115711854875,
      "calib/mean_conf": 0.7682,
      "calib/mu_c": 0.7837804878048782,
      "calib/mu_w": 0.7384883720930233,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.12619999999999998,
      "calib/std_conf": 0.14708487345747012,
      "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": 1832.0,
      "completions/max_terminated_length": 1832.0,
      "completions/mean_length": 540.84765625,
      "completions/mean_terminated_length": 540.84765625,
      "completions/min_length": 213.0,
      "completions/min_terminated_length": 213.0,
      "epoch": 0.1504,
      "grad_norm": 0.00863255001604557,
      "learning_rate": 1.638888888888889e-06,
      "loss": 0.0139,
      "num_tokens": 31244738.0,
      "reward": 1.8233741521835327,
      "reward_std": 0.3620089292526245,
      "rewards/accuracy_reward_step": 0.640625,
      "rewards/final_brier_reward_step": 0.7274668216705322,
      "rewards/format_reward_step": 0.953125,
      "rewards/stepwise_brier_reward": 0.8160301446914673,
      "step": 141
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.666687437686939,
      "calib/avg_num_step_conf": 5.15234375,
      "calib/ece": 0.2458255905511811,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.2047244094488189,
      "calib/gap": 0.09083346211365928,
      "calib/mean_conf": 0.7734657480314961,
      "calib/mu_c": 0.8156639705882354,
      "calib/mu_w": 0.7248305084745761,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.24192913385826773,
      "calib/std_conf": 0.16074253380253575,
      "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": 1350.0,
      "completions/max_terminated_length": 1350.0,
      "completions/mean_length": 540.95703125,
      "completions/mean_terminated_length": 543.0784912109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.15146666666666667,
      "grad_norm": 0.008594821207225323,
      "learning_rate": 1.6111111111111113e-06,
      "loss": 0.0318,
      "num_tokens": 31488383.0,
      "reward": 1.6679813861846924,
      "reward_std": 0.3934059739112854,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.7005878686904907,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.8150873780250549,
      "step": 142
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6364356435643563,
      "calib/avg_num_step_conf": 5.18359375,
      "calib/ece": 0.1970916334661355,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.199203187250996,
      "calib/gap": 0.057742574257425794,
      "calib/mean_conf": 0.7787649402390437,
      "calib/mu_c": 0.8019999999999999,
      "calib/mu_w": 0.7442574257425741,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.18912350597609567,
      "calib/std_conf": 0.14385856726796234,
      "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": 2608.0,
      "completions/max_terminated_length": 2608.0,
      "completions/mean_length": 526.29296875,
      "completions/mean_terminated_length": 528.3568725585938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 60.0,
      "epoch": 0.15253333333333333,
      "grad_norm": 0.008822314441204071,
      "learning_rate": 1.5833333333333333e-06,
      "loss": 0.0228,
      "num_tokens": 31730450.0,
      "reward": 1.7379977703094482,
      "reward_std": 0.318717360496521,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.7136461138725281,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.7930324673652649,
      "step": 143
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6373626373626374,
      "calib/avg_num_step_conf": 4.76171875,
      "calib/ece": 0.058292857142857125,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.19047619047619047,
      "calib/gap": 0.07005604395604392,
      "calib/mean_conf": 0.7684531746031745,
      "calib/mu_c": 0.7879131868131867,
      "calib/mu_w": 0.7178571428571427,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.052261904761904746,
      "calib/std_conf": 0.147642762327517,
      "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": 0.00390625,
      "completions/max_length": 2846.0,
      "completions/max_terminated_length": 2846.0,
      "completions/mean_length": 528.30859375,
      "completions/mean_terminated_length": 530.3804321289062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.1536,
      "grad_norm": 0.009168135933578014,
      "learning_rate": 1.5555555555555558e-06,
      "loss": 0.0112,
      "num_tokens": 31969825.0,
      "reward": 1.9473841190338135,
      "reward_std": 0.4583970606327057,
      "rewards/accuracy_reward_step": 0.7109375,
      "rewards/final_brier_reward_step": 0.77021324634552,
      "rewards/format_reward_step": 0.9609375,
      "rewards/stepwise_brier_reward": 0.8318236470222473,
      "step": 144
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5327559167379526,
      "calib/avg_num_step_conf": 4.66015625,
      "calib/ece": 0.14095617529880475,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.21115537848605578,
      "calib/gap": 0.020917450812660388,
      "calib/mean_conf": 0.7571314741035856,
      "calib/mu_c": 0.7641317365269461,
      "calib/mu_w": 0.7432142857142857,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.11637450199203184,
      "calib/std_conf": 0.15597188761604533,
      "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": 2701.0,
      "completions/max_terminated_length": 2701.0,
      "completions/mean_length": 501.65625,
      "completions/mean_terminated_length": 501.65625,
      "completions/min_length": 149.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.15466666666666667,
      "grad_norm": 0.009162410162389278,
      "learning_rate": 1.527777777777778e-06,
      "loss": 0.0618,
      "num_tokens": 32200953.0,
      "reward": 1.860499620437622,
      "reward_std": 0.4525335431098938,
      "rewards/accuracy_reward_step": 0.65234375,
      "rewards/final_brier_reward_step": 0.7391742467880249,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8278242945671082,
      "step": 145
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5856652949245542,
      "calib/avg_num_step_conf": 4.390625,
      "calib/ece": 0.30004115226337447,
      "calib/final_conf_rate": 0.94921875,
      "calib/format_rate": 0.92578125,
      "calib/frac_conf_gt_0.9": 0.17695473251028807,
      "calib/gap": 0.054592592592592526,
      "calib/mean_conf": 0.7444855967078189,
      "calib/mu_c": 0.7748148148148147,
      "calib/mu_w": 0.7202222222222222,
      "calib/nonempty_final_conf_rate": 0.94921875,
      "calib/nonempty_reasoning_rate": 0.96875,
      "calib/nonempty_step_conf_rate": 0.9453125,
      "calib/pce": 0.30004115226337447,
      "calib/std_conf": 0.1555318857598609,
      "calib/step_conf_rate": 0.9453125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 2179.0,
      "completions/max_terminated_length": 2179.0,
      "completions/mean_length": 513.6171875,
      "completions/mean_terminated_length": 521.7698974609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.15573333333333333,
      "grad_norm": 0.00974523089826107,
      "learning_rate": 1.5e-06,
      "loss": -0.0017,
      "num_tokens": 32439655.0,
      "reward": 1.4424173831939697,
      "reward_std": 0.4637851119041443,
      "rewards/accuracy_reward_step": 0.42578125,
      "rewards/final_brier_reward_step": 0.6153711080551147,
      "rewards/format_reward_step": 0.92578125,
      "rewards/stepwise_brier_reward": 0.7480484843254089,
      "step": 146
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6449354838709678,
      "calib/avg_num_step_conf": 4.2890625,
      "calib/ece": 0.2845381526104418,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.9375,
      "calib/frac_conf_gt_0.9": 0.21686746987951808,
      "calib/gap": 0.06404387096774233,
      "calib/mean_conf": 0.7633333333333333,
      "calib/mu_c": 0.7954838709677422,
      "calib/mu_w": 0.7314399999999999,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.953125,
      "calib/pce": 0.27493975903614465,
      "calib/std_conf": 0.15510759749456335,
      "calib/step_conf_rate": 0.953125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1756.0,
      "completions/max_terminated_length": 1756.0,
      "completions/mean_length": 511.671875,
      "completions/mean_terminated_length": 513.678466796875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.1568,
      "grad_norm": 0.008800083771348,
      "learning_rate": 1.4722222222222225e-06,
      "loss": 0.0058,
      "num_tokens": 32674323.0,
      "reward": 1.5476961135864258,
      "reward_std": 0.46623843908309937,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.6436504125595093,
      "rewards/format_reward_step": 0.9375,
      "rewards/stepwise_brier_reward": 0.7658840417861938,
      "step": 147
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6432334897149045,
      "calib/avg_num_step_conf": 4.046875,
      "calib/ece": 0.13487903225806452,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.9375,
      "calib/frac_conf_gt_0.9": 0.16129032258064516,
      "calib/gap": 0.06621364128473473,
      "calib/mean_conf": 0.7796370967741935,
      "calib/mu_c": 0.8023312883435583,
      "calib/mu_w": 0.7361176470588235,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.12862903225806452,
      "calib/std_conf": 0.13306964188563553,
      "calib/step_conf_rate": 0.96484375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 2642.0,
      "completions/max_terminated_length": 2642.0,
      "completions/mean_length": 525.359375,
      "completions/mean_terminated_length": 529.4960327148438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 117.0,
      "epoch": 0.15786666666666666,
      "grad_norm": 0.008699139580130577,
      "learning_rate": 1.4444444444444445e-06,
      "loss": 0.0429,
      "num_tokens": 32913927.0,
      "reward": 1.7964141368865967,
      "reward_std": 0.5030028223991394,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.7241019606590271,
      "rewards/format_reward_step": 0.9375,
      "rewards/stepwise_brier_reward": 0.7662421464920044,
      "step": 148
    },
    {
      "calib/answer_extract_rate": 0.9375,
      "calib/auroc": 0.6462651956995292,
      "calib/avg_num_step_conf": 4.3125,
      "calib/ece": 0.2471079166666667,
      "calib/final_conf_rate": 0.9375,
      "calib/format_rate": 0.93359375,
      "calib/frac_conf_gt_0.9": 0.21666666666666667,
      "calib/gap": 0.0641043285784555,
      "calib/mean_conf": 0.7912254166666666,
      "calib/mu_c": 0.8198052631578947,
      "calib/mu_w": 0.7557009345794392,
      "calib/nonempty_final_conf_rate": 0.9375,
      "calib/nonempty_reasoning_rate": 0.96875,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.24208333333333337,
      "calib/std_conf": 0.13939393242053158,
      "calib/step_conf_rate": 0.96484375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 2583.0,
      "completions/max_terminated_length": 2583.0,
      "completions/mean_length": 555.1015625,
      "completions/mean_terminated_length": 566.1593627929688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 146.0,
      "epoch": 0.15893333333333334,
      "grad_norm": 0.008257905952632427,
      "learning_rate": 1.4166666666666667e-06,
      "loss": 0.021,
      "num_tokens": 33160489.0,
      "reward": 1.6061928272247314,
      "reward_std": 0.3906460106372833,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.6626788973808289,
      "rewards/format_reward_step": 0.93359375,
      "rewards/stepwise_brier_reward": 0.7777170538902283,
      "step": 149
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.6704482758620689,
      "calib/avg_num_step_conf": 4.23828125,
      "calib/ece": 0.2226122448979591,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.3142857142857143,
      "calib/gap": 0.0738448275862067,
      "calib/mean_conf": 0.8062040816326531,
      "calib/mu_c": 0.8363448275862069,
      "calib/mu_w": 0.7625000000000002,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.2184897959183673,
      "calib/std_conf": 0.1353358603082756,
      "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": 0.0078125,
      "completions/max_length": 1885.0,
      "completions/max_terminated_length": 1885.0,
      "completions/mean_length": 484.81640625,
      "completions/mean_terminated_length": 488.63385009765625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 56.0,
      "epoch": 0.16,
      "grad_norm": 0.008547094650566578,
      "learning_rate": 1.3888888888888892e-06,
      "loss": -0.0373,
      "num_tokens": 33389562.0,
      "reward": 1.6901917457580566,
      "reward_std": 0.531447172164917,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.6897379159927368,
      "rewards/format_reward_step": 0.9453125,
      "rewards/stepwise_brier_reward": 0.7819663286209106,
      "step": 150
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5765810276679844,
      "calib/avg_num_step_conf": 4.12109375,
      "calib/ece": 0.3268825910931174,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.92578125,
      "calib/frac_conf_gt_0.9": 0.2793522267206478,
      "calib/gap": 0.05539393939393944,
      "calib/mean_conf": 0.7863967611336032,
      "calib/mu_c": 0.8160000000000001,
      "calib/mu_w": 0.7606060606060606,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.9453125,
      "calib/pce": 0.32384615384615384,
      "calib/std_conf": 0.16178524673895403,
      "calib/step_conf_rate": 0.9453125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 2880.0,
      "completions/max_terminated_length": 2880.0,
      "completions/mean_length": 540.63671875,
      "completions/mean_terminated_length": 547.0474243164062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 173.0,
      "epoch": 0.16106666666666666,
      "grad_norm": 0.00871346890926361,
      "learning_rate": 1.3611111111111112e-06,
      "loss": 0.0432,
      "num_tokens": 33634989.0,
      "reward": 1.469602108001709,
      "reward_std": 0.5533789396286011,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6017488241195679,
      "rewards/format_reward_step": 0.92578125,
      "rewards/stepwise_brier_reward": 0.7063466310501099,
      "step": 151
    },
    {
      "calib/answer_extract_rate": 0.9453125,
      "calib/auroc": 0.6225941989699106,
      "calib/avg_num_step_conf": 3.84765625,
      "calib/ece": 0.27156378600823045,
      "calib/final_conf_rate": 0.94921875,
      "calib/format_rate": 0.9296875,
      "calib/frac_conf_gt_0.9": 0.20164609053497942,
      "calib/gap": 0.056595960965030945,
      "calib/mean_conf": 0.7793004115226336,
      "calib/mu_c": 0.807016129032258,
      "calib/mu_w": 0.750420168067227,
      "calib/nonempty_final_conf_rate": 0.94921875,
      "calib/nonempty_reasoning_rate": 0.96875,
      "calib/nonempty_step_conf_rate": 0.953125,
      "calib/pce": 0.27028806584362136,
      "calib/std_conf": 0.13491318760273607,
      "calib/step_conf_rate": 0.953125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 2934.0,
      "completions/max_terminated_length": 2934.0,
      "completions/mean_length": 512.609375,
      "completions/mean_terminated_length": 516.6456909179688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 185.0,
      "epoch": 0.16213333333333332,
      "grad_norm": 0.009242072701454163,
      "learning_rate": 1.3333333333333334e-06,
      "loss": 0.0205,
      "num_tokens": 33871609.0,
      "reward": 1.538311243057251,
      "reward_std": 0.5286429524421692,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.6414843797683716,
      "rewards/format_reward_step": 0.9296875,
      "rewards/stepwise_brier_reward": 0.7461355924606323,
      "step": 152
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5912703804347826,
      "calib/avg_num_step_conf": 3.7578125,
      "calib/ece": 0.14952380952380953,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.17063492063492064,
      "calib/gap": 0.05000815217391308,
      "calib/mean_conf": 0.7755555555555556,
      "calib/mu_c": 0.7938124999999999,
      "calib/mu_w": 0.7438043478260868,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.953125,
      "calib/pce": 0.14507936507936509,
      "calib/std_conf": 0.14745220132464965,
      "calib/step_conf_rate": 0.953125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 1603.0,
      "completions/max_terminated_length": 1603.0,
      "completions/mean_length": 486.48828125,
      "completions/mean_terminated_length": 490.31890869140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 160.0,
      "epoch": 0.1632,
      "grad_norm": 0.009091148152947426,
      "learning_rate": 1.3055555555555556e-06,
      "loss": -0.0206,
      "num_tokens": 34103470.0,
      "reward": 1.777563452720642,
      "reward_std": 0.49572986364364624,
      "rewards/accuracy_reward_step": 0.625,
      "rewards/final_brier_reward_step": 0.7101125121116638,
      "rewards/format_reward_step": 0.9453125,
      "rewards/stepwise_brier_reward": 0.7595163583755493,
      "step": 153
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6654799745708836,
      "calib/avg_num_step_conf": 3.671875,
      "calib/ece": 0.21308300395256913,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.18972332015810275,
      "calib/gap": 0.07676923076923092,
      "calib/mean_conf": 0.778300395256917,
      "calib/mu_c": 0.8116783216783217,
      "calib/mu_w": 0.7349090909090907,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.21308300395256913,
      "calib/std_conf": 0.13480140816813654,
      "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": 1763.0,
      "completions/max_terminated_length": 1763.0,
      "completions/mean_length": 476.5859375,
      "completions/mean_terminated_length": 480.3385925292969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.16426666666666667,
      "grad_norm": 0.010077430866658688,
      "learning_rate": 1.2777777777777779e-06,
      "loss": -0.0233,
      "num_tokens": 34329916.0,
      "reward": 1.7182092666625977,
      "reward_std": 0.44295963644981384,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.7198730111122131,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8248393535614014,
      "step": 154
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5980925578486553,
      "calib/avg_num_step_conf": 4.4140625,
      "calib/ece": 0.26960474308300403,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.16205533596837945,
      "calib/gap": 0.04396185115697304,
      "calib/mean_conf": 0.74098814229249,
      "calib/mu_c": 0.7635772357723577,
      "calib/mu_w": 0.7196153846153847,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.26221343873517794,
      "calib/std_conf": 0.15352064614312008,
      "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": 2849.0,
      "completions/max_terminated_length": 2849.0,
      "completions/mean_length": 468.8984375,
      "completions/mean_terminated_length": 470.7372741699219,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 30.0,
      "epoch": 0.16533333333333333,
      "grad_norm": 0.010738848708570004,
      "learning_rate": 1.25e-06,
      "loss": 0.0526,
      "num_tokens": 34557170.0,
      "reward": 1.5779985189437866,
      "reward_std": 0.45197001099586487,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.6746238470077515,
      "rewards/format_reward_step": 0.984375,
      "rewards/stepwise_brier_reward": 0.7858077883720398,
      "step": 155
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5769230769230769,
      "calib/avg_num_step_conf": 4.21484375,
      "calib/ece": 0.18044176706827317,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.18875502008032127,
      "calib/gap": 0.04775265251989391,
      "calib/mean_conf": 0.7513654618473897,
      "calib/mu_c": 0.7713103448275862,
      "calib/mu_w": 0.7235576923076923,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.17473895582329327,
      "calib/std_conf": 0.1617442046033757,
      "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": 2301.0,
      "completions/max_terminated_length": 2301.0,
      "completions/mean_length": 524.9921875,
      "completions/mean_terminated_length": 527.051025390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 147.0,
      "epoch": 0.1664,
      "grad_norm": 0.008976864628493786,
      "learning_rate": 1.2222222222222223e-06,
      "loss": 0.0153,
      "num_tokens": 34796328.0,
      "reward": 1.7078721523284912,
      "reward_std": 0.4199150800704956,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.7016226053237915,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.7939286231994629,
      "step": 156
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.7503644314868805,
      "calib/avg_num_step_conf": 4.50390625,
      "calib/ece": 0.16372698412698417,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.25,
      "calib/gap": 0.14685120593692047,
      "calib/mean_conf": 0.7669873015873016,
      "calib/mu_c": 0.824096103896104,
      "calib/mu_w": 0.6772448979591835,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.15980158730158733,
      "calib/std_conf": 0.1715093754594668,
      "calib/step_conf_rate": 0.98046875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1270.0,
      "completions/max_terminated_length": 1270.0,
      "completions/mean_length": 520.0,
      "completions/mean_terminated_length": 522.0392456054688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.16746666666666668,
      "grad_norm": 0.008186625316739082,
      "learning_rate": 1.1944444444444446e-06,
      "loss": 0.0133,
      "num_tokens": 35033176.0,
      "reward": 1.7840977907180786,
      "reward_std": 0.42261362075805664,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.7521061897277832,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.8139725923538208,
      "step": 157
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5396075823079481,
      "calib/avg_num_step_conf": 3.890625,
      "calib/ece": 0.1828571428571429,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.20634920634920634,
      "calib/gap": 0.01812304622547367,
      "calib/mean_conf": 0.7465079365079365,
      "calib/mu_c": 0.7534838709677418,
      "calib/mu_w": 0.7353608247422682,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.95703125,
      "calib/pce": 0.1571428571428572,
      "calib/std_conf": 0.17592529736928728,
      "calib/step_conf_rate": 0.95703125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 1633.0,
      "completions/max_terminated_length": 1633.0,
      "completions/mean_length": 486.40625,
      "completions/mean_terminated_length": 492.1739196777344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 148.0,
      "epoch": 0.16853333333333334,
      "grad_norm": 0.00946896430104971,
      "learning_rate": 1.1666666666666668e-06,
      "loss": 0.0187,
      "num_tokens": 35262936.0,
      "reward": 1.749363660812378,
      "reward_std": 0.4539514183998108,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.6959766149520874,
      "rewards/format_reward_step": 0.953125,
      "rewards/stepwise_brier_reward": 0.7624151706695557,
      "step": 158
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5826985514485514,
      "calib/avg_num_step_conf": 4.1328125,
      "calib/ece": 0.17113725490196077,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.1411764705882353,
      "calib/gap": 0.04818119380619379,
      "calib/mean_conf": 0.7312156862745098,
      "calib/mu_c": 0.7523776223776224,
      "calib/mu_w": 0.7041964285714286,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.17078431372549022,
      "calib/std_conf": 0.15794418395160004,
      "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": 1090.0,
      "completions/max_terminated_length": 1090.0,
      "completions/mean_length": 449.3671875,
      "completions/mean_terminated_length": 451.1294250488281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 182.0,
      "epoch": 0.1696,
      "grad_norm": 0.010996346361935139,
      "learning_rate": 1.138888888888889e-06,
      "loss": 0.0079,
      "num_tokens": 35482758.0,
      "reward": 1.725920557975769,
      "reward_std": 0.3647944927215576,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.7167355418205261,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8510092496871948,
      "step": 159
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6352752880921895,
      "calib/avg_num_step_conf": 4.265625,
      "calib/ece": 0.21239087301587306,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.15079365079365079,
      "calib/gap": 0.06363412291933424,
      "calib/mean_conf": 0.7571527777777778,
      "calib/mu_c": 0.7849295774647888,
      "calib/mu_w": 0.7212954545454545,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.20302579365079368,
      "calib/std_conf": 0.14683990537229405,
      "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": 3005.0,
      "completions/max_terminated_length": 3005.0,
      "completions/mean_length": 506.578125,
      "completions/mean_terminated_length": 510.5669250488281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 175.0,
      "epoch": 0.17066666666666666,
      "grad_norm": 0.009727086871862411,
      "learning_rate": 1.111111111111111e-06,
      "loss": 0.0081,
      "num_tokens": 35717282.0,
      "reward": 1.7033100128173828,
      "reward_std": 0.49738264083862305,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.7124208807945251,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8117566108703613,
      "step": 160
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6746942473199455,
      "calib/avg_num_step_conf": 4.05859375,
      "calib/ece": 0.12596837944664024,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.1422924901185771,
      "calib/gap": 0.0892246715989734,
      "calib/mean_conf": 0.7208300395256917,
      "calib/mu_c": 0.7469273743016761,
      "calib/mu_w": 0.6577027027027027,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.0696442687747035,
      "calib/std_conf": 0.18357802717660351,
      "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": 1716.0,
      "completions/max_terminated_length": 1716.0,
      "completions/mean_length": 447.89453125,
      "completions/mean_terminated_length": 449.6510009765625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 37.0,
      "epoch": 0.17173333333333332,
      "grad_norm": 0.009979375638067722,
      "learning_rate": 1.0833333333333335e-06,
      "loss": -0.0135,
      "num_tokens": 35935863.0,
      "reward": 1.9391924142837524,
      "reward_std": 0.4121323525905609,
      "rewards/accuracy_reward_step": 0.69921875,
      "rewards/final_brier_reward_step": 0.7794542908668518,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.8288778066635132,
      "step": 161
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6850654177817425,
      "calib/avg_num_step_conf": 4.2578125,
      "calib/ece": 0.06758893280632408,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.21739130434782608,
      "calib/gap": 0.10630315194766571,
      "calib/mean_conf": 0.7413438735177865,
      "calib/mu_c": 0.7732768361581921,
      "calib/mu_w": 0.6669736842105264,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.05466403162055331,
      "calib/std_conf": 0.16049507254825174,
      "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": 1450.0,
      "completions/max_terminated_length": 1450.0,
      "completions/mean_length": 437.6015625,
      "completions/mean_terminated_length": 442.79052734375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 160.0,
      "epoch": 0.1728,
      "grad_norm": 0.010758143849670887,
      "learning_rate": 1.0555555555555557e-06,
      "loss": -0.0304,
      "num_tokens": 36152033.0,
      "reward": 1.9452869892120361,
      "reward_std": 0.47155964374542236,
      "rewards/accuracy_reward_step": 0.69140625,
      "rewards/final_brier_reward_step": 0.7935965061187744,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8781765699386597,
      "step": 162
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.709619341563786,
      "calib/avg_num_step_conf": 3.95703125,
      "calib/ece": 0.16178571428571428,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.1111111111111111,
      "calib/gap": 0.11678240740740764,
      "calib/mean_conf": 0.7332142857142856,
      "calib/mu_c": 0.783263888888889,
      "calib/mu_w": 0.6664814814814813,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.16178571428571428,
      "calib/std_conf": 0.154088415514984,
      "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": 1832.0,
      "completions/max_terminated_length": 1832.0,
      "completions/mean_length": 490.46875,
      "completions/mean_terminated_length": 494.3307189941406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 140.0,
      "epoch": 0.17386666666666667,
      "grad_norm": 0.008721891790628433,
      "learning_rate": 1.0277777777777777e-06,
      "loss": -0.0098,
      "num_tokens": 36382425.0,
      "reward": 1.71864652633667,
      "reward_std": 0.40552768111228943,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.7435265779495239,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.8107466697692871,
      "step": 163
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6272925484160574,
      "calib/avg_num_step_conf": 3.7734375,
      "calib/ece": 0.22195219123505974,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.199203187250996,
      "calib/gap": 0.061930870847762054,
      "calib/mean_conf": 0.7621912350597609,
      "calib/mu_c": 0.7900724637681159,
      "calib/mu_w": 0.7281415929203539,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.21717131474103585,
      "calib/std_conf": 0.1554673563399058,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 2140.0,
      "completions/max_terminated_length": 2140.0,
      "completions/mean_length": 537.4609375,
      "completions/mean_terminated_length": 541.6929321289062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 219.0,
      "epoch": 0.17493333333333333,
      "grad_norm": 0.009260274469852448,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0175,
      "num_tokens": 36626151.0,
      "reward": 1.6682591438293457,
      "reward_std": 0.4575202763080597,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.6931565999984741,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.8001925945281982,
      "step": 164
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.605976266209934,
      "calib/avg_num_step_conf": 4.08203125,
      "calib/ece": 0.297186328125,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.1640625,
      "calib/gap": 0.0432713604110595,
      "calib/mean_conf": 0.752188671875,
      "calib/mu_c": 0.7748385245901639,
      "calib/mu_w": 0.7315671641791044,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.28640625000000003,
      "calib/std_conf": 0.1580191070471735,
      "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": 1269.0,
      "completions/max_terminated_length": 1269.0,
      "completions/mean_length": 496.58203125,
      "completions/mean_terminated_length": 498.5294494628906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.176,
      "grad_norm": 0.0088722612708807,
      "learning_rate": 9.722222222222224e-07,
      "loss": 0.0252,
      "num_tokens": 36858852.0,
      "reward": 1.5780149698257446,
      "reward_std": 0.3902446925640106,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.6691632866859436,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.799146294593811,
      "step": 165
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6497965768799102,
      "calib/avg_num_step_conf": 3.984375,
      "calib/ece": 0.1483200000000001,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.204,
      "calib/gap": 0.060657968574635235,
      "calib/mean_conf": 0.7619199999999999,
      "calib/mu_c": 0.7832716049382716,
      "calib/mu_w": 0.7226136363636364,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.96875,
      "calib/pce": 0.13112000000000007,
      "calib/std_conf": 0.15161369858954038,
      "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": 0.0078125,
      "completions/max_length": 1301.0,
      "completions/max_terminated_length": 1301.0,
      "completions/mean_length": 499.40625,
      "completions/mean_terminated_length": 503.3385925292969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 194.0,
      "epoch": 0.17706666666666668,
      "grad_norm": 0.008880876004695892,
      "learning_rate": 9.444444444444445e-07,
      "loss": 0.0227,
      "num_tokens": 37092884.0,
      "reward": 1.8140075206756592,
      "reward_std": 0.5116109251976013,
      "rewards/accuracy_reward_step": 0.6328125,
      "rewards/final_brier_reward_step": 0.7289254069328308,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.8161675930023193,
      "step": 166
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6138914263343338,
      "calib/avg_num_step_conf": 3.984375,
      "calib/ece": 0.19490196078431377,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.3137254901960784,
      "calib/gap": 0.052887250424115906,
      "calib/mean_conf": 0.8062745098039217,
      "calib/mu_c": 0.8263924050632911,
      "calib/mu_w": 0.7735051546391752,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.19078431372549026,
      "calib/std_conf": 0.1368892608731845,
      "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": 1249.0,
      "completions/max_terminated_length": 1249.0,
      "completions/mean_length": 471.6640625,
      "completions/mean_terminated_length": 473.5137634277344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 77.0,
      "epoch": 0.17813333333333334,
      "grad_norm": 0.010015462525188923,
      "learning_rate": 9.166666666666666e-07,
      "loss": -0.0172,
      "num_tokens": 37319238.0,
      "reward": 1.80796480178833,
      "reward_std": 0.3873631954193115,
      "rewards/accuracy_reward_step": 0.6171875,
      "rewards/final_brier_reward_step": 0.729062557220459,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.8152968883514404,
      "step": 167
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6985240098844586,
      "calib/avg_num_step_conf": 3.8046875,
      "calib/ece": 0.1361023622047244,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.2204724409448819,
      "calib/gap": 0.1040573031456623,
      "calib/mean_conf": 0.7620866141732283,
      "calib/mu_c": 0.8001863354037267,
      "calib/mu_w": 0.6961290322580644,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.13216535433070864,
      "calib/std_conf": 0.16048393957087956,
      "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": 2428.0,
      "completions/max_terminated_length": 2428.0,
      "completions/mean_length": 540.51953125,
      "completions/mean_terminated_length": 542.6392211914062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 158.0,
      "epoch": 0.1792,
      "grad_norm": 0.008988683111965656,
      "learning_rate": 8.88888888888889e-07,
      "loss": 0.0365,
      "num_tokens": 37562283.0,
      "reward": 1.8311870098114014,
      "reward_std": 0.49814075231552124,
      "rewards/accuracy_reward_step": 0.62890625,
      "rewards/final_brier_reward_step": 0.7569624781608582,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8334102034568787,
      "step": 168
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5962589459986988,
      "calib/avg_num_step_conf": 3.89453125,
      "calib/ece": 0.2535059760956175,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.30677290836653387,
      "calib/gap": 0.04688484059856868,
      "calib/mean_conf": 0.7648207171314743,
      "calib/mu_c": 0.7846206896551724,
      "calib/mu_w": 0.7377358490566037,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.22031872509960154,
      "calib/std_conf": 0.18912323320979038,
      "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": 2773.0,
      "completions/max_terminated_length": 2773.0,
      "completions/mean_length": 522.45703125,
      "completions/mean_terminated_length": 522.45703125,
      "completions/min_length": 182.0,
      "completions/min_terminated_length": 182.0,
      "epoch": 0.18026666666666666,
      "grad_norm": 0.00886106863617897,
      "learning_rate": 8.611111111111112e-07,
      "loss": 0.0092,
      "num_tokens": 37800216.0,
      "reward": 1.7088391780853271,
      "reward_std": 0.41909223794937134,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.6909284591674805,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.8006783723831177,
      "step": 169
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.711261777438248,
      "calib/avg_num_step_conf": 3.9296875,
      "calib/ece": 0.20894531249999995,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.4140625,
      "calib/gap": 0.11834861217214143,
      "calib/mean_conf": 0.8105078125,
      "calib/mu_c": 0.8576623376623376,
      "calib/mu_w": 0.7393137254901961,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.20894531249999995,
      "calib/std_conf": 0.1636764384432434,
      "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": 1050.0,
      "completions/max_terminated_length": 1050.0,
      "completions/mean_length": 499.76953125,
      "completions/mean_terminated_length": 501.72943115234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.18133333333333335,
      "grad_norm": 0.008470607921481133,
      "learning_rate": 8.333333333333333e-07,
      "loss": -0.004,
      "num_tokens": 38032309.0,
      "reward": 1.7914400100708008,
      "reward_std": 0.450106143951416,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.7427699565887451,
      "rewards/format_reward_step": 0.99609375,
      "rewards/stepwise_brier_reward": 0.8214274644851685,
      "step": 170
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6854213553388347,
      "calib/avg_num_step_conf": 3.8203125,
      "calib/ece": 0.2778656126482214,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.28063241106719367,
      "calib/gap": 0.095874593648412,
      "calib/mean_conf": 0.7833201581027668,
      "calib/mu_c": 0.8303100775193798,
      "calib/mu_w": 0.7344354838709678,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.27565217391304353,
      "calib/std_conf": 0.15994330847418,
      "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": 1302.0,
      "completions/max_terminated_length": 1302.0,
      "completions/mean_length": 482.8125,
      "completions/mean_terminated_length": 484.7059020996094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 159.0,
      "epoch": 0.1824,
      "grad_norm": 0.009204911068081856,
      "learning_rate": 8.055555555555557e-07,
      "loss": 0.0116,
      "num_tokens": 38262805.0,
      "reward": 1.6148879528045654,
      "reward_std": 0.4567514955997467,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.6848187446594238,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.7903579473495483,
      "step": 171
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5284090909090909,
      "calib/avg_num_step_conf": 3.828125,
      "calib/ece": 0.17785714285714288,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.38095238095238093,
      "calib/gap": 0.022935144124168638,
      "calib/mean_conf": 0.8225396825396825,
      "calib/mu_c": 0.8305487804878049,
      "calib/mu_w": 0.8076136363636363,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.17480158730158735,
      "calib/std_conf": 0.13261075938498024,
      "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": 2585.0,
      "completions/max_terminated_length": 2585.0,
      "completions/mean_length": 472.421875,
      "completions/mean_terminated_length": 474.2745361328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 184.0,
      "epoch": 0.18346666666666667,
      "grad_norm": 0.008253687061369419,
      "learning_rate": 7.777777777777779e-07,
      "loss": 0.0282,
      "num_tokens": 38487097.0,
      "reward": 1.8315176963806152,
      "reward_std": 0.5292471051216125,
      "rewards/accuracy_reward_step": 0.640625,
      "rewards/final_brier_reward_step": 0.7211250066757202,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.8002580404281616,
      "step": 172
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5893203211517165,
      "calib/avg_num_step_conf": 3.53125,
      "calib/ece": 0.18480314960629918,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.41732283464566927,
      "calib/gap": 0.05265227021040986,
      "calib/mean_conf": 0.8462204724409449,
      "calib/mu_c": 0.8640476190476192,
      "calib/mu_w": 0.8113953488372093,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.18480314960629918,
      "calib/std_conf": 0.11058846983125918,
      "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": 1404.0,
      "completions/max_terminated_length": 1404.0,
      "completions/mean_length": 503.09375,
      "completions/mean_terminated_length": 505.06671142578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 142.0,
      "epoch": 0.18453333333333333,
      "grad_norm": 0.009023896418511868,
      "learning_rate": 7.5e-07,
      "loss": -0.0156,
      "num_tokens": 38719049.0,
      "reward": 1.8713914155960083,
      "reward_std": 0.4353087544441223,
      "rewards/accuracy_reward_step": 0.65625,
      "rewards/final_brier_reward_step": 0.7473710775375366,
      "rewards/format_reward_step": 0.9921875,
      "rewards/stepwise_brier_reward": 0.816319465637207,
      "step": 173
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.49092269651223325,
      "calib/avg_num_step_conf": 3.98828125,
      "calib/ece": 0.290160642570281,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.39759036144578314,
      "calib/gap": 0.030691697032795284,
      "calib/mean_conf": 0.8165863453815261,
      "calib/mu_c": 0.8305147058823529,
      "calib/mu_w": 0.7998230088495576,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2802811244979919,
      "calib/std_conf": 0.16990664095573377,
      "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": 2592.0,
      "completions/max_terminated_length": 2592.0,
      "completions/mean_length": 602.7421875,
      "completions/mean_terminated_length": 602.7421875,
      "completions/min_length": 98.0,
      "completions/min_terminated_length": 98.0,
      "epoch": 0.1856,
      "grad_norm": 0.008487635292112827,
      "learning_rate": 7.222222222222222e-07,
      "loss": 0.0487,
      "num_tokens": 38977583.0,
      "reward": 1.6311129331588745,
      "reward_std": 0.4975399971008301,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6444879174232483,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.7549636960029602,
      "step": 174
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.7164092664092664,
      "calib/avg_num_step_conf": 3.671875,
      "calib/ece": 0.3578884462151395,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.3466135458167331,
      "calib/gap": 0.11584684684684643,
      "calib/mean_conf": 0.7922310756972113,
      "calib/mu_c": 0.8568468468468465,
      "calib/mu_w": 0.7410000000000001,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.96875,
      "calib/pce": 0.3539442231075698,
      "calib/std_conf": 0.1748778810977799,
      "calib/step_conf_rate": 0.96875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3004.0,
      "completions/max_terminated_length": 3004.0,
      "completions/mean_length": 549.328125,
      "completions/mean_terminated_length": 549.328125,
      "completions/min_length": 163.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.18666666666666668,
      "grad_norm": 0.008262657560408115,
      "learning_rate": 6.944444444444446e-07,
      "loss": 0.0231,
      "num_tokens": 39224035.0,
      "reward": 1.4807465076446533,
      "reward_std": 0.4046202301979065,
      "rewards/accuracy_reward_step": 0.43359375,
      "rewards/final_brier_reward_step": 0.6380871534347534,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.7458364963531494,
      "step": 175
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6982714138286894,
      "calib/avg_num_step_conf": 4.11328125,
      "calib/ece": 0.28112,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.444,
      "calib/gap": 0.1105147058823529,
      "calib/mean_conf": 0.8251200000000001,
      "calib/mu_c": 0.875514705882353,
      "calib/mu_w": 0.7650000000000001,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.28112,
      "calib/std_conf": 0.15838492857592226,
      "calib/step_conf_rate": 0.97265625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 1655.0,
      "completions/max_terminated_length": 1655.0,
      "completions/mean_length": 512.05078125,
      "completions/mean_terminated_length": 518.12255859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.18773333333333334,
      "grad_norm": 0.009166779927909374,
      "learning_rate": 6.666666666666667e-07,
      "loss": -0.0182,
      "num_tokens": 39459184.0,
      "reward": 1.6427857875823975,
      "reward_std": 0.4436536133289337,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6745976805686951,
      "rewards/format_reward_step": 0.9609375,
      "rewards/stepwise_brier_reward": 0.78717041015625,
      "step": 176
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.627831715210356,
      "calib/avg_num_step_conf": 4.078125,
      "calib/ece": 0.2414574898785425,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9375,
      "calib/frac_conf_gt_0.9": 0.42105263157894735,
      "calib/gap": 0.07607335490830625,
      "calib/mean_conf": 0.817165991902834,
      "calib/mu_c": 0.8488888888888888,
      "calib/mu_w": 0.7728155339805826,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.9765625,
      "calib/nonempty_step_conf_rate": 0.953125,
      "calib/pce": 0.23781376518218622,
      "calib/std_conf": 0.16609367868317582,
      "calib/step_conf_rate": 0.953125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 2964.0,
      "completions/max_terminated_length": 2964.0,
      "completions/mean_length": 517.6640625,
      "completions/mean_terminated_length": 525.8809814453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 152.0,
      "epoch": 0.1888,
      "grad_norm": 0.008578428067266941,
      "learning_rate": 6.388888888888889e-07,
      "loss": 0.0231,
      "num_tokens": 39695538.0,
      "reward": 1.6687235832214355,
      "reward_std": 0.4203333258628845,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.674630880355835,
      "rewards/format_reward_step": 0.9375,
      "rewards/stepwise_brier_reward": 0.7502630949020386,
      "step": 177
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6879578754578755,
      "calib/avg_num_step_conf": 3.6328125,
      "calib/ece": 0.22165354330708661,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.4763779527559055,
      "calib/gap": 0.09830455259026694,
      "calib/mean_conf": 0.829763779527559,
      "calib/mu_c": 0.8676923076923078,
      "calib/mu_w": 0.7693877551020408,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.21862204724409448,
      "calib/std_conf": 0.16257749002515856,
      "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": 2298.0,
      "completions/max_terminated_length": 2298.0,
      "completions/mean_length": 491.14453125,
      "completions/mean_terminated_length": 493.07061767578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 216.0,
      "epoch": 0.18986666666666666,
      "grad_norm": 0.008935781195759773,
      "learning_rate": 6.111111111111112e-07,
      "loss": 0.0221,
      "num_tokens": 39927343.0,
      "reward": 1.7922996282577515,
      "reward_std": 0.49695172905921936,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.7280288934707642,
      "rewards/format_reward_step": 0.98828125,
      "rewards/stepwise_brier_reward": 0.8083571195602417,
      "step": 178
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6983082706766918,
      "calib/avg_num_step_conf": 4.015625,
      "calib/ece": 0.2287999999999999,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.42,
      "calib/gap": 0.10298737916219125,
      "calib/mean_conf": 0.8368000000000001,
      "calib/mu_c": 0.877171052631579,
      "calib/mu_w": 0.7741836734693878,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.2287999999999999,
      "calib/std_conf": 0.1451101650471117,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 2859.0,
      "completions/max_terminated_length": 2859.0,
      "completions/mean_length": 537.34765625,
      "completions/mean_terminated_length": 543.7193603515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 191.0,
      "epoch": 0.19093333333333334,
      "grad_norm": 0.008550455793738365,
      "learning_rate": 5.833333333333334e-07,
      "loss": 0.0053,
      "num_tokens": 40171168.0,
      "reward": 1.759314775466919,
      "reward_std": 0.5290068984031677,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.7181859016418457,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.8112608790397644,
      "step": 179
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.7223080021482277,
      "calib/avg_num_step_conf": 3.484375,
      "calib/ece": 0.2504,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.512,
      "calib/gap": 0.08547261009667007,
      "calib/mean_conf": 0.8536,
      "calib/mu_c": 0.8871052631578947,
      "calib/mu_w": 0.8016326530612247,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.248,
      "calib/std_conf": 0.14477237305508256,
      "calib/step_conf_rate": 0.96484375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2635.0,
      "completions/max_terminated_length": 2635.0,
      "completions/mean_length": 589.91015625,
      "completions/mean_terminated_length": 592.2235717773438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 217.0,
      "epoch": 0.192,
      "grad_norm": 0.009398216381669044,
      "learning_rate": 5.555555555555555e-07,
      "loss": 0.0138,
      "num_tokens": 40426041.0,
      "reward": 1.7339614629745483,
      "reward_std": 0.43195515871047974,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.6881320476531982,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.7711513042449951,
      "step": 180
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6823293938238659,
      "calib/avg_num_step_conf": 3.63671875,
      "calib/ece": 0.32776892430278887,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.49800796812749004,
      "calib/gap": 0.09456601855381896,
      "calib/mean_conf": 0.8330278884462151,
      "calib/mu_c": 0.8789922480620156,
      "calib/mu_w": 0.7844262295081966,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.32342629482071716,
      "calib/std_conf": 0.1619441003942825,
      "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": 2949.0,
      "completions/max_terminated_length": 2949.0,
      "completions/mean_length": 501.921875,
      "completions/mean_terminated_length": 507.87353515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.19306666666666666,
      "grad_norm": 0.009197493083775043,
      "learning_rate": 5.277777777777779e-07,
      "loss": 0.0065,
      "num_tokens": 40660797.0,
      "reward": 1.5986297130584717,
      "reward_std": 0.4643535315990448,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.6527894735336304,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.7729793787002563,
      "step": 181
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6768916155419223,
      "calib/avg_num_step_conf": 3.78515625,
      "calib/ece": 0.2026482213438735,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.5098814229249012,
      "calib/gap": 0.09388957055214697,
      "calib/mean_conf": 0.8424901185770752,
      "calib/mu_c": 0.8758895705521471,
      "calib/mu_w": 0.7820000000000001,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.20043478260869568,
      "calib/std_conf": 0.15115015209860883,
      "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": 2014.0,
      "completions/max_terminated_length": 2014.0,
      "completions/mean_length": 519.12890625,
      "completions/mean_terminated_length": 521.1647338867188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.19413333333333332,
      "grad_norm": 0.00853924173861742,
      "learning_rate": 5.000000000000001e-07,
      "loss": -0.0122,
      "num_tokens": 40899854.0,
      "reward": 1.8242754936218262,
      "reward_std": 0.40249642729759216,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.7367503643035889,
      "rewards/format_reward_step": 0.98046875,
      "rewards/stepwise_brier_reward": 0.7791011333465576,
      "step": 182
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6730018491884118,
      "calib/avg_num_step_conf": 3.41796875,
      "calib/ece": 0.20656,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.452,
      "calib/gap": 0.09167317307033751,
      "calib/mean_conf": 0.83456,
      "calib/mu_c": 0.8686624203821656,
      "calib/mu_w": 0.7769892473118281,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.20656,
      "calib/std_conf": 0.14758592886857475,
      "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": 2830.0,
      "completions/max_terminated_length": 2830.0,
      "completions/mean_length": 535.10546875,
      "completions/mean_terminated_length": 537.2039794921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 196.0,
      "epoch": 0.1952,
      "grad_norm": 0.008417710661888123,
      "learning_rate": 4.7222222222222226e-07,
      "loss": 0.0049,
      "num_tokens": 41143521.0,
      "reward": 1.777566909790039,
      "reward_std": 0.5069254636764526,
      "rewards/accuracy_reward_step": 0.61328125,
      "rewards/final_brier_reward_step": 0.7230734825134277,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.7778192758560181,
      "step": 183
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6357766394921001,
      "calib/avg_num_step_conf": 3.6171875,
      "calib/ece": 0.20892000000000005,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.568,
      "calib/gap": 0.049339153019262616,
      "calib/mean_conf": 0.86332,
      "calib/mu_c": 0.8797005988023954,
      "calib/mu_w": 0.8303614457831328,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.20212000000000005,
      "calib/std_conf": 0.13884587714440785,
      "calib/step_conf_rate": 0.97265625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 2762.0,
      "completions/max_terminated_length": 2762.0,
      "completions/mean_length": 530.875,
      "completions/mean_terminated_length": 537.1699829101562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.19626666666666667,
      "grad_norm": 0.008153742179274559,
      "learning_rate": 4.444444444444445e-07,
      "loss": 0.009,
      "num_tokens": 41384705.0,
      "reward": 1.826616644859314,
      "reward_std": 0.5993893146514893,
      "rewards/accuracy_reward_step": 0.65234375,
      "rewards/final_brier_reward_step": 0.7080210447311401,
      "rewards/format_reward_step": 0.953125,
      "rewards/stepwise_brier_reward": 0.7781328558921814,
      "step": 184
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.7254847645429363,
      "calib/avg_num_step_conf": 4.16796875,
      "calib/ece": 0.2365182186234816,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.582995951417004,
      "calib/gap": 0.13288157894736852,
      "calib/mean_conf": 0.8446153846153847,
      "calib/mu_c": 0.8957236842105264,
      "calib/mu_w": 0.7628421052631579,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.23287449392712534,
      "calib/std_conf": 0.17386227577425378,
      "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": 2969.0,
      "completions/max_terminated_length": 2969.0,
      "completions/mean_length": 553.6484375,
      "completions/mean_terminated_length": 558.0078735351562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 146.0,
      "epoch": 0.19733333333333333,
      "grad_norm": 0.007938927970826626,
      "learning_rate": 4.1666666666666667e-07,
      "loss": 0.0311,
      "num_tokens": 41633359.0,
      "reward": 1.7404544353485107,
      "reward_std": 0.5316325426101685,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.712269127368927,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.7729862332344055,
      "step": 185
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5913487554112554,
      "calib/avg_num_step_conf": 3.46484375,
      "calib/ece": 0.28547999999999996,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.576,
      "calib/gap": 0.03818993506493518,
      "calib/mean_conf": 0.8654,
      "calib/mu_c": 0.8800649350649351,
      "calib/mu_w": 0.8418749999999999,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.26744,
      "calib/std_conf": 0.14395290896678678,
      "calib/step_conf_rate": 0.96484375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2966.0,
      "completions/max_terminated_length": 2966.0,
      "completions/mean_length": 577.9453125,
      "completions/mean_terminated_length": 577.9453125,
      "completions/min_length": 188.0,
      "completions/min_terminated_length": 188.0,
      "epoch": 0.1984,
      "grad_norm": 0.007273213006556034,
      "learning_rate": 3.8888888888888895e-07,
      "loss": 0.0173,
      "num_tokens": 41886353.0,
      "reward": 1.7397516965866089,
      "reward_std": 0.5561968088150024,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.6718722581863403,
      "rewards/format_reward_step": 0.953125,
      "rewards/stepwise_brier_reward": 0.7715095281600952,
      "step": 186
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.49982406755805764,
      "calib/avg_num_step_conf": 4.875,
      "calib/ece": 0.2998353909465021,
      "calib/final_conf_rate": 0.94921875,
      "calib/format_rate": 0.93359375,
      "calib/frac_conf_gt_0.9": 0.6378600823045267,
      "calib/gap": 0.013572836030964264,
      "calib/mean_conf": 0.8896296296296297,
      "calib/mu_c": 0.8951034482758622,
      "calib/mu_w": 0.8815306122448979,
      "calib/nonempty_final_conf_rate": 0.94921875,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.2963786008230453,
      "calib/std_conf": 0.11272287783732576,
      "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": 0.0078125,
      "completions/max_length": 3039.0,
      "completions/max_terminated_length": 3039.0,
      "completions/mean_length": 621.875,
      "completions/mean_terminated_length": 626.7716674804688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 204.0,
      "epoch": 0.19946666666666665,
      "grad_norm": 0.007824388332664967,
      "learning_rate": 3.611111111111111e-07,
      "loss": 0.069,
      "num_tokens": 42147097.0,
      "reward": 1.6536288261413574,
      "reward_std": 0.5715471506118774,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.6232359409332275,
      "rewards/format_reward_step": 0.93359375,
      "rewards/stepwise_brier_reward": 0.7256544828414917,
      "step": 187
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6511424731182796,
      "calib/avg_num_step_conf": 3.65234375,
      "calib/ece": 0.23110671936758895,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.5889328063241107,
      "calib/gap": 0.09423387096774183,
      "calib/mean_conf": 0.8541106719367589,
      "calib/mu_c": 0.8887499999999999,
      "calib/mu_w": 0.7945161290322581,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.22640316205533598,
      "calib/std_conf": 0.16959991168692434,
      "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": 2935.0,
      "completions/max_terminated_length": 2935.0,
      "completions/mean_length": 599.0703125,
      "completions/mean_terminated_length": 599.0703125,
      "completions/min_length": 165.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.20053333333333334,
      "grad_norm": 0.007716072723269463,
      "learning_rate": 3.3333333333333335e-07,
      "loss": 0.0562,
      "num_tokens": 42404531.0,
      "reward": 1.789067029953003,
      "reward_std": 0.418638676404953,
      "rewards/accuracy_reward_step": 0.625,
      "rewards/final_brier_reward_step": 0.7137839794158936,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.76279616355896,
      "step": 188
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6447690217391305,
      "calib/avg_num_step_conf": 3.6015625,
      "calib/ece": 0.23900793650793645,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.6190476190476191,
      "calib/gap": 0.0731521739130434,
      "calib/mean_conf": 0.8732936507936507,
      "calib/mu_c": 0.9,
      "calib/mu_w": 0.8268478260869566,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.23869047619047612,
      "calib/std_conf": 0.1371617957510612,
      "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": 1873.0,
      "completions/max_terminated_length": 1873.0,
      "completions/mean_length": 512.40234375,
      "completions/mean_terminated_length": 514.4118041992188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 178.0,
      "epoch": 0.2016,
      "grad_norm": 0.009424522519111633,
      "learning_rate": 3.055555555555556e-07,
      "loss": 0.0108,
      "num_tokens": 42643474.0,
      "reward": 1.7970731258392334,
      "reward_std": 0.40603047609329224,
      "rewards/accuracy_reward_step": 0.625,
      "rewards/final_brier_reward_step": 0.7096503973007202,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.7755172252655029,
      "step": 189
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6672170987038883,
      "calib/avg_num_step_conf": 3.421875,
      "calib/ece": 0.32212598425196853,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.5196850393700787,
      "calib/gap": 0.07488160518444664,
      "calib/mean_conf": 0.8554330708661417,
      "calib/mu_c": 0.8902205882352942,
      "calib/mu_w": 0.8153389830508475,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.953125,
      "calib/pce": 0.32106299212598427,
      "calib/std_conf": 0.1486329088172709,
      "calib/step_conf_rate": 0.953125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1772.0,
      "completions/max_terminated_length": 1772.0,
      "completions/mean_length": 564.8984375,
      "completions/mean_terminated_length": 567.11376953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 70.0,
      "epoch": 0.20266666666666666,
      "grad_norm": 0.008034134283661842,
      "learning_rate": 2.7777777777777776e-07,
      "loss": -0.0175,
      "num_tokens": 42893696.0,
      "reward": 1.6131715774536133,
      "reward_std": 0.5308533906936646,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.625710129737854,
      "rewards/format_reward_step": 0.953125,
      "rewards/stepwise_brier_reward": 0.7332261800765991,
      "step": 190
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6234649122807018,
      "calib/avg_num_step_conf": 3.65625,
      "calib/ece": 0.3415415019762845,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.5770750988142292,
      "calib/gap": 0.06813220551378463,
      "calib/mean_conf": 0.8672332015810277,
      "calib/mu_c": 0.8995488721804512,
      "calib/mu_w": 0.8314166666666666,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.3415415019762845,
      "calib/std_conf": 0.1448491547536018,
      "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": 2841.0,
      "completions/max_terminated_length": 2841.0,
      "completions/mean_length": 545.61328125,
      "completions/mean_terminated_length": 547.7529907226562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 189.0,
      "epoch": 0.20373333333333332,
      "grad_norm": 0.008736473508179188,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0121,
      "num_tokens": 43137541.0,
      "reward": 1.602749228477478,
      "reward_std": 0.46529421210289,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.6267433166503906,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.7217535972595215,
      "step": 191
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6254540651211941,
      "calib/avg_num_step_conf": 3.34375,
      "calib/ece": 0.29011999999999993,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.556,
      "calib/gap": 0.052795059771481245,
      "calib/mean_conf": 0.8671599999999999,
      "calib/mu_c": 0.8889115646258501,
      "calib/mu_w": 0.8361165048543688,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.9453125,
      "calib/pce": 0.28464,
      "calib/std_conf": 0.13686027327168393,
      "calib/step_conf_rate": 0.9453125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3037.0,
      "completions/max_terminated_length": 3037.0,
      "completions/mean_length": 566.765625,
      "completions/mean_terminated_length": 566.765625,
      "completions/min_length": 164.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.2048,
      "grad_norm": 0.008317789062857628,
      "learning_rate": 2.2222222222222224e-07,
      "loss": 0.0263,
      "num_tokens": 43387609.0,
      "reward": 1.6767215728759766,
      "reward_std": 0.4984460473060608,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.6436020135879517,
      "rewards/format_reward_step": 0.94140625,
      "rewards/stepwise_brier_reward": 0.735159158706665,
      "step": 192
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5463848039215687,
      "calib/avg_num_step_conf": 3.65234375,
      "calib/ece": 0.3450390625,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.59375,
      "calib/gap": 0.03662254901960793,
      "calib/mean_conf": 0.8762890624999999,
      "calib/mu_c": 0.8934558823529412,
      "calib/mu_w": 0.8568333333333332,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9609375,
      "calib/pce": 0.3450390625,
      "calib/std_conf": 0.13279116901312032,
      "calib/step_conf_rate": 0.9609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1100.0,
      "completions/max_terminated_length": 1100.0,
      "completions/mean_length": 532.97265625,
      "completions/mean_terminated_length": 535.0628051757812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 257.0,
      "epoch": 0.20586666666666667,
      "grad_norm": 0.009877479635179043,
      "learning_rate": 1.9444444444444447e-07,
      "loss": 0.0408,
      "num_tokens": 43629762.0,
      "reward": 1.6049281358718872,
      "reward_std": 0.5841389894485474,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6001319885253906,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.7180180549621582,
      "step": 193
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.7548754062838571,
      "calib/avg_num_step_conf": 3.546875,
      "calib/ece": 0.26605691056910563,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.5934959349593496,
      "calib/gap": 0.15619989165763792,
      "calib/mean_conf": 0.8430487804878048,
      "calib/mu_c": 0.9090845070422533,
      "calib/mu_w": 0.7528846153846154,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.9765625,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.26593495934959344,
      "calib/std_conf": 0.19721447833765263,
      "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": 2721.0,
      "completions/max_terminated_length": 2721.0,
      "completions/mean_length": 561.62109375,
      "completions/mean_terminated_length": 561.62109375,
      "completions/min_length": 208.0,
      "completions/min_terminated_length": 208.0,
      "epoch": 0.20693333333333333,
      "grad_norm": 0.009010870940983295,
      "learning_rate": 1.6666666666666668e-07,
      "loss": 0.1407,
      "num_tokens": 43879481.0,
      "reward": 1.6807918548583984,
      "reward_std": 0.5311864614486694,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6933391094207764,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.7876406908035278,
      "step": 194
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.6227896497576297,
      "calib/avg_num_step_conf": 3.38671875,
      "calib/ece": 0.26774193548387093,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.6290322580645161,
      "calib/gap": 0.06824742268041206,
      "calib/mean_conf": 0.8733064516129033,
      "calib/mu_c": 0.8999999999999998,
      "calib/mu_w": 0.8317525773195877,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.26608870967741927,
      "calib/std_conf": 0.14243120452658123,
      "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": 0.01953125,
      "completions/max_length": 2492.0,
      "completions/max_terminated_length": 2492.0,
      "completions/mean_length": 514.96875,
      "completions/mean_terminated_length": 525.2271118164062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.208,
      "grad_norm": 0.008795440196990967,
      "learning_rate": 1.3888888888888888e-07,
      "loss": -0.0087,
      "num_tokens": 44117297.0,
      "reward": 1.7167940139770508,
      "reward_std": 0.4254692494869232,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.6748417615890503,
      "rewards/format_reward_step": 0.95703125,
      "rewards/stepwise_brier_reward": 0.7392090559005737,
      "step": 195
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.7245424836601306,
      "calib/avg_num_step_conf": 3.74609375,
      "calib/ece": 0.30369047619047623,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.6825396825396826,
      "calib/gap": 0.09125882352941184,
      "calib/mean_conf": 0.8989285714285714,
      "calib/mu_c": 0.9358666666666666,
      "calib/mu_w": 0.8446078431372548,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.30369047619047623,
      "calib/std_conf": 0.12059723591632436,
      "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": 2154.0,
      "completions/max_terminated_length": 2154.0,
      "completions/mean_length": 489.078125,
      "completions/mean_terminated_length": 490.99609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 216.0,
      "epoch": 0.20906666666666668,
      "grad_norm": 0.009376071393489838,
      "learning_rate": 1.1111111111111112e-07,
      "loss": 0.0425,
      "num_tokens": 44345045.0,
      "reward": 1.7269446849822998,
      "reward_std": 0.2757290005683899,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.6787136793136597,
      "rewards/format_reward_step": 0.9765625,
      "rewards/stepwise_brier_reward": 0.7603150010108948,
      "step": 196
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5699561403508772,
      "calib/avg_num_step_conf": 3.62109375,
      "calib/ece": 0.3515810276679841,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.5691699604743083,
      "calib/gap": 0.01779323308270664,
      "calib/mean_conf": 0.8588537549407116,
      "calib/mu_c": 0.8672932330827067,
      "calib/mu_w": 0.8495,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.34237154150197624,
      "calib/std_conf": 0.15121337421417586,
      "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": 1421.0,
      "completions/max_terminated_length": 1421.0,
      "completions/mean_length": 516.578125,
      "completions/mean_terminated_length": 518.6039428710938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.21013333333333334,
      "grad_norm": 0.009442859329283237,
      "learning_rate": 8.333333333333334e-08,
      "loss": 0.0148,
      "num_tokens": 44582345.0,
      "reward": 1.599530816078186,
      "reward_std": 0.5991620421409607,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.6044124364852905,
      "rewards/format_reward_step": 0.96875,
      "rewards/stepwise_brier_reward": 0.7155857682228088,
      "step": 197
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5945955086580087,
      "calib/avg_num_step_conf": 3.81640625,
      "calib/ece": 0.25832000000000005,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.652,
      "calib/gap": 0.0575757575757575,
      "calib/mean_conf": 0.8687999999999999,
      "calib/mu_c": 0.8909090909090909,
      "calib/mu_w": 0.8333333333333334,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.25556000000000006,
      "calib/std_conf": 0.14774356161944924,
      "calib/step_conf_rate": 0.9765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 3027.0,
      "completions/max_terminated_length": 3027.0,
      "completions/mean_length": 505.69140625,
      "completions/mean_terminated_length": 511.6877746582031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 155.0,
      "epoch": 0.2112,
      "grad_norm": 0.00891911145299673,
      "learning_rate": 5.555555555555556e-08,
      "loss": 0.0257,
      "num_tokens": 44817186.0,
      "reward": 1.738128900527954,
      "reward_std": 0.48579591512680054,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.6744344234466553,
      "rewards/format_reward_step": 0.9609375,
      "rewards/stepwise_brier_reward": 0.7468315958976746,
      "step": 198
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6567143643368188,
      "calib/avg_num_step_conf": 3.5703125,
      "calib/ece": 0.2606,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.628,
      "calib/gap": 0.060875068794716714,
      "calib/mean_conf": 0.85956,
      "calib/mu_c": 0.8819620253164556,
      "calib/mu_w": 0.8210869565217389,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.24408000000000002,
      "calib/std_conf": 0.1639981902339169,
      "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": 2492.0,
      "completions/max_terminated_length": 2492.0,
      "completions/mean_length": 596.85546875,
      "completions/mean_terminated_length": 596.85546875,
      "completions/min_length": 187.0,
      "completions/min_terminated_length": 187.0,
      "epoch": 0.21226666666666666,
      "grad_norm": 0.008912494406104088,
      "learning_rate": 2.777777777777778e-08,
      "loss": -0.0006,
      "num_tokens": 45074181.0,
      "reward": 1.7760467529296875,
      "reward_std": 0.5032174587249756,
      "rewards/accuracy_reward_step": 0.6171875,
      "rewards/final_brier_reward_step": 0.6886956691741943,
      "rewards/format_reward_step": 0.96484375,
      "rewards/stepwise_brier_reward": 0.7826792597770691,
      "step": 199
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.8373956369512524,
      "calib/avg_num_step_conf": 3.48046875,
      "calib/ece": 0.22738095238095232,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.5595238095238095,
      "calib/gap": 0.18070832211150023,
      "calib/mean_conf": 0.8543650793650794,
      "calib/mu_c": 0.921772151898734,
      "calib/mu_w": 0.7410638297872337,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.22738095238095232,
      "calib/std_conf": 0.14974579534996532,
      "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": 3046.0,
      "completions/max_terminated_length": 3046.0,
      "completions/mean_length": 578.37890625,
      "completions/mean_terminated_length": 580.6470947265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 177.0,
      "epoch": 0.21333333333333335,
      "grad_norm": 0.008349462412297726,
      "learning_rate": 0.0,
      "loss": 0.031,
      "num_tokens": 45330294.0,
      "reward": 1.8081409931182861,
      "reward_std": 0.4588835835456848,
      "rewards/accuracy_reward_step": 0.6171875,
      "rewards/final_brier_reward_step": 0.7590082287788391,
      "rewards/format_reward_step": 0.97265625,
      "rewards/stepwise_brier_reward": 0.8251185417175293,
      "step": 200
    },
    {
      "epoch": 0.21333333333333335,
      "step": 200,
      "total_flos": 0.0,
      "train_loss": 0.025571908718557097,
      "train_runtime": 10962.4299,
      "train_samples_per_second": 4.67,
      "train_steps_per_second": 0.018
    }
  ],
  "logging_steps": 1,
  "max_steps": 200,
  "num_input_tokens_seen": 45330294,
  "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
}