{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.21333333333333335,
  "eval_steps": 500,
  "global_step": 200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "aux_brier/lambda": 0.05,
      "aux_brier/loss": 5.791089203391117e-07,
      "aux_brier/mean_group_std": 0.06289231620091193,
      "aux_brier/mean_r": 0.4665906001184907,
      "aux_brier/n_active_tok": 24.615384615384617,
      "aux_brier/n_groups": 5.3076923076923075,
      "aux_brier/n_step_records": 6.153846153846154,
      "calib/answer_extract_rate": 0.08203125,
      "calib/auroc": 0.6944444444444445,
      "calib/avg_num_step_conf": 0.3359375,
      "calib/ece": 0.6230769230769231,
      "calib/final_conf_rate": 0.05078125,
      "calib/format_rate": 0.04296875,
      "calib/frac_conf_gt_0.9": 0.7692307692307693,
      "calib/gap": 0.03861111111111115,
      "calib/mean_conf": 0.9307692307692309,
      "calib/mu_c": 0.9575,
      "calib/mu_w": 0.9188888888888889,
      "calib/nonempty_final_conf_rate": 0.05078125,
      "calib/nonempty_reasoning_rate": 0.09765625,
      "calib/nonempty_step_conf_rate": 0.0703125,
      "calib/pce": 0.6230769230769231,
      "calib/std_conf": 0.07965903671384378,
      "calib/step_conf_rate": 0.0703125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08984375,
      "completions/max_length": 2955.0,
      "completions/max_terminated_length": 2955.0,
      "completions/mean_length": 613.67578125,
      "completions/mean_terminated_length": 674.2532348632812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0010666666666666667,
      "grad_norm": 0.16193221509456635,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0332,
      "num_tokens": 264685.0,
      "reward": 0.03929531201720238,
      "reward_std": 0.08434611558914185,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.01655624993145466,
      "rewards/format_reward_step_strict": 0.0390625,
      "step": 1
    },
    {
      "aux_brier/lambda": 0.05000000000000002,
      "aux_brier/loss": 2.461345396504181e-08,
      "aux_brier/mean_group_std": 0.046398653263787254,
      "aux_brier/mean_r": 0.430243897442093,
      "aux_brier/n_active_tok": 28.42105263157895,
      "aux_brier/n_groups": 5.894736842105263,
      "aux_brier/n_step_records": 7.105263157894737,
      "calib/answer_extract_rate": 0.13671875,
      "calib/auroc": 0.5338345864661654,
      "calib/avg_num_step_conf": 0.55078125,
      "calib/ece": 0.6261538461538463,
      "calib/final_conf_rate": 0.1015625,
      "calib/format_rate": 0.08984375,
      "calib/frac_conf_gt_0.9": 0.7692307692307693,
      "calib/gap": 0.002406015037593856,
      "calib/mean_conf": 0.8953846153846153,
      "calib/mu_c": 0.897142857142857,
      "calib/mu_w": 0.8947368421052632,
      "calib/nonempty_final_conf_rate": 0.1015625,
      "calib/nonempty_reasoning_rate": 0.14453125,
      "calib/nonempty_step_conf_rate": 0.109375,
      "calib/pce": 0.6261538461538463,
      "calib/std_conf": 0.18653172073466937,
      "calib/step_conf_rate": 0.109375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0546875,
      "completions/max_length": 3001.0,
      "completions/max_terminated_length": 3001.0,
      "completions/mean_length": 646.4609375,
      "completions/mean_terminated_length": 683.8594970703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0021333333333333334,
      "grad_norm": 0.006253078579902649,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0643,
      "num_tokens": 533467.0,
      "reward": 0.08358447253704071,
      "reward_std": 0.15892045199871063,
      "rewards/accuracy_reward_step": 0.03125,
      "rewards/final_brier_reward_step": 0.02965039201080799,
      "rewards/format_reward_step_strict": 0.08984375,
      "step": 2
    },
    {
      "aux_brier/lambda": 0.05,
      "aux_brier/loss": -2.8176203592522597e-08,
      "aux_brier/mean_group_std": 0.029847013798721494,
      "aux_brier/mean_r": 0.41571989539143306,
      "aux_brier/n_active_tok": 22.285714285714285,
      "aux_brier/n_groups": 4.785714285714286,
      "aux_brier/n_step_records": 5.571428571428571,
      "calib/answer_extract_rate": 0.08203125,
      "calib/auroc": 0.7692307692307692,
      "calib/avg_num_step_conf": 0.3046875,
      "calib/ece": 0.644705882352941,
      "calib/final_conf_rate": 0.06640625,
      "calib/format_rate": 0.05078125,
      "calib/frac_conf_gt_0.9": 0.7647058823529411,
      "calib/gap": 0.13076923076923086,
      "calib/mean_conf": 0.88,
      "calib/mu_c": 0.98,
      "calib/mu_w": 0.8492307692307691,
      "calib/nonempty_final_conf_rate": 0.06640625,
      "calib/nonempty_reasoning_rate": 0.09375,
      "calib/nonempty_step_conf_rate": 0.06640625,
      "calib/pce": 0.644705882352941,
      "calib/std_conf": 0.19991174523362287,
      "calib/step_conf_rate": 0.06640625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08984375,
      "completions/max_length": 3060.0,
      "completions/max_terminated_length": 3060.0,
      "completions/mean_length": 614.921875,
      "completions/mean_terminated_length": 675.622314453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0032,
      "grad_norm": 0.004395780153572559,
      "learning_rate": 7.5e-07,
      "loss": 0.0312,
      "num_tokens": 796143.0,
      "reward": 0.04669833928346634,
      "reward_std": 0.10021394491195679,
      "rewards/accuracy_reward_step": 0.015625,
      "rewards/final_brier_reward_step": 0.022730860859155655,
      "rewards/format_reward_step_strict": 0.05078125,
      "step": 3
    },
    {
      "aux_brier/lambda": 0.049999999999999996,
      "aux_brier/loss": 3.7930230269787235e-09,
      "aux_brier/mean_group_std": 0.030711086994661366,
      "aux_brier/mean_r": 0.4590041716009507,
      "aux_brier/n_active_tok": 22.545454545454547,
      "aux_brier/n_groups": 4.818181818181818,
      "aux_brier/n_step_records": 5.636363636363637,
      "calib/answer_extract_rate": 0.05859375,
      "calib/auroc": 0.6,
      "calib/avg_num_step_conf": 0.25390625,
      "calib/ece": 0.6273272727272727,
      "calib/final_conf_rate": 0.04296875,
      "calib/format_rate": 0.03125,
      "calib/frac_conf_gt_0.9": 0.5454545454545454,
      "calib/gap": 0.25493999999999994,
      "calib/mean_conf": 0.7182363636363637,
      "calib/mu_c": 0.95,
      "calib/mu_w": 0.69506,
      "calib/nonempty_final_conf_rate": 0.04296875,
      "calib/nonempty_reasoning_rate": 0.078125,
      "calib/nonempty_step_conf_rate": 0.0546875,
      "calib/pce": 0.6273272727272727,
      "calib/std_conf": 0.3309647255928898,
      "calib/step_conf_rate": 0.0546875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.08984375,
      "completions/max_length": 2977.0,
      "completions/max_terminated_length": 2977.0,
      "completions/mean_length": 669.00390625,
      "completions/mean_terminated_length": 735.0429077148438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.004266666666666667,
      "grad_norm": 0.0034748422913253307,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0095,
      "num_tokens": 1073576.0,
      "reward": 0.01978631690144539,
      "reward_std": 0.04911034554243088,
      "rewards/accuracy_reward_step": 0.00390625,
      "rewards/final_brier_reward_step": 0.008832764811813831,
      "rewards/format_reward_step_strict": 0.02734375,
      "step": 4
    },
    {
      "aux_brier/lambda": 0.049999999999999996,
      "aux_brier/loss": -7.947286206899663e-09,
      "aux_brier/mean_group_std": 0.007505333331238634,
      "aux_brier/mean_r": 0.3610654883250641,
      "aux_brier/n_active_tok": 26.0,
      "aux_brier/n_groups": 6.1,
      "aux_brier/n_step_records": 6.5,
      "calib/answer_extract_rate": 0.06640625,
      "calib/avg_num_step_conf": 0.296875,
      "calib/ece": 0.965,
      "calib/final_conf_rate": 0.03125,
      "calib/format_rate": 0.0234375,
      "calib/frac_conf_gt_0.9": 1.0,
      "calib/mean_conf": 0.965,
      "calib/mu_c": NaN,
      "calib/mu_w": 0.965,
      "calib/nonempty_final_conf_rate": 0.03125,
      "calib/nonempty_reasoning_rate": 0.08203125,
      "calib/nonempty_step_conf_rate": 0.046875,
      "calib/pce": 0.965,
      "calib/std_conf": 0.013228756555322964,
      "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": 0.10546875,
      "completions/max_length": 2969.0,
      "completions/max_terminated_length": 2969.0,
      "completions/mean_length": 641.55859375,
      "completions/mean_terminated_length": 717.2008666992188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 8.0,
      "epoch": 0.005333333333333333,
      "grad_norm": 0.0031038725282996893,
      "learning_rate": 1.25e-06,
      "loss": 0.0112,
      "num_tokens": 1344503.0,
      "reward": 0.014093163423240185,
      "reward_std": 0.03986148536205292,
      "rewards/accuracy_reward_step": 0.00390625,
      "rewards/final_brier_reward_step": 0.0016851562540978193,
      "rewards/format_reward_step_strict": 0.01953125,
      "step": 5
    },
    {
      "aux_brier/lambda": 0.05000000000000001,
      "aux_brier/loss": -4.303009086303224e-09,
      "aux_brier/mean_group_std": 0.017132194048956134,
      "aux_brier/mean_r": 0.5234074353874353,
      "aux_brier/n_active_tok": 22.11764705882353,
      "aux_brier/n_groups": 5.176470588235294,
      "aux_brier/n_step_records": 5.529411764705882,
      "calib/answer_extract_rate": 0.0703125,
      "calib/avg_num_step_conf": 0.375,
      "calib/ece": 0.8025,
      "calib/final_conf_rate": 0.0625,
      "calib/format_rate": 0.046875,
      "calib/frac_conf_gt_0.9": 0.625,
      "calib/mean_conf": 0.8025,
      "calib/mu_c": NaN,
      "calib/mu_w": 0.8025,
      "calib/nonempty_final_conf_rate": 0.0625,
      "calib/nonempty_reasoning_rate": 0.09375,
      "calib/nonempty_step_conf_rate": 0.078125,
      "calib/pce": 0.8025,
      "calib/std_conf": 0.2742147880767921,
      "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": 0.08984375,
      "completions/max_length": 3051.0,
      "completions/max_terminated_length": 3051.0,
      "completions/mean_length": 554.87890625,
      "completions/mean_terminated_length": 609.65234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0064,
      "grad_norm": 0.0043120114132761955,
      "learning_rate": 1.5e-06,
      "loss": 0.0394,
      "num_tokens": 1592504.0,
      "reward": 0.024134960025548935,
      "reward_std": 0.05962400883436203,
      "rewards/accuracy_reward_step": 0.0,
      "rewards/final_brier_reward_step": 0.010602343827486038,
      "rewards/format_reward_step_strict": 0.04296875,
      "step": 6
    },
    {
      "aux_brier/lambda": 0.049999999999999996,
      "aux_brier/loss": -7.968535374981332e-08,
      "aux_brier/mean_group_std": 0.01891592284558502,
      "aux_brier/mean_r": 0.6214788348051656,
      "aux_brier/n_active_tok": 28.363636363636363,
      "aux_brier/n_groups": 6.0,
      "aux_brier/n_step_records": 7.090909090909091,
      "calib/answer_extract_rate": 0.06640625,
      "calib/auroc": 0.25,
      "calib/avg_num_step_conf": 0.31640625,
      "calib/ece": 0.6187777777777778,
      "calib/final_conf_rate": 0.03515625,
      "calib/format_rate": 0.02734375,
      "calib/frac_conf_gt_0.9": 0.4444444444444444,
      "calib/gap": -0.12342857142857144,
      "calib/mean_conf": 0.841,
      "calib/mu_c": 0.745,
      "calib/mu_w": 0.8684285714285714,
      "calib/nonempty_final_conf_rate": 0.03515625,
      "calib/nonempty_reasoning_rate": 0.08203125,
      "calib/nonempty_step_conf_rate": 0.05859375,
      "calib/pce": 0.6187777777777778,
      "calib/std_conf": 0.13927271408602937,
      "calib/step_conf_rate": 0.05859375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1171875,
      "completions/max_length": 2951.0,
      "completions/max_terminated_length": 2951.0,
      "completions/mean_length": 671.2109375,
      "completions/mean_terminated_length": 760.3097534179688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.007466666666666667,
      "grad_norm": 0.13974224030971527,
      "learning_rate": 1.75e-06,
      "loss": 0.009,
      "num_tokens": 1871758.0,
      "reward": 0.02218896523118019,
      "reward_std": 0.0627598688006401,
      "rewards/accuracy_reward_step": 0.0078125,
      "rewards/final_brier_reward_step": 0.010630855336785316,
      "rewards/format_reward_step_strict": 0.0234375,
      "step": 7
    },
    {
      "aux_brier/lambda": 0.05000000000000001,
      "aux_brier/loss": -1.0630785437134213e-08,
      "aux_brier/mean_group_std": 0.06213361844459951,
      "aux_brier/mean_r": 0.3225575777875755,
      "aux_brier/n_active_tok": 23.466666666666665,
      "aux_brier/n_groups": 5.066666666666666,
      "aux_brier/n_step_records": 5.866666666666666,
      "calib/answer_extract_rate": 0.09765625,
      "calib/auroc": 0.3125,
      "calib/avg_num_step_conf": 0.34765625,
      "calib/ece": 0.5566666666666668,
      "calib/final_conf_rate": 0.0703125,
      "calib/format_rate": 0.0546875,
      "calib/frac_conf_gt_0.9": 0.7222222222222222,
      "calib/gap": 0.03750000000000009,
      "calib/mean_conf": 0.89,
      "calib/mu_c": 0.915,
      "calib/mu_w": 0.8775,
      "calib/nonempty_final_conf_rate": 0.0703125,
      "calib/nonempty_reasoning_rate": 0.109375,
      "calib/nonempty_step_conf_rate": 0.07421875,
      "calib/pce": 0.5566666666666668,
      "calib/std_conf": 0.20264912204760888,
      "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": 0.1015625,
      "completions/max_length": 3052.0,
      "completions/max_terminated_length": 3052.0,
      "completions/mean_length": 625.6796875,
      "completions/mean_terminated_length": 696.40869140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.008533333333333334,
      "grad_norm": 0.002846751594915986,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0284,
      "num_tokens": 2138444.0,
      "reward": 0.055156148970127106,
      "reward_std": 0.08708842843770981,
      "rewards/accuracy_reward_step": 0.0234375,
      "rewards/final_brier_reward_step": 0.025312108919024467,
      "rewards/format_reward_step_strict": 0.05078125,
      "step": 8
    },
    {
      "aux_brier/lambda": 0.05,
      "aux_brier/loss": 4.903836682755371e-08,
      "aux_brier/mean_group_std": 0.06103745202557255,
      "aux_brier/mean_r": 0.5146568030651996,
      "aux_brier/n_active_tok": 20.857142857142858,
      "aux_brier/n_groups": 4.071428571428571,
      "aux_brier/n_step_records": 5.214285714285714,
      "calib/answer_extract_rate": 0.08203125,
      "calib/auroc": 0.35833333333333334,
      "calib/avg_num_step_conf": 0.296875,
      "calib/ece": 0.6105941176470588,
      "calib/final_conf_rate": 0.06640625,
      "calib/format_rate": 0.0546875,
      "calib/frac_conf_gt_0.9": 0.6470588235294118,
      "calib/gap": -0.022841666666666538,
      "calib/mean_conf": 0.7941235294117647,
      "calib/mu_c": 0.778,
      "calib/mu_w": 0.8008416666666666,
      "calib/nonempty_final_conf_rate": 0.06640625,
      "calib/nonempty_reasoning_rate": 0.1015625,
      "calib/nonempty_step_conf_rate": 0.08203125,
      "calib/pce": 0.5553,
      "calib/std_conf": 0.3163816524936332,
      "calib/step_conf_rate": 0.08203125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.07421875,
      "completions/max_length": 3042.0,
      "completions/max_terminated_length": 3042.0,
      "completions/mean_length": 716.05859375,
      "completions/mean_terminated_length": 773.464111328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 6.0,
      "epoch": 0.0096,
      "grad_norm": 0.2166428118944168,
      "learning_rate": 2.25e-06,
      "loss": 0.0302,
      "num_tokens": 2429291.0,
      "reward": 0.04808994382619858,
      "reward_std": 0.10048744082450867,
      "rewards/accuracy_reward_step": 0.01953125,
      "rewards/final_brier_reward_step": 0.02048475854098797,
      "rewards/format_reward_step_strict": 0.046875,
      "step": 9
    },
    {
      "aux_brier/lambda": 0.05000000000000001,
      "aux_brier/loss": -6.612390301974358e-09,
      "aux_brier/mean_group_std": 0.022029799530182782,
      "aux_brier/mean_r": 0.48289992708497376,
      "aux_brier/n_active_tok": 21.0,
      "aux_brier/n_groups": 4.5625,
      "aux_brier/n_step_records": 5.25,
      "calib/answer_extract_rate": 0.08984375,
      "calib/auroc": 0.6388888888888888,
      "calib/avg_num_step_conf": 0.33984375,
      "calib/ece": 0.672,
      "calib/final_conf_rate": 0.05859375,
      "calib/format_rate": 0.05859375,
      "calib/frac_conf_gt_0.9": 0.6666666666666666,
      "calib/gap": 0.10166666666666657,
      "calib/mean_conf": 0.8720000000000001,
      "calib/mu_c": 0.9533333333333333,
      "calib/mu_w": 0.8516666666666667,
      "calib/nonempty_final_conf_rate": 0.05859375,
      "calib/nonempty_reasoning_rate": 0.109375,
      "calib/nonempty_step_conf_rate": 0.078125,
      "calib/pce": 0.672,
      "calib/std_conf": 0.174783675820522,
      "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": 0.0625,
      "completions/max_length": 3015.0,
      "completions/max_terminated_length": 3015.0,
      "completions/mean_length": 644.64453125,
      "completions/mean_terminated_length": 687.620849609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.010666666666666666,
      "grad_norm": 0.05089627206325531,
      "learning_rate": 2.5e-06,
      "loss": 0.0552,
      "num_tokens": 2701120.0,
      "reward": 0.04478124901652336,
      "reward_std": 0.08551986515522003,
      "rewards/accuracy_reward_step": 0.01171875,
      "rewards/final_brier_reward_step": 0.022874999791383743,
      "rewards/format_reward_step_strict": 0.0546875,
      "step": 10
    },
    {
      "aux_brier/lambda": 0.05000000000000001,
      "aux_brier/loss": 9.959429873617655e-08,
      "aux_brier/mean_group_std": 0.06935623024823882,
      "aux_brier/mean_r": 0.3516953115711509,
      "aux_brier/n_active_tok": 28.705882352941178,
      "aux_brier/n_groups": 5.588235294117647,
      "aux_brier/n_step_records": 7.176470588235294,
      "calib/answer_extract_rate": 0.09765625,
      "calib/auroc": 0.7,
      "calib/avg_num_step_conf": 0.51953125,
      "calib/ece": 0.6410526315789475,
      "calib/final_conf_rate": 0.07421875,
      "calib/format_rate": 0.05859375,
      "calib/frac_conf_gt_0.9": 0.7368421052631579,
      "calib/gap": 0.07571428571428584,
      "calib/mean_conf": 0.9042105263157897,
      "calib/mu_c": 0.96,
      "calib/mu_w": 0.8842857142857141,
      "calib/nonempty_final_conf_rate": 0.07421875,
      "calib/nonempty_reasoning_rate": 0.13671875,
      "calib/nonempty_step_conf_rate": 0.10546875,
      "calib/pce": 0.6410526315789475,
      "calib/std_conf": 0.1669325672074201,
      "calib/step_conf_rate": 0.10546875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.04296875,
      "completions/max_length": 2996.0,
      "completions/max_terminated_length": 2996.0,
      "completions/mean_length": 710.73828125,
      "completions/mean_terminated_length": 742.64892578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.011733333333333333,
      "grad_norm": 0.12358586490154266,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.0153,
      "num_tokens": 2987549.0,
      "reward": 0.053199902176856995,
      "reward_std": 0.08878539502620697,
      "rewards/accuracy_reward_step": 0.01953125,
      "rewards/final_brier_reward_step": 0.025299610570073128,
      "rewards/format_reward_step_strict": 0.0546875,
      "step": 11
    },
    {
      "aux_brier/lambda": 0.05000000000000001,
      "aux_brier/loss": 8.21229944788134e-08,
      "aux_brier/mean_group_std": 0.0582216399601433,
      "aux_brier/mean_r": 0.5254478349033717,
      "aux_brier/n_active_tok": 25.904761904761905,
      "aux_brier/n_groups": 4.9523809523809526,
      "aux_brier/n_step_records": 6.476190476190476,
      "calib/answer_extract_rate": 0.16796875,
      "calib/auroc": 0.611111111111111,
      "calib/avg_num_step_conf": 0.625,
      "calib/ece": 0.5332142857142858,
      "calib/final_conf_rate": 0.109375,
      "calib/format_rate": 0.09375,
      "calib/frac_conf_gt_0.9": 0.7857142857142857,
      "calib/gap": 0.042999999999999816,
      "calib/mean_conf": 0.8903571428571428,
      "calib/mu_c": 0.9179999999999999,
      "calib/mu_w": 0.8750000000000001,
      "calib/nonempty_final_conf_rate": 0.109375,
      "calib/nonempty_reasoning_rate": 0.1875,
      "calib/nonempty_step_conf_rate": 0.12890625,
      "calib/pce": 0.5332142857142858,
      "calib/std_conf": 0.1983590205168602,
      "calib/step_conf_rate": 0.12890625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2852.0,
      "completions/max_terminated_length": 2852.0,
      "completions/mean_length": 621.9765625,
      "completions/mean_terminated_length": 674.6864624023438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.0128,
      "grad_norm": 0.12016642838716507,
      "learning_rate": 3e-06,
      "loss": 0.0694,
      "num_tokens": 3250951.0,
      "reward": 0.0971623957157135,
      "reward_std": 0.17875269055366516,
      "rewards/accuracy_reward_step": 0.0390625,
      "rewards/final_brier_reward_step": 0.04489960893988609,
      "rewards/format_reward_step_strict": 0.09375,
      "step": 12
    },
    {
      "aux_brier/lambda": 0.05000000000000002,
      "aux_brier/loss": -1.3304560525284615e-07,
      "aux_brier/mean_group_std": 0.05968226273717765,
      "aux_brier/mean_r": 0.38506495852629286,
      "aux_brier/n_active_tok": 27.04,
      "aux_brier/n_groups": 4.76,
      "aux_brier/n_step_records": 6.76,
      "calib/answer_extract_rate": 0.15234375,
      "calib/auroc": 0.6538461538461539,
      "calib/avg_num_step_conf": 0.671875,
      "calib/ece": 0.7648387096774194,
      "calib/final_conf_rate": 0.12109375,
      "calib/format_rate": 0.1015625,
      "calib/frac_conf_gt_0.9": 0.8387096774193549,
      "calib/gap": 0.059461538461538455,
      "calib/mean_conf": 0.9261290322580644,
      "calib/mu_c": 0.976,
      "calib/mu_w": 0.9165384615384615,
      "calib/nonempty_final_conf_rate": 0.12109375,
      "calib/nonempty_reasoning_rate": 0.19140625,
      "calib/nonempty_step_conf_rate": 0.1640625,
      "calib/pce": 0.7648387096774194,
      "calib/std_conf": 0.1230703292078214,
      "calib/step_conf_rate": 0.1640625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05859375,
      "completions/max_length": 3043.0,
      "completions/max_terminated_length": 3043.0,
      "completions/mean_length": 671.5390625,
      "completions/mean_terminated_length": 713.3361206054688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.013866666666666666,
      "grad_norm": 0.058262571692466736,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.073,
      "num_tokens": 3527457.0,
      "reward": 0.07933026552200317,
      "reward_std": 0.1612529754638672,
      "rewards/accuracy_reward_step": 0.0234375,
      "rewards/final_brier_reward_step": 0.028258593752980232,
      "rewards/format_reward_step_strict": 0.09765625,
      "step": 13
    },
    {
      "aux_brier/lambda": 0.05000000000000002,
      "aux_brier/loss": 6.968066774248882e-08,
      "aux_brier/mean_group_std": 0.0971092681259337,
      "aux_brier/mean_r": 0.4211621161943812,
      "aux_brier/n_active_tok": 31.84,
      "aux_brier/n_groups": 5.44,
      "aux_brier/n_step_records": 7.96,
      "calib/answer_extract_rate": 0.2109375,
      "calib/auroc": 0.563063063063063,
      "calib/avg_num_step_conf": 0.80078125,
      "calib/ece": 0.7269565217391302,
      "calib/final_conf_rate": 0.1796875,
      "calib/format_rate": 0.15234375,
      "calib/frac_conf_gt_0.9": 0.7391304347826086,
      "calib/gap": 0.010570570570570759,
      "calib/mean_conf": 0.9226086956521737,
      "calib/mu_c": 0.9311111111111112,
      "calib/mu_w": 0.9205405405405405,
      "calib/nonempty_final_conf_rate": 0.1796875,
      "calib/nonempty_reasoning_rate": 0.23828125,
      "calib/nonempty_step_conf_rate": 0.1875,
      "calib/pce": 0.7269565217391302,
      "calib/std_conf": 0.09987894374014226,
      "calib/step_conf_rate": 0.1875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.09765625,
      "completions/max_length": 3046.0,
      "completions/max_terminated_length": 3046.0,
      "completions/mean_length": 616.3046875,
      "completions/mean_terminated_length": 683.0043334960938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.014933333333333333,
      "grad_norm": 0.12956944108009338,
      "learning_rate": 3.5e-06,
      "loss": 0.0887,
      "num_tokens": 3790631.0,
      "reward": 0.12147372961044312,
      "reward_std": 0.20730483531951904,
      "rewards/accuracy_reward_step": 0.03515625,
      "rewards/final_brier_reward_step": 0.04839491844177246,
      "rewards/format_reward_step_strict": 0.1484375,
      "step": 14
    },
    {
      "aux_brier/lambda": 0.05000000000000002,
      "aux_brier/loss": 3.00324828937339e-08,
      "aux_brier/mean_group_std": 0.06625814184144391,
      "aux_brier/mean_r": 0.4505305222592945,
      "aux_brier/n_active_tok": 42.0,
      "aux_brier/n_groups": 6.269230769230769,
      "aux_brier/n_step_records": 10.5,
      "calib/answer_extract_rate": 0.21875,
      "calib/auroc": 0.42647058823529405,
      "calib/avg_num_step_conf": 1.08203125,
      "calib/ece": 0.7490476190476191,
      "calib/final_conf_rate": 0.1640625,
      "calib/format_rate": 0.12109375,
      "calib/frac_conf_gt_0.9": 0.8333333333333334,
      "calib/gap": -0.04117647058823537,
      "calib/mean_conf": 0.9133333333333333,
      "calib/mu_c": 0.8799999999999999,
      "calib/mu_w": 0.9211764705882353,
      "calib/nonempty_final_conf_rate": 0.1640625,
      "calib/nonempty_reasoning_rate": 0.2734375,
      "calib/nonempty_step_conf_rate": 0.2109375,
      "calib/pce": 0.735952380952381,
      "calib/std_conf": 0.13688136446027974,
      "calib/step_conf_rate": 0.2109375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1015625,
      "completions/max_length": 2831.0,
      "completions/max_terminated_length": 2831.0,
      "completions/mean_length": 545.08984375,
      "completions/mean_terminated_length": 606.7086791992188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 5.0,
      "epoch": 0.016,
      "grad_norm": 0.1261015087366104,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.0349,
      "num_tokens": 4038054.0,
      "reward": 0.11347460746765137,
      "reward_std": 0.22185897827148438,
      "rewards/accuracy_reward_step": 0.04296875,
      "rewards/final_brier_reward_step": 0.039835937321186066,
      "rewards/format_reward_step_strict": 0.12109375,
      "step": 15
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.0546628842286839e-08,
      "aux_brier/mean_group_std": 0.15131436740359547,
      "aux_brier/mean_r": 0.541980853761356,
      "aux_brier/n_active_tok": 47.58620689655172,
      "aux_brier/n_groups": 6.758620689655173,
      "aux_brier/n_step_records": 11.89655172413793,
      "calib/answer_extract_rate": 0.2578125,
      "calib/auroc": 0.4716599190283401,
      "calib/avg_num_step_conf": 1.3984375,
      "calib/ece": 0.6116379310344826,
      "calib/final_conf_rate": 0.2265625,
      "calib/format_rate": 0.1875,
      "calib/frac_conf_gt_0.9": 0.8448275862068966,
      "calib/gap": 0.025310391363022933,
      "calib/mean_conf": 0.9150862068965516,
      "calib/mu_c": 0.9321052631578945,
      "calib/mu_w": 0.9067948717948716,
      "calib/nonempty_final_conf_rate": 0.2265625,
      "calib/nonempty_reasoning_rate": 0.34375,
      "calib/nonempty_step_conf_rate": 0.28125,
      "calib/pce": 0.5995689655172411,
      "calib/std_conf": 0.16442294961918918,
      "calib/step_conf_rate": 0.28125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.1015625,
      "completions/max_length": 2951.0,
      "completions/max_terminated_length": 2951.0,
      "completions/mean_length": 597.69921875,
      "completions/mean_terminated_length": 665.2651977539062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 3.0,
      "epoch": 0.017066666666666667,
      "grad_norm": 0.2648871839046478,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0331,
      "num_tokens": 4299913.0,
      "reward": 0.1786659061908722,
      "reward_std": 0.3179003596305847,
      "rewards/accuracy_reward_step": 0.07421875,
      "rewards/final_brier_reward_step": 0.074038565158844,
      "rewards/format_reward_step_strict": 0.171875,
      "step": 16
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.06125258872936e-08,
      "aux_brier/mean_group_std": 0.123763447321381,
      "aux_brier/mean_r": 0.5180065285588968,
      "aux_brier/n_active_tok": 56.375,
      "aux_brier/n_groups": 7.71875,
      "aux_brier/n_step_records": 14.09375,
      "calib/answer_extract_rate": 0.375,
      "calib/auroc": 0.41557734204793023,
      "calib/avg_num_step_conf": 1.8203125,
      "calib/ece": 0.696056338028169,
      "calib/final_conf_rate": 0.27734375,
      "calib/format_rate": 0.2265625,
      "calib/frac_conf_gt_0.9": 0.6901408450704225,
      "calib/gap": -0.08327886710239651,
      "calib/mean_conf": 0.8709859154929577,
      "calib/mu_c": 0.8076470588235294,
      "calib/mu_w": 0.8909259259259259,
      "calib/nonempty_final_conf_rate": 0.27734375,
      "calib/nonempty_reasoning_rate": 0.4375,
      "calib/nonempty_step_conf_rate": 0.328125,
      "calib/pce": 0.6638028169014084,
      "calib/std_conf": 0.2130737816375937,
      "calib/step_conf_rate": 0.328125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3022.0,
      "completions/max_terminated_length": 3022.0,
      "completions/mean_length": 535.3828125,
      "completions/mean_terminated_length": 557.1463012695312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.018133333333333335,
      "grad_norm": 0.15465809404850006,
      "learning_rate": 4.25e-06,
      "loss": 0.083,
      "num_tokens": 4540499.0,
      "reward": 0.19698895514011383,
      "reward_std": 0.3120965361595154,
      "rewards/accuracy_reward_step": 0.06640625,
      "rewards/final_brier_reward_step": 0.0770183578133583,
      "rewards/format_reward_step_strict": 0.22265625,
      "step": 17
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.4943371377583044e-07,
      "aux_brier/mean_group_std": 0.1019210505216714,
      "aux_brier/mean_r": 0.5275322096514212,
      "aux_brier/n_active_tok": 42.064516129032256,
      "aux_brier/n_groups": 6.419354838709677,
      "aux_brier/n_step_records": 10.516129032258064,
      "calib/answer_extract_rate": 0.2734375,
      "calib/auroc": 0.7057823129251701,
      "calib/avg_num_step_conf": 1.2890625,
      "calib/ece": 0.6642622950819672,
      "calib/final_conf_rate": 0.23828125,
      "calib/format_rate": 0.17578125,
      "calib/frac_conf_gt_0.9": 0.5901639344262295,
      "calib/gap": 0.10355442176870766,
      "calib/mean_conf": 0.860983606557377,
      "calib/mu_c": 0.9441666666666668,
      "calib/mu_w": 0.8406122448979592,
      "calib/nonempty_final_conf_rate": 0.23828125,
      "calib/nonempty_reasoning_rate": 0.34375,
      "calib/nonempty_step_conf_rate": 0.27734375,
      "calib/pce": 0.6642622950819672,
      "calib/std_conf": 0.21424442084311687,
      "calib/step_conf_rate": 0.27734375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.05078125,
      "completions/max_length": 3057.0,
      "completions/max_terminated_length": 3057.0,
      "completions/mean_length": 618.37890625,
      "completions/mean_terminated_length": 651.4608764648438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 6.0,
      "epoch": 0.0192,
      "grad_norm": 0.2550928592681885,
      "learning_rate": 4.5e-06,
      "loss": 0.1194,
      "num_tokens": 4809524.0,
      "reward": 0.1505630910396576,
      "reward_std": 0.2417924702167511,
      "rewards/accuracy_reward_step": 0.046875,
      "rewards/final_brier_reward_step": 0.07100234925746918,
      "rewards/format_reward_step_strict": 0.171875,
      "step": 18
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.1574589175999961e-08,
      "aux_brier/mean_group_std": 0.15655222438005406,
      "aux_brier/mean_r": 0.5297212505808176,
      "aux_brier/n_active_tok": 86.75,
      "aux_brier/n_groups": 7.53125,
      "aux_brier/n_step_records": 21.6875,
      "calib/answer_extract_rate": 0.578125,
      "calib/auroc": 0.4634055265123226,
      "calib/avg_num_step_conf": 2.79296875,
      "calib/ece": 0.6772984496124032,
      "calib/final_conf_rate": 0.50390625,
      "calib/format_rate": 0.42578125,
      "calib/frac_conf_gt_0.9": 0.6511627906976745,
      "calib/gap": -0.00889880507841656,
      "calib/mean_conf": 0.8351821705426357,
      "calib/mu_c": 0.8280769230769229,
      "calib/mu_w": 0.8369757281553395,
      "calib/nonempty_final_conf_rate": 0.50390625,
      "calib/nonempty_reasoning_rate": 0.64453125,
      "calib/nonempty_step_conf_rate": 0.53125,
      "calib/pce": 0.6554651162790698,
      "calib/std_conf": 0.2665359101275789,
      "calib/step_conf_rate": 0.53125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.04296875,
      "completions/max_length": 2936.0,
      "completions/max_terminated_length": 2936.0,
      "completions/mean_length": 462.88671875,
      "completions/mean_terminated_length": 483.66937255859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 8.0,
      "epoch": 0.020266666666666665,
      "grad_norm": 0.02498788572847843,
      "learning_rate": 4.75e-06,
      "loss": 0.1313,
      "num_tokens": 5032783.0,
      "reward": 0.34538060426712036,
      "reward_std": 0.4372578263282776,
      "rewards/accuracy_reward_step": 0.10546875,
      "rewards/final_brier_reward_step": 0.14714746177196503,
      "rewards/format_reward_step_strict": 0.40625,
      "step": 19
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.524534731789373e-08,
      "aux_brier/mean_group_std": 0.18416291747432723,
      "aux_brier/mean_r": 0.4724621196341125,
      "aux_brier/n_active_tok": 98.25,
      "aux_brier/n_groups": 7.5,
      "aux_brier/n_step_records": 24.5625,
      "calib/answer_extract_rate": 0.68359375,
      "calib/auroc": 0.5297591477535896,
      "calib/avg_num_step_conf": 3.16796875,
      "calib/ece": 0.7053416149068322,
      "calib/final_conf_rate": 0.62890625,
      "calib/format_rate": 0.546875,
      "calib/frac_conf_gt_0.9": 0.782608695652174,
      "calib/gap": 0.0243816581750812,
      "calib/mean_conf": 0.9078260869565217,
      "calib/mu_c": 0.9270588235294118,
      "calib/mu_w": 0.9026771653543306,
      "calib/nonempty_final_conf_rate": 0.62890625,
      "calib/nonempty_reasoning_rate": 0.79296875,
      "calib/nonempty_step_conf_rate": 0.69140625,
      "calib/pce": 0.7009937888198756,
      "calib/std_conf": 0.15433164117749484,
      "calib/step_conf_rate": 0.69140625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3072.0,
      "completions/max_terminated_length": 3072.0,
      "completions/mean_length": 368.48046875,
      "completions/mean_terminated_length": 377.3240051269531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 23.0,
      "epoch": 0.021333333333333333,
      "grad_norm": 0.12939509749412537,
      "learning_rate": 5e-06,
      "loss": 0.0943,
      "num_tokens": 5231986.0,
      "reward": 0.4509944021701813,
      "reward_std": 0.4627910852432251,
      "rewards/accuracy_reward_step": 0.140625,
      "rewards/final_brier_reward_step": 0.18679022789001465,
      "rewards/format_reward_step_strict": 0.52734375,
      "step": 20
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.6855419965897696e-09,
      "aux_brier/mean_group_std": 0.2078869918818616,
      "aux_brier/mean_r": 0.4831021750609932,
      "aux_brier/n_active_tok": 121.0,
      "aux_brier/n_groups": 8.4375,
      "aux_brier/n_step_records": 30.25,
      "calib/answer_extract_rate": 0.79296875,
      "calib/auroc": 0.5836913049918531,
      "calib/avg_num_step_conf": 3.85546875,
      "calib/ece": 0.6835785000000001,
      "calib/final_conf_rate": 0.78125,
      "calib/format_rate": 0.70703125,
      "calib/frac_conf_gt_0.9": 0.72,
      "calib/gap": 0.04802620352540332,
      "calib/mean_conf": 0.8985785000000002,
      "calib/mu_c": 0.9362790697674418,
      "calib/mu_w": 0.8882528662420385,
      "calib/nonempty_final_conf_rate": 0.78125,
      "calib/nonempty_reasoning_rate": 0.8671875,
      "calib/nonempty_step_conf_rate": 0.80078125,
      "calib/pce": 0.6835785000000001,
      "calib/std_conf": 0.15646988460323602,
      "calib/step_conf_rate": 0.80078125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2907.0,
      "completions/max_terminated_length": 2907.0,
      "completions/mean_length": 335.38671875,
      "completions/mean_terminated_length": 339.3636474609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 2.0,
      "epoch": 0.0224,
      "grad_norm": 0.044533684849739075,
      "learning_rate": 4.9722222222222224e-06,
      "loss": 0.0741,
      "num_tokens": 5420805.0,
      "reward": 0.5556639432907104,
      "reward_std": 0.49617576599121094,
      "rewards/accuracy_reward_step": 0.16796875,
      "rewards/final_brier_reward_step": 0.2460932433605194,
      "rewards/format_reward_step_strict": 0.65234375,
      "step": 21
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.322772361233575e-09,
      "aux_brier/mean_group_std": 0.2046128928256343,
      "aux_brier/mean_r": 0.5185869317094062,
      "aux_brier/n_active_tok": 132.625,
      "aux_brier/n_groups": 9.46875,
      "aux_brier/n_step_records": 33.15625,
      "calib/answer_extract_rate": 0.84765625,
      "calib/auroc": 0.5711273006134969,
      "calib/avg_num_step_conf": 4.18359375,
      "calib/ece": 0.6619905213270142,
      "calib/final_conf_rate": 0.82421875,
      "calib/format_rate": 0.7421875,
      "calib/frac_conf_gt_0.9": 0.7251184834123223,
      "calib/gap": 0.040318251533742244,
      "calib/mean_conf": 0.889478672985782,
      "calib/mu_c": 0.9206249999999999,
      "calib/mu_w": 0.8803067484662577,
      "calib/nonempty_final_conf_rate": 0.82421875,
      "calib/nonempty_reasoning_rate": 0.921875,
      "calib/nonempty_step_conf_rate": 0.84765625,
      "calib/pce": 0.6619905213270142,
      "calib/std_conf": 0.1804344763509982,
      "calib/step_conf_rate": 0.84765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2683.0,
      "completions/max_terminated_length": 2683.0,
      "completions/mean_length": 329.69140625,
      "completions/mean_terminated_length": 330.9843444824219,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 24.0,
      "epoch": 0.023466666666666667,
      "grad_norm": 0.07365760207176208,
      "learning_rate": 4.944444444444445e-06,
      "loss": 0.1207,
      "num_tokens": 5607022.0,
      "reward": 0.6302892565727234,
      "reward_std": 0.49790266156196594,
      "rewards/accuracy_reward_step": 0.1953125,
      "rewards/final_brier_reward_step": 0.26334452629089355,
      "rewards/format_reward_step_strict": 0.73828125,
      "step": 22
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.4648843964237823e-08,
      "aux_brier/mean_group_std": 0.21768981241071053,
      "aux_brier/mean_r": 0.5264782417382557,
      "aux_brier/n_active_tok": 125.875,
      "aux_brier/n_groups": 7.625,
      "aux_brier/n_step_records": 31.46875,
      "calib/answer_extract_rate": 0.89453125,
      "calib/auroc": 0.4203296703296703,
      "calib/avg_num_step_conf": 3.984375,
      "calib/ece": 0.6808136363636363,
      "calib/final_conf_rate": 0.859375,
      "calib/format_rate": 0.80078125,
      "calib/frac_conf_gt_0.9": 0.6590909090909091,
      "calib/gap": -0.05642261904761903,
      "calib/mean_conf": 0.8780863636363636,
      "calib/mu_c": 0.8350000000000001,
      "calib/mu_w": 0.8914226190476191,
      "calib/nonempty_final_conf_rate": 0.859375,
      "calib/nonempty_reasoning_rate": 0.95703125,
      "calib/nonempty_step_conf_rate": 0.90625,
      "calib/pce": 0.6612681818181817,
      "calib/std_conf": 0.18692190103563802,
      "calib/step_conf_rate": 0.90625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2147.0,
      "completions/max_terminated_length": 2147.0,
      "completions/mean_length": 318.7109375,
      "completions/mean_terminated_length": 321.220458984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 24.0,
      "epoch": 0.024533333333333334,
      "grad_norm": 0.10605355352163315,
      "learning_rate": 4.9166666666666665e-06,
      "loss": 0.1017,
      "num_tokens": 5792548.0,
      "reward": 0.6638668775558472,
      "reward_std": 0.45015454292297363,
      "rewards/accuracy_reward_step": 0.20703125,
      "rewards/final_brier_reward_step": 0.28046756982803345,
      "rewards/format_reward_step_strict": 0.7734375,
      "step": 23
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.2244822426605886e-09,
      "aux_brier/mean_group_std": 0.20294305429502568,
      "aux_brier/mean_r": 0.5130447519385217,
      "aux_brier/n_active_tok": 148.75,
      "aux_brier/n_groups": 9.0625,
      "aux_brier/n_step_records": 37.1875,
      "calib/answer_extract_rate": 0.875,
      "calib/auroc": 0.4318121693121693,
      "calib/avg_num_step_conf": 4.75390625,
      "calib/ece": 0.7022972972972974,
      "calib/final_conf_rate": 0.8671875,
      "calib/format_rate": 0.796875,
      "calib/frac_conf_gt_0.9": 0.7072072072072072,
      "calib/gap": -0.010209523809523668,
      "calib/mean_conf": 0.8868018018018017,
      "calib/mu_c": 0.8785238095238096,
      "calib/mu_w": 0.8887333333333333,
      "calib/nonempty_final_conf_rate": 0.8671875,
      "calib/nonempty_reasoning_rate": 0.9609375,
      "calib/nonempty_step_conf_rate": 0.9140625,
      "calib/pce": 0.6999549549549551,
      "calib/std_conf": 0.17966138419183594,
      "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": 2745.0,
      "completions/max_terminated_length": 2745.0,
      "completions/mean_length": 325.76171875,
      "completions/mean_terminated_length": 328.3267822265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 15.0,
      "epoch": 0.0256,
      "grad_norm": 0.21002741158008575,
      "learning_rate": 4.888888888888889e-06,
      "loss": 0.1087,
      "num_tokens": 5980455.0,
      "reward": 0.610745370388031,
      "reward_std": 0.43826824426651,
      "rewards/accuracy_reward_step": 0.16796875,
      "rewards/final_brier_reward_step": 0.263293981552124,
      "rewards/format_reward_step_strict": 0.75390625,
      "step": 24
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.988896718846389e-08,
      "aux_brier/mean_group_std": 0.20363710790214565,
      "aux_brier/mean_r": 0.49830946569913254,
      "aux_brier/n_active_tok": 150.875,
      "aux_brier/n_groups": 8.125,
      "aux_brier/n_step_records": 37.71875,
      "calib/answer_extract_rate": 0.91796875,
      "calib/auroc": 0.6020542400323822,
      "calib/avg_num_step_conf": 4.765625,
      "calib/ece": 0.6479831223628691,
      "calib/final_conf_rate": 0.92578125,
      "calib/format_rate": 0.87890625,
      "calib/frac_conf_gt_0.9": 0.6582278481012658,
      "calib/gap": 0.05367668488160293,
      "calib/mean_conf": 0.875831223628692,
      "calib/mu_c": 0.9172777777777777,
      "calib/mu_w": 0.8636010928961748,
      "calib/nonempty_final_conf_rate": 0.92578125,
      "calib/nonempty_reasoning_rate": 0.96875,
      "calib/nonempty_step_conf_rate": 0.95703125,
      "calib/pce": 0.6479831223628691,
      "calib/std_conf": 0.1821509001281357,
      "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": 0.0,
      "completions/max_length": 1848.0,
      "completions/max_terminated_length": 1848.0,
      "completions/mean_length": 280.63671875,
      "completions/mean_terminated_length": 280.63671875,
      "completions/min_length": 6.0,
      "completions/min_terminated_length": 6.0,
      "epoch": 0.02666666666666667,
      "grad_norm": 0.008940259926021099,
      "learning_rate": 4.861111111111111e-06,
      "loss": 0.0274,
      "num_tokens": 6155522.0,
      "reward": 0.7232168316841125,
      "reward_std": 0.4639264941215515,
      "rewards/accuracy_reward_step": 0.2109375,
      "rewards/final_brier_reward_step": 0.33817988634109497,
      "rewards/format_reward_step_strict": 0.85546875,
      "step": 25
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 8.733723841358731e-09,
      "aux_brier/mean_group_std": 0.1880441359486682,
      "aux_brier/mean_r": 0.5253828549221222,
      "aux_brier/n_active_tok": 154.125,
      "aux_brier/n_groups": 8.25,
      "aux_brier/n_step_records": 38.53125,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.4772074659009333,
      "calib/avg_num_step_conf": 4.86328125,
      "calib/ece": 0.7233775933609959,
      "calib/final_conf_rate": 0.94140625,
      "calib/format_rate": 0.9140625,
      "calib/frac_conf_gt_0.9": 0.7012448132780082,
      "calib/gap": 0.010950227327111817,
      "calib/mean_conf": 0.8897676348547717,
      "calib/mu_c": 0.8988095238095238,
      "calib/mu_w": 0.887859296482412,
      "calib/nonempty_final_conf_rate": 0.94140625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.719435684647303,
      "calib/std_conf": 0.17055198386843828,
      "calib/step_conf_rate": 0.984375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2863.0,
      "completions/max_terminated_length": 2863.0,
      "completions/mean_length": 283.8046875,
      "completions/mean_terminated_length": 283.8046875,
      "completions/min_length": 57.0,
      "completions/min_terminated_length": 57.0,
      "epoch": 0.027733333333333332,
      "grad_norm": 0.11429033428430557,
      "learning_rate": 4.833333333333333e-06,
      "loss": 0.0064,
      "num_tokens": 6333416.0,
      "reward": 0.68150794506073,
      "reward_std": 0.4080682396888733,
      "rewards/accuracy_reward_step": 0.1640625,
      "rewards/final_brier_reward_step": 0.2963443994522095,
      "rewards/format_reward_step_strict": 0.88671875,
      "step": 26
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.146805901630632e-08,
      "aux_brier/mean_group_std": 0.20170100575564176,
      "aux_brier/mean_r": 0.5008420967947467,
      "aux_brier/n_active_tok": 158.0,
      "aux_brier/n_groups": 8.4375,
      "aux_brier/n_step_records": 39.5,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.4906553398058252,
      "calib/avg_num_step_conf": 4.96484375,
      "calib/ece": 0.7226029810298104,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.91796875,
      "calib/frac_conf_gt_0.9": 0.6829268292682927,
      "calib/gap": 0.0549862459546927,
      "calib/mean_conf": 0.8852046070460704,
      "calib/mu_c": 0.93125,
      "calib/mu_w": 0.8762637540453073,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.7226029810298104,
      "calib/std_conf": 0.17142669879331107,
      "calib/step_conf_rate": 0.9765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2698.0,
      "completions/max_terminated_length": 2698.0,
      "completions/mean_length": 271.6875,
      "completions/mean_terminated_length": 271.6875,
      "completions/min_length": 44.0,
      "completions/min_terminated_length": 44.0,
      "epoch": 0.0288,
      "grad_norm": 0.2054324895143509,
      "learning_rate": 4.805555555555556e-06,
      "loss": 0.0588,
      "num_tokens": 6508184.0,
      "reward": 0.690642237663269,
      "reward_std": 0.35970115661621094,
      "rewards/accuracy_reward_step": 0.16796875,
      "rewards/final_brier_reward_step": 0.30944421887397766,
      "rewards/format_reward_step_strict": 0.890625,
      "step": 27
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -9.251436984547556e-09,
      "aux_brier/mean_group_std": 0.21422411978879408,
      "aux_brier/mean_r": 0.5064897092037213,
      "aux_brier/n_active_tok": 154.25,
      "aux_brier/n_groups": 9.34375,
      "aux_brier/n_step_records": 38.5625,
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.5271692745376956,
      "calib/avg_num_step_conf": 4.86328125,
      "calib/ece": 0.6659421487603305,
      "calib/final_conf_rate": 0.9453125,
      "calib/format_rate": 0.90234375,
      "calib/frac_conf_gt_0.9": 0.7066115702479339,
      "calib/gap": 0.01013504030346124,
      "calib/mean_conf": 0.8903223140495867,
      "calib/mu_c": 0.8980701754385964,
      "calib/mu_w": 0.8879351351351351,
      "calib/nonempty_final_conf_rate": 0.9453125,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.95703125,
      "calib/pce": 0.6603636363636363,
      "calib/std_conf": 0.16952703847708325,
      "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": 0.0,
      "completions/max_length": 1584.0,
      "completions/max_terminated_length": 1584.0,
      "completions/mean_length": 283.50390625,
      "completions/mean_terminated_length": 283.50390625,
      "completions/min_length": 40.0,
      "completions/min_terminated_length": 40.0,
      "epoch": 0.029866666666666666,
      "grad_norm": 0.1324453055858612,
      "learning_rate": 4.777777777777778e-06,
      "loss": 0.0586,
      "num_tokens": 6687705.0,
      "reward": 0.7451667785644531,
      "reward_std": 0.43373051285743713,
      "rewards/accuracy_reward_step": 0.22265625,
      "rewards/final_brier_reward_step": 0.3400421738624573,
      "rewards/format_reward_step_strict": 0.875,
      "step": 28
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.774421712341656e-08,
      "aux_brier/mean_group_std": 0.19610744163840346,
      "aux_brier/mean_r": 0.5103295116870583,
      "aux_brier/n_active_tok": 169.875,
      "aux_brier/n_groups": 8.90625,
      "aux_brier/n_step_records": 42.46875,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5384146341463415,
      "calib/avg_num_step_conf": 5.484375,
      "calib/ece": 0.7245673469387756,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.9140625,
      "calib/frac_conf_gt_0.9": 0.6816326530612244,
      "calib/gap": 0.053980487804878186,
      "calib/mean_conf": 0.8878326530612244,
      "calib/mu_c": 0.933,
      "calib/mu_w": 0.8790195121951219,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.7245673469387756,
      "calib/std_conf": 0.16122579095315817,
      "calib/step_conf_rate": 0.9765625,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 1917.0,
      "completions/max_terminated_length": 1917.0,
      "completions/mean_length": 303.91015625,
      "completions/mean_terminated_length": 306.30316162109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 48.0,
      "epoch": 0.030933333333333334,
      "grad_norm": 0.34857821464538574,
      "learning_rate": 4.75e-06,
      "loss": -0.0121,
      "num_tokens": 6872634.0,
      "reward": 0.680101752281189,
      "reward_std": 0.38362351059913635,
      "rewards/accuracy_reward_step": 0.16015625,
      "rewards/final_brier_reward_step": 0.2985318899154663,
      "rewards/format_reward_step_strict": 0.890625,
      "step": 29
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.349526157657822e-08,
      "aux_brier/mean_group_std": 0.18645527569371767,
      "aux_brier/mean_r": 0.5258975753340903,
      "aux_brier/n_active_tok": 172.125,
      "aux_brier/n_groups": 10.78125,
      "aux_brier/n_step_records": 43.03125,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5746729461015174,
      "calib/avg_num_step_conf": 5.41796875,
      "calib/ece": 0.6881557377049181,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.9296875,
      "calib/frac_conf_gt_0.9": 0.6475409836065574,
      "calib/gap": 0.02557299843014127,
      "calib/mean_conf": 0.882827868852459,
      "calib/mu_c": 0.903265306122449,
      "calib/mu_w": 0.8776923076923078,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.6850819672131149,
      "calib/std_conf": 0.16179258086614187,
      "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": 2010.0,
      "completions/max_terminated_length": 2010.0,
      "completions/mean_length": 304.546875,
      "completions/mean_terminated_length": 304.546875,
      "completions/min_length": 59.0,
      "completions/min_terminated_length": 59.0,
      "epoch": 0.032,
      "grad_norm": 0.10014064610004425,
      "learning_rate": 4.722222222222222e-06,
      "loss": 0.0001,
      "num_tokens": 7057582.0,
      "reward": 0.7322098612785339,
      "reward_std": 0.4103948175907135,
      "rewards/accuracy_reward_step": 0.1953125,
      "rewards/final_brier_reward_step": 0.33508944511413574,
      "rewards/format_reward_step_strict": 0.90625,
      "step": 30
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 8.70296445326657e-09,
      "aux_brier/mean_group_std": 0.2190587943030789,
      "aux_brier/mean_r": 0.5495102326387148,
      "aux_brier/n_active_tok": 175.125,
      "aux_brier/n_groups": 9.6875,
      "aux_brier/n_step_records": 43.78125,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.524795918367347,
      "calib/avg_num_step_conf": 5.5,
      "calib/ece": 0.6690361445783132,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.5823293172690763,
      "calib/gap": 0.03365714285714294,
      "calib/mean_conf": 0.8658232931726908,
      "calib/mu_c": 0.8928571428571429,
      "calib/mu_w": 0.8592,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.6690361445783132,
      "calib/std_conf": 0.17100525162676766,
      "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": 1299.0,
      "completions/max_terminated_length": 1299.0,
      "completions/mean_length": 294.66796875,
      "completions/mean_terminated_length": 294.66796875,
      "completions/min_length": 98.0,
      "completions/min_terminated_length": 98.0,
      "epoch": 0.03306666666666667,
      "grad_norm": 0.3468322455883026,
      "learning_rate": 4.694444444444445e-06,
      "loss": 0.0431,
      "num_tokens": 7238929.0,
      "reward": 0.7521989941596985,
      "reward_std": 0.38824474811553955,
      "rewards/accuracy_reward_step": 0.19140625,
      "rewards/final_brier_reward_step": 0.3603585958480835,
      "rewards/format_reward_step_strict": 0.94140625,
      "step": 31
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.008295861363976e-10,
      "aux_brier/mean_group_std": 0.20375160642834253,
      "aux_brier/mean_r": 0.6171608731266107,
      "aux_brier/n_active_tok": 158.5,
      "aux_brier/n_groups": 7.75,
      "aux_brier/n_step_records": 39.625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5288600288600288,
      "calib/avg_num_step_conf": 4.9765625,
      "calib/ece": 0.6176574803149606,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.5078740157480315,
      "calib/gap": 0.053775252525252504,
      "calib/mean_conf": 0.8318307086614174,
      "calib/mu_c": 0.87375,
      "calib/mu_w": 0.8199747474747475,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.614507874015748,
      "calib/std_conf": 0.2009100723388808,
      "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": 803.0,
      "completions/max_terminated_length": 803.0,
      "completions/mean_length": 263.7734375,
      "completions/mean_terminated_length": 264.807861328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 72.0,
      "epoch": 0.034133333333333335,
      "grad_norm": 0.08650796115398407,
      "learning_rate": 4.666666666666667e-06,
      "loss": 0.0278,
      "num_tokens": 7413159.0,
      "reward": 0.8118621706962585,
      "reward_std": 0.3442915380001068,
      "rewards/accuracy_reward_step": 0.21875,
      "rewards/final_brier_reward_step": 0.42713624238967896,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 32
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 7.882408420423204e-09,
      "aux_brier/mean_group_std": 0.20335982369599354,
      "aux_brier/mean_r": 0.5829862248054526,
      "aux_brier/n_active_tok": 172.375,
      "aux_brier/n_groups": 8.75,
      "aux_brier/n_step_records": 43.09375,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5207657770017461,
      "calib/avg_num_step_conf": 5.41015625,
      "calib/ece": 0.7021686746987952,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.5461847389558233,
      "calib/gap": 6.734846595157773e-05,
      "calib/mean_conf": 0.8520481927710845,
      "calib/mu_c": 0.8521052631578946,
      "calib/mu_w": 0.852037914691943,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.7008032128514057,
      "calib/std_conf": 0.1779767529344418,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 916.0,
      "completions/max_terminated_length": 916.0,
      "completions/mean_length": 299.81640625,
      "completions/mean_terminated_length": 302.1771545410156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 68.0,
      "epoch": 0.0352,
      "grad_norm": 0.06140900403261185,
      "learning_rate": 4.638888888888889e-06,
      "loss": 0.0056,
      "num_tokens": 7596784.0,
      "reward": 0.7030861377716064,
      "reward_std": 0.3202725648880005,
      "rewards/accuracy_reward_step": 0.1484375,
      "rewards/final_brier_reward_step": 0.3357820212841034,
      "rewards/format_reward_step_strict": 0.94140625,
      "step": 33
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 4.224410730946371e-09,
      "aux_brier/mean_group_std": 0.1973998358719577,
      "aux_brier/mean_r": 0.675027915675693,
      "aux_brier/n_active_tok": 173.5,
      "aux_brier/n_groups": 9.71875,
      "aux_brier/n_step_records": 43.375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5004618226600985,
      "calib/avg_num_step_conf": 5.46484375,
      "calib/ece": 0.5999203187250997,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.42231075697211157,
      "calib/gap": 0.016860632183907986,
      "calib/mean_conf": 0.7911553784860557,
      "calib/mu_c": 0.8047916666666666,
      "calib/mu_w": 0.7879310344827586,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.5999203187250997,
      "calib/std_conf": 0.2017015757457608,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 274.69921875,
      "completions/mean_terminated_length": 275.7764892578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 97.0,
      "epoch": 0.03626666666666667,
      "grad_norm": 1.1419422626495361,
      "learning_rate": 4.611111111111112e-06,
      "loss": 0.0116,
      "num_tokens": 7772219.0,
      "reward": 0.78791344165802,
      "reward_std": 0.3862859606742859,
      "rewards/accuracy_reward_step": 0.19140625,
      "rewards/final_brier_reward_step": 0.44071638584136963,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 34
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.123872044183631e-08,
      "aux_brier/mean_group_std": 0.19222505546668744,
      "aux_brier/mean_r": 0.6863559674116386,
      "aux_brier/n_active_tok": 180.5,
      "aux_brier/n_groups": 9.625,
      "aux_brier/n_step_records": 45.125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6022884283246979,
      "calib/avg_num_step_conf": 5.6640625,
      "calib/ece": 0.53300395256917,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.3438735177865613,
      "calib/gap": 0.08325215889464588,
      "calib/mean_conf": 0.7701581027667985,
      "calib/mu_c": 0.8336666666666666,
      "calib/mu_w": 0.7504145077720207,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.53300395256917,
      "calib/std_conf": 0.22595947095308164,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 887.0,
      "completions/max_terminated_length": 887.0,
      "completions/mean_length": 310.3203125,
      "completions/mean_terminated_length": 311.5372619628906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 27.0,
      "epoch": 0.037333333333333336,
      "grad_norm": 0.09675668179988861,
      "learning_rate": 4.583333333333333e-06,
      "loss": 0.0073,
      "num_tokens": 7960917.0,
      "reward": 0.8488951921463013,
      "reward_std": 0.3726450204849243,
      "rewards/accuracy_reward_step": 0.234375,
      "rewards/final_brier_reward_step": 0.5049558877944946,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 35
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 4.84285989088562e-09,
      "aux_brier/mean_group_std": 0.1746511756104062,
      "aux_brier/mean_r": 0.7607258928054881,
      "aux_brier/n_active_tok": 173.0,
      "aux_brier/n_groups": 9.40625,
      "aux_brier/n_step_records": 43.25,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5042504665146175,
      "calib/avg_num_step_conf": 5.45703125,
      "calib/ece": 0.37484000000000006,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.212,
      "calib/gap": 0.0037231322136981326,
      "calib/mean_conf": 0.68972,
      "calib/mu_c": 0.6920879120879121,
      "calib/mu_w": 0.688364779874214,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.35028000000000004,
      "calib/std_conf": 0.2423491728890363,
      "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": 1628.0,
      "completions/max_terminated_length": 1628.0,
      "completions/mean_length": 304.76171875,
      "completions/mean_terminated_length": 305.9568786621094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 7.0,
      "epoch": 0.0384,
      "grad_norm": 0.5897600054740906,
      "learning_rate": 4.555555555555556e-06,
      "loss": 0.0127,
      "num_tokens": 8141648.0,
      "reward": 0.9805706739425659,
      "reward_std": 0.4639264941215515,
      "rewards/accuracy_reward_step": 0.35546875,
      "rewards/final_brier_reward_step": 0.5863453149795532,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 36
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.0809664130760908e-08,
      "aux_brier/mean_group_std": 0.18644022102253568,
      "aux_brier/mean_r": 0.7817912325901277,
      "aux_brier/n_active_tok": 164.875,
      "aux_brier/n_groups": 8.8125,
      "aux_brier/n_step_records": 41.21875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.49517835808158395,
      "calib/avg_num_step_conf": 5.24609375,
      "calib/ece": 0.40549800796812757,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.18725099601593626,
      "calib/gap": -0.00244410308926446,
      "calib/mean_conf": 0.641195219123506,
      "calib/mu_c": 0.6393548387096774,
      "calib/mu_w": 0.6417989417989418,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.39984063745019927,
      "calib/std_conf": 0.25375931623243464,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2081.0,
      "completions/max_terminated_length": 2081.0,
      "completions/mean_length": 302.0078125,
      "completions/mean_terminated_length": 303.1921691894531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 69.0,
      "epoch": 0.039466666666666664,
      "grad_norm": 0.08607197552919388,
      "learning_rate": 4.527777777777778e-06,
      "loss": 0.0528,
      "num_tokens": 8326058.0,
      "reward": 0.870861291885376,
      "reward_std": 0.3708493709564209,
      "rewards/accuracy_reward_step": 0.2421875,
      "rewards/final_brier_reward_step": 0.5771952867507935,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 37
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.5499739727808937e-08,
      "aux_brier/mean_group_std": 0.15065610893712256,
      "aux_brier/mean_r": 0.84442506153201,
      "aux_brier/n_active_tok": 163.625,
      "aux_brier/n_groups": 8.625,
      "aux_brier/n_step_records": 40.90625,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.4324565155403478,
      "calib/avg_num_step_conf": 5.1328125,
      "calib/ece": 0.31868525896414346,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.10756972111553785,
      "calib/gap": -0.05270958083832333,
      "calib/mean_conf": 0.5775697211155378,
      "calib/mu_c": 0.5425,
      "calib/mu_w": 0.5952095808383233,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.28079681274900403,
      "calib/std_conf": 0.2625375520844664,
      "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": 728.0,
      "completions/max_terminated_length": 728.0,
      "completions/mean_length": 277.3203125,
      "completions/mean_terminated_length": 278.4078674316406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 73.0,
      "epoch": 0.04053333333333333,
      "grad_norm": 0.2101420909166336,
      "learning_rate": 4.5e-06,
      "loss": 0.0435,
      "num_tokens": 8503940.0,
      "reward": 0.970810055732727,
      "reward_std": 0.363656222820282,
      "rewards/accuracy_reward_step": 0.33203125,
      "rewards/final_brier_reward_step": 0.6098023653030396,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 38
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 8.453383709694506e-08,
      "aux_brier/mean_group_std": 0.1489221191832934,
      "aux_brier/mean_r": 0.8456605949715051,
      "aux_brier/n_active_tok": 180.25,
      "aux_brier/n_groups": 10.25,
      "aux_brier/n_step_records": 45.0625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4156654166985536,
      "calib/avg_num_step_conf": 5.64453125,
      "calib/ece": 0.31293650793650796,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0992063492063492,
      "calib/gap": -0.0772311930818092,
      "calib/mean_conf": 0.518968253968254,
      "calib/mu_c": 0.46410958904109584,
      "calib/mu_w": 0.541340782122905,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.27111111111111114,
      "calib/std_conf": 0.2711712671634785,
      "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": 1931.0,
      "completions/max_terminated_length": 1931.0,
      "completions/mean_length": 318.3359375,
      "completions/mean_terminated_length": 318.3359375,
      "completions/min_length": 89.0,
      "completions/min_terminated_length": 89.0,
      "epoch": 0.0416,
      "grad_norm": 0.030873866751790047,
      "learning_rate": 4.472222222222223e-06,
      "loss": 0.0565,
      "num_tokens": 8691522.0,
      "reward": 0.9219582080841064,
      "reward_std": 0.33394742012023926,
      "rewards/accuracy_reward_step": 0.28515625,
      "rewards/final_brier_reward_step": 0.6175203323364258,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 39
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.1069860050838543e-07,
      "aux_brier/mean_group_std": 0.09696652906739323,
      "aux_brier/mean_r": 0.8914522087968476,
      "aux_brier/n_active_tok": 184.75,
      "aux_brier/n_groups": 9.90625,
      "aux_brier/n_step_records": 46.1875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5167284604519774,
      "calib/avg_num_step_conf": 5.7734375,
      "calib/ece": 0.28800796812749,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.06374501992031872,
      "calib/gap": 0.024824329096045095,
      "calib/mean_conf": 0.4860956175298805,
      "calib/mu_c": 0.5050847457627118,
      "calib/mu_w": 0.4802604166666667,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.26952191235059764,
      "calib/std_conf": 0.25475379650435753,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2183.0,
      "completions/max_terminated_length": 2183.0,
      "completions/mean_length": 331.79296875,
      "completions/mean_terminated_length": 331.79296875,
      "completions/min_length": 85.0,
      "completions/min_terminated_length": 85.0,
      "epoch": 0.042666666666666665,
      "grad_norm": 0.009598830714821815,
      "learning_rate": 4.444444444444444e-06,
      "loss": -0.0005,
      "num_tokens": 8883221.0,
      "reward": 0.8843405246734619,
      "reward_std": 0.35157209634780884,
      "rewards/accuracy_reward_step": 0.23046875,
      "rewards/final_brier_reward_step": 0.6779870986938477,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 40
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.0429956512347793e-07,
      "aux_brier/mean_group_std": 0.08730693682591503,
      "aux_brier/mean_r": 0.911575622717098,
      "aux_brier/n_active_tok": 180.375,
      "aux_brier/n_groups": 9.90625,
      "aux_brier/n_step_records": 45.09375,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5180523773201119,
      "calib/avg_num_step_conf": 5.6640625,
      "calib/ece": 0.21460317460317457,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.047619047619047616,
      "calib/gap": 0.0029824561403508087,
      "calib/mean_conf": 0.40301587301587305,
      "calib/mu_c": 0.4046491228070175,
      "calib/mu_w": 0.40166666666666667,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0826190476190476,
      "calib/std_conf": 0.24391082758159677,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 806.0,
      "completions/max_terminated_length": 806.0,
      "completions/mean_length": 299.2421875,
      "completions/mean_terminated_length": 300.41571044921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 66.0,
      "epoch": 0.04373333333333333,
      "grad_norm": 0.17322491109371185,
      "learning_rate": 4.416666666666667e-06,
      "loss": 0.0106,
      "num_tokens": 9067075.0,
      "reward": 1.1116578578948975,
      "reward_std": 0.40159034729003906,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.6810062527656555,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 41
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.4525956246180804e-07,
      "aux_brier/mean_group_std": 0.08007366569528057,
      "aux_brier/mean_r": 0.9243059512665607,
      "aux_brier/n_active_tok": 168.5,
      "aux_brier/n_groups": 8.90625,
      "aux_brier/n_step_records": 42.125,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.4832434646099243,
      "calib/avg_num_step_conf": 5.296875,
      "calib/ece": 0.16892857142857146,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.05555555555555555,
      "calib/gap": -0.018466316292403306,
      "calib/mean_conf": 0.3443253968253968,
      "calib/mu_c": 0.3325274725274725,
      "calib/mu_w": 0.3509937888198758,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.07607142857142857,
      "calib/std_conf": 0.23050139721248117,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 686.0,
      "completions/max_terminated_length": 686.0,
      "completions/mean_length": 271.64453125,
      "completions/mean_terminated_length": 272.7098083496094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 107.0,
      "epoch": 0.0448,
      "grad_norm": 0.08163885772228241,
      "learning_rate": 4.388888888888889e-06,
      "loss": 0.0416,
      "num_tokens": 9240984.0,
      "reward": 1.0217323303222656,
      "reward_std": 0.28946220874786377,
      "rewards/accuracy_reward_step": 0.359375,
      "rewards/final_brier_reward_step": 0.6963043212890625,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 42
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 9.231101651130391e-08,
      "aux_brier/mean_group_std": 0.0746223491502667,
      "aux_brier/mean_r": 0.9358663043217771,
      "aux_brier/n_active_tok": 182.625,
      "aux_brier/n_groups": 10.71875,
      "aux_brier/n_step_records": 45.65625,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.480614690539969,
      "calib/avg_num_step_conf": 5.7109375,
      "calib/ece": 0.19223529411764706,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0196078431372549,
      "calib/gap": 0.002710418722684349,
      "calib/mean_conf": 0.31803921568627447,
      "calib/mu_c": 0.3198780487804878,
      "calib/mu_w": 0.31716763005780346,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.09435294117647058,
      "calib/std_conf": 0.2150915865834661,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1937.0,
      "completions/max_terminated_length": 1937.0,
      "completions/mean_length": 312.6328125,
      "completions/mean_terminated_length": 312.6328125,
      "completions/min_length": 82.0,
      "completions/min_terminated_length": 82.0,
      "epoch": 0.04586666666666667,
      "grad_norm": 0.11977484822273254,
      "learning_rate": 4.361111111111112e-06,
      "loss": 0.023,
      "num_tokens": 9426242.0,
      "reward": 0.9972071051597595,
      "reward_std": 0.2933819591999054,
      "rewards/accuracy_reward_step": 0.3203125,
      "rewards/final_brier_reward_step": 0.7310156226158142,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 43
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.160131162802891e-07,
      "aux_brier/mean_group_std": 0.06065450534212685,
      "aux_brier/mean_r": 0.9425842242965002,
      "aux_brier/n_active_tok": 197.25,
      "aux_brier/n_groups": 11.53125,
      "aux_brier/n_step_records": 49.3125,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5035491815152833,
      "calib/avg_num_step_conf": 6.1640625,
      "calib/ece": 0.1675686274509804,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.011764705882352941,
      "calib/gap": -0.01358322468491957,
      "calib/mean_conf": 0.3016078431372549,
      "calib/mu_c": 0.2921794871794872,
      "calib/mu_w": 0.3057627118644068,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0816470588235294,
      "calib/std_conf": 0.19989056598513535,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1780.0,
      "completions/max_terminated_length": 1780.0,
      "completions/mean_length": 337.79296875,
      "completions/mean_terminated_length": 337.79296875,
      "completions/min_length": 74.0,
      "completions/min_terminated_length": 74.0,
      "epoch": 0.046933333333333334,
      "grad_norm": 0.032078322023153305,
      "learning_rate": 4.333333333333334e-06,
      "loss": 0.0346,
      "num_tokens": 9619037.0,
      "reward": 0.9874945878982544,
      "reward_std": 0.27549248933792114,
      "rewards/accuracy_reward_step": 0.3046875,
      "rewards/final_brier_reward_step": 0.7390409708023071,
      "rewards/format_reward_step_strict": 0.99609375,
      "step": 44
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.5251341978217425e-07,
      "aux_brier/mean_group_std": 0.05199722409692429,
      "aux_brier/mean_r": 0.9553006842650337,
      "aux_brier/n_active_tok": 182.625,
      "aux_brier/n_groups": 9.84375,
      "aux_brier/n_step_records": 45.65625,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.4215377067563779,
      "calib/avg_num_step_conf": 5.72265625,
      "calib/ece": 0.23852589641434263,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0199203187250996,
      "calib/gap": -0.04835085506027481,
      "calib/mean_conf": 0.24354581673306774,
      "calib/mu_c": 0.21195402298850569,
      "calib/mu_w": 0.2603048780487805,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.06772908366533865,
      "calib/std_conf": 0.17582751670290586,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 866.0,
      "completions/max_terminated_length": 866.0,
      "completions/mean_length": 314.13671875,
      "completions/mean_terminated_length": 315.36865234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 58.0,
      "epoch": 0.048,
      "grad_norm": 0.0748474970459938,
      "learning_rate": 4.305555555555556e-06,
      "loss": 0.0373,
      "num_tokens": 9804504.0,
      "reward": 1.0080393552780151,
      "reward_std": 0.3095443546772003,
      "rewards/accuracy_reward_step": 0.34375,
      "rewards/final_brier_reward_step": 0.6962199211120605,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 45
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 8.267610718259988e-08,
      "aux_brier/mean_group_std": 0.06327006275738548,
      "aux_brier/mean_r": 0.9465535762646131,
      "aux_brier/n_active_tok": 199.75,
      "aux_brier/n_groups": 11.46875,
      "aux_brier/n_step_records": 49.9375,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.43403643336529246,
      "calib/avg_num_step_conf": 6.26171875,
      "calib/ece": 0.2702629921259843,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.003937007874015748,
      "calib/gap": -0.04364034515819751,
      "calib/mean_conf": 0.22115433070866142,
      "calib/mu_c": 0.1955542857142857,
      "calib/mu_w": 0.23919463087248322,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.03901574803149606,
      "calib/std_conf": 0.15636737217101962,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1708.0,
      "completions/max_terminated_length": 1708.0,
      "completions/mean_length": 331.9765625,
      "completions/mean_terminated_length": 331.9765625,
      "completions/min_length": 109.0,
      "completions/min_terminated_length": 109.0,
      "epoch": 0.04906666666666667,
      "grad_norm": 0.17392946779727936,
      "learning_rate": 4.277777777777778e-06,
      "loss": 0.0383,
      "num_tokens": 9994258.0,
      "reward": 1.0756181478500366,
      "reward_std": 0.3038777709007263,
      "rewards/accuracy_reward_step": 0.4140625,
      "rewards/final_brier_reward_step": 0.6696599721908569,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 46
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.2866831400625287e-07,
      "aux_brier/mean_group_std": 0.05879260840197176,
      "aux_brier/mean_r": 0.9562548058580053,
      "aux_brier/n_active_tok": 213.875,
      "aux_brier/n_groups": 13.4375,
      "aux_brier/n_step_records": 53.46875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.4816504250466515,
      "calib/avg_num_step_conf": 6.703125,
      "calib/ece": 0.23188000000000003,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.008,
      "calib/gap": -0.016387449028958417,
      "calib/mean_conf": 0.18284,
      "calib/mu_c": 0.17241758241758243,
      "calib/mu_w": 0.18880503144654084,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.025360000000000004,
      "calib/std_conf": 0.1448486603320859,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2975.0,
      "completions/max_terminated_length": 2975.0,
      "completions/mean_length": 374.66015625,
      "completions/mean_terminated_length": 376.1294250488281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 99.0,
      "epoch": 0.050133333333333335,
      "grad_norm": 0.08345242589712143,
      "learning_rate": 4.25e-06,
      "loss": 0.0466,
      "num_tokens": 10196147.0,
      "reward": 1.0144307613372803,
      "reward_std": 0.32323795557022095,
      "rewards/accuracy_reward_step": 0.35546875,
      "rewards/final_brier_reward_step": 0.6905355453491211,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 47
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.769369222086151e-07,
      "aux_brier/mean_group_std": 0.04574837180941228,
      "aux_brier/mean_r": 0.9644007629165204,
      "aux_brier/n_active_tok": 179.5,
      "aux_brier/n_groups": 9.03125,
      "aux_brier/n_step_records": 44.875,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.4432930339666091,
      "calib/avg_num_step_conf": 5.6875,
      "calib/ece": 0.16326953125,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0179101077391233,
      "calib/mean_conf": 0.14508984375000003,
      "calib/mu_c": 0.13158730158730159,
      "calib/mu_w": 0.14949740932642488,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.031132812499999985,
      "calib/std_conf": 0.10950665172422444,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1037.0,
      "completions/max_terminated_length": 1037.0,
      "completions/mean_length": 293.421875,
      "completions/mean_terminated_length": 294.57257080078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 97.0,
      "epoch": 0.0512,
      "grad_norm": 0.11976519972085953,
      "learning_rate": 4.222222222222223e-06,
      "loss": 0.0628,
      "num_tokens": 10374951.0,
      "reward": 0.9385948181152344,
      "reward_std": 0.18301494419574738,
      "rewards/accuracy_reward_step": 0.24609375,
      "rewards/final_brier_reward_step": 0.7856290936470032,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 48
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.169830846225466e-08,
      "aux_brier/mean_group_std": 0.04859701273715004,
      "aux_brier/mean_r": 0.9599143498316884,
      "aux_brier/n_active_tok": 194.75,
      "aux_brier/n_groups": 11.84375,
      "aux_brier/n_step_records": 48.6875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.517244884130641,
      "calib/avg_num_step_conf": 6.12109375,
      "calib/ece": 0.2568924302788845,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": -0.009650359127253039,
      "calib/mean_conf": 0.14071713147410359,
      "calib/mu_c": 0.1346808510638298,
      "calib/mu_w": 0.14433121019108283,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.011553784860557768,
      "calib/std_conf": 0.11051258543343341,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2021.0,
      "completions/max_terminated_length": 2021.0,
      "completions/mean_length": 342.5234375,
      "completions/mean_terminated_length": 342.5234375,
      "completions/min_length": 102.0,
      "completions/min_terminated_length": 102.0,
      "epoch": 0.05226666666666667,
      "grad_norm": 0.2972572147846222,
      "learning_rate": 4.194444444444445e-06,
      "loss": 0.1037,
      "num_tokens": 10567173.0,
      "reward": 1.0237152576446533,
      "reward_std": 0.2431778907775879,
      "rewards/accuracy_reward_step": 0.3671875,
      "rewards/final_brier_reward_step": 0.6807984113693237,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 49
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.9017204059299537e-07,
      "aux_brier/mean_group_std": 0.047627489063467456,
      "aux_brier/mean_r": 0.9671897326182052,
      "aux_brier/n_active_tok": 197.5,
      "aux_brier/n_groups": 11.25,
      "aux_brier/n_step_records": 49.375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5290701364646144,
      "calib/avg_num_step_conf": 6.21484375,
      "calib/ece": 0.360547619047619,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002993970168200605,
      "calib/mean_conf": 0.1061984126984127,
      "calib/mu_c": 0.10782608695652178,
      "calib/mu_w": 0.10483211678832118,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0051984126984126995,
      "calib/std_conf": 0.07634929338674379,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2177.0,
      "completions/max_terminated_length": 2177.0,
      "completions/mean_length": 356.31640625,
      "completions/mean_terminated_length": 357.7137451171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 63.0,
      "epoch": 0.05333333333333334,
      "grad_norm": 0.02677062712609768,
      "learning_rate": 4.166666666666667e-06,
      "loss": 0.0327,
      "num_tokens": 10763750.0,
      "reward": 1.0912978649139404,
      "reward_std": 0.3045734167098999,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.6151912212371826,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 50
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.5887713887274835e-06,
      "aux_brier/mean_group_std": 0.04609024460167228,
      "aux_brier/mean_r": 0.9644180483138113,
      "aux_brier/n_active_tok": 207.625,
      "aux_brier/n_groups": 13.09375,
      "aux_brier/n_step_records": 51.90625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4327722772277227,
      "calib/avg_num_step_conf": 6.66015625,
      "calib/ece": 0.31158964143426293,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.014207326732673276,
      "calib/mean_conf": 0.10522310756972113,
      "calib/mu_c": 0.09673267326732674,
      "calib/mu_w": 0.11094000000000001,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.007211155378486056,
      "calib/std_conf": 0.08254356584769627,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2336.0,
      "completions/max_terminated_length": 2336.0,
      "completions/mean_length": 333.21875,
      "completions/mean_terminated_length": 337.16998291015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 87.0,
      "epoch": 0.0544,
      "grad_norm": 0.05267907679080963,
      "learning_rate": 4.138888888888889e-06,
      "loss": 0.0192,
      "num_tokens": 10958350.0,
      "reward": 1.0391311645507812,
      "reward_std": 0.29196834564208984,
      "rewards/accuracy_reward_step": 0.39453125,
      "rewards/final_brier_reward_step": 0.6408998966217041,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 51
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.583419999151019e-07,
      "aux_brier/mean_group_std": 0.06857622843521455,
      "aux_brier/mean_r": 0.9559836778689764,
      "aux_brier/n_active_tok": 169.75,
      "aux_brier/n_groups": 8.65625,
      "aux_brier/n_step_records": 42.4375,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5377759960527939,
      "calib/avg_num_step_conf": 5.31640625,
      "calib/ece": 0.44505882352941173,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.007583569754533101,
      "calib/mean_conf": 0.08043137254901961,
      "calib/mu_c": 0.08402985074626865,
      "calib/mu_w": 0.07644628099173555,
      "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": 0.052433059060385125,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 888.0,
      "completions/max_terminated_length": 888.0,
      "completions/mean_length": 304.37109375,
      "completions/mean_terminated_length": 305.5647277832031,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 86.0,
      "epoch": 0.055466666666666664,
      "grad_norm": 0.03042164258658886,
      "learning_rate": 4.111111111111111e-06,
      "loss": -0.007,
      "num_tokens": 11144221.0,
      "reward": 1.163251280784607,
      "reward_std": 0.3101959526538849,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.551442563533783,
      "rewards/format_reward_step_strict": 0.99609375,
      "step": 52
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.88559796332477e-07,
      "aux_brier/mean_group_std": 0.08276254140809931,
      "aux_brier/mean_r": 0.948797932380574,
      "aux_brier/n_active_tok": 199.5,
      "aux_brier/n_groups": 11.84375,
      "aux_brier/n_step_records": 49.875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5718158890290037,
      "calib/avg_num_step_conf": 6.24609375,
      "calib/ece": 0.4018968253968254,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.010276166456494332,
      "calib/mean_conf": 0.08643650793650795,
      "calib/mu_c": 0.09173770491803279,
      "calib/mu_w": 0.08146153846153846,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0021031746031746033,
      "calib/std_conf": 0.06867348510203528,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2008.0,
      "completions/max_terminated_length": 2008.0,
      "completions/mean_length": 359.95703125,
      "completions/mean_terminated_length": 359.95703125,
      "completions/min_length": 117.0,
      "completions/min_terminated_length": 117.0,
      "epoch": 0.05653333333333333,
      "grad_norm": 0.011011505499482155,
      "learning_rate": 4.083333333333334e-06,
      "loss": 0.0329,
      "num_tokens": 11342194.0,
      "reward": 1.116516351699829,
      "reward_std": 0.30897217988967896,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.5832530856132507,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 53
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.2435113564135403e-06,
      "aux_brier/mean_group_std": 0.07204236926393826,
      "aux_brier/mean_r": 0.9523674597628599,
      "aux_brier/n_active_tok": 171.75,
      "aux_brier/n_groups": 10.78125,
      "aux_brier/n_step_records": 42.9375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.501830114855484,
      "calib/avg_num_step_conf": 5.3671875,
      "calib/ece": 0.4752569169960474,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0042010602044680045,
      "calib/mean_conf": 0.07415019762845851,
      "calib/mu_c": 0.07604316546762589,
      "calib/mu_w": 0.07184210526315789,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05885408675384341,
      "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": 2405.0,
      "completions/max_terminated_length": 2405.0,
      "completions/mean_length": 307.6484375,
      "completions/mean_terminated_length": 307.6484375,
      "completions/min_length": 86.0,
      "completions/min_terminated_length": 86.0,
      "epoch": 0.0576,
      "grad_norm": 0.037044648081064224,
      "learning_rate": 4.055555555555556e-06,
      "loss": 0.0125,
      "num_tokens": 11527184.0,
      "reward": 1.1587554216384888,
      "reward_std": 0.2927601933479309,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.5100218653678894,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 54
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 7.267408025923849e-07,
      "aux_brier/mean_group_std": 0.06855691800515884,
      "aux_brier/mean_r": 0.9563366510048938,
      "aux_brier/n_active_tok": 178.75,
      "aux_brier/n_groups": 11.09375,
      "aux_brier/n_step_records": 44.6875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.46380687491085437,
      "calib/avg_num_step_conf": 5.60546875,
      "calib/ece": 0.2638616600790514,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.006634288974468709,
      "calib/mean_conf": 0.06570355731225296,
      "calib/mu_c": 0.06121951219512194,
      "calib/mu_w": 0.06785380116959065,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.002727272727272727,
      "calib/std_conf": 0.052667154069498685,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2005.0,
      "completions/max_terminated_length": 2005.0,
      "completions/mean_length": 330.3359375,
      "completions/mean_terminated_length": 331.6313781738281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.058666666666666666,
      "grad_norm": 0.009194295853376389,
      "learning_rate": 4.027777777777779e-06,
      "loss": 0.0647,
      "num_tokens": 11719574.0,
      "reward": 0.9875449538230896,
      "reward_std": 0.23884600400924683,
      "rewards/accuracy_reward_step": 0.3203125,
      "rewards/final_brier_reward_step": 0.7001798152923584,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 55
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.0774575854854973e-06,
      "aux_brier/mean_group_std": 0.06709565865688792,
      "aux_brier/mean_r": 0.9565804266181223,
      "aux_brier/n_active_tok": 193.125,
      "aux_brier/n_groups": 11.15625,
      "aux_brier/n_step_records": 48.28125,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5015060240963856,
      "calib/avg_num_step_conf": 6.0390625,
      "calib/ece": 0.28295275590551183,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00040115005476451526,
      "calib/mean_conf": 0.06507874015748032,
      "calib/mu_c": 0.06534090909090909,
      "calib/mu_w": 0.06493975903614457,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0007874015748031496,
      "calib/std_conf": 0.0431691611819474,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 353.73046875,
      "completions/mean_terminated_length": 353.73046875,
      "completions/min_length": 84.0,
      "completions/min_terminated_length": 84.0,
      "epoch": 0.05973333333333333,
      "grad_norm": 0.0191681906580925,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.079,
      "num_tokens": 11916969.0,
      "reward": 1.0097177028656006,
      "reward_std": 0.24075153470039368,
      "rewards/accuracy_reward_step": 0.34375,
      "rewards/final_brier_reward_step": 0.6873081922531128,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 56
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.2804876683425448e-06,
      "aux_brier/mean_group_std": 0.04012048894455258,
      "aux_brier/mean_r": 0.9748410301637029,
      "aux_brier/n_active_tok": 187.375,
      "aux_brier/n_groups": 10.96875,
      "aux_brier/n_step_records": 46.84375,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.39824218750000007,
      "calib/avg_num_step_conf": 5.8671875,
      "calib/ece": 0.4264919354838709,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.018453125000000008,
      "calib/mean_conf": 0.0630241935483871,
      "calib/mu_c": 0.0535,
      "calib/mu_w": 0.071953125,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00282258064516129,
      "calib/std_conf": 0.05080925171495352,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2632.0,
      "completions/max_terminated_length": 2632.0,
      "completions/mean_length": 373.07421875,
      "completions/mean_terminated_length": 373.07421875,
      "completions/min_length": 100.0,
      "completions/min_terminated_length": 100.0,
      "epoch": 0.0608,
      "grad_norm": 0.010963428765535355,
      "learning_rate": 3.972222222222223e-06,
      "loss": 0.0625,
      "num_tokens": 12119268.0,
      "reward": 1.0871237516403198,
      "reward_std": 0.2706947922706604,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5438074469566345,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 57
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.280510507521541e-06,
      "aux_brier/mean_group_std": 0.04747223413674383,
      "aux_brier/mean_r": 0.9704677822945551,
      "aux_brier/n_active_tok": 210.375,
      "aux_brier/n_groups": 14.4375,
      "aux_brier/n_step_records": 52.59375,
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.43560659433563476,
      "calib/avg_num_step_conf": 6.75390625,
      "calib/ece": 0.3235191836734694,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.009850796111032824,
      "calib/mean_conf": 0.060154285714285716,
      "calib/mu_c": 0.054082978723404264,
      "calib/mu_w": 0.06393377483443709,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.05175294955608909,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2440.0,
      "completions/max_terminated_length": 2440.0,
      "completions/mean_length": 427.703125,
      "completions/mean_terminated_length": 429.38043212890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 109.0,
      "epoch": 0.06186666666666667,
      "grad_norm": 0.05008196085691452,
      "learning_rate": 3.944444444444445e-06,
      "loss": 0.1937,
      "num_tokens": 12335080.0,
      "reward": 0.9957273602485657,
      "reward_std": 0.343005508184433,
      "rewards/accuracy_reward_step": 0.3671875,
      "rewards/final_brier_reward_step": 0.623534619808197,
      "rewards/format_reward_step_strict": 0.9453125,
      "step": 58
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.077476075727617e-07,
      "aux_brier/mean_group_std": 0.043504242172368354,
      "aux_brier/mean_r": 0.9694438174815395,
      "aux_brier/n_active_tok": 199.25,
      "aux_brier/n_groups": 10.84375,
      "aux_brier/n_step_records": 49.8125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.45559304768574527,
      "calib/avg_num_step_conf": 6.28515625,
      "calib/ece": 0.39793650793650787,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.007465461259311128,
      "calib/mean_conf": 0.05420634920634921,
      "calib/mu_c": 0.05008849557522125,
      "calib/mu_w": 0.05755395683453238,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.001865079365079365,
      "calib/std_conf": 0.040046662754626844,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1402.0,
      "completions/max_terminated_length": 1402.0,
      "completions/mean_length": 351.8046875,
      "completions/mean_terminated_length": 353.184326171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 107.0,
      "epoch": 0.06293333333333333,
      "grad_norm": 0.06653322279453278,
      "learning_rate": 3.916666666666667e-06,
      "loss": 0.0079,
      "num_tokens": 12531390.0,
      "reward": 1.0792728662490845,
      "reward_std": 0.21965667605400085,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.5827163457870483,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 59
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.519963707679599e-07,
      "aux_brier/mean_group_std": 0.05771456903880541,
      "aux_brier/mean_r": 0.9655047520637373,
      "aux_brier/n_active_tok": 208.875,
      "aux_brier/n_groups": 14.25,
      "aux_brier/n_step_records": 52.21875,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.4223856209150327,
      "calib/avg_num_step_conf": 6.703125,
      "calib/ece": 0.36349593495934956,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.010249183006535947,
      "calib/mean_conf": 0.05276422764227642,
      "calib/mu_c": 0.046764705882352944,
      "calib/mu_w": 0.05701388888888889,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0008130081300813009,
      "calib/std_conf": 0.03601717776731212,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 401.3125,
      "completions/mean_terminated_length": 402.88629150390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.064,
      "grad_norm": 0.0890054702758789,
      "learning_rate": 3.88888888888889e-06,
      "loss": 0.1602,
      "num_tokens": 12742982.0,
      "reward": 1.031773328781128,
      "reward_std": 0.31017613410949707,
      "rewards/accuracy_reward_step": 0.40234375,
      "rewards/final_brier_reward_step": 0.5958437919616699,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 60
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.4815887647845472e-06,
      "aux_brier/mean_group_std": 0.049969137749620404,
      "aux_brier/mean_r": 0.9610750055704818,
      "aux_brier/n_active_tok": 177.875,
      "aux_brier/n_groups": 11.65625,
      "aux_brier/n_step_records": 44.46875,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4649503722084367,
      "calib/avg_num_step_conf": 5.55859375,
      "calib/ece": 0.46775590551181107,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00020223325062034003,
      "calib/mean_conf": 0.04405511811023623,
      "calib/mu_c": 0.044153846153846155,
      "calib/mu_w": 0.043951612903225815,
      "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": 0.038776729584501,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2494.0,
      "completions/max_terminated_length": 2494.0,
      "completions/mean_length": 313.25,
      "completions/mean_terminated_length": 313.25,
      "completions/min_length": 101.0,
      "completions/min_terminated_length": 101.0,
      "epoch": 0.06506666666666666,
      "grad_norm": 0.019003357738256454,
      "learning_rate": 3.861111111111112e-06,
      "loss": 0.0693,
      "num_tokens": 12927238.0,
      "reward": 1.1353565454483032,
      "reward_std": 0.2609567642211914,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5258011817932129,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 61
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.6686910742368397e-06,
      "aux_brier/mean_group_std": 0.04529141929014861,
      "aux_brier/mean_r": 0.9706526473128498,
      "aux_brier/n_active_tok": 221.375,
      "aux_brier/n_groups": 15.5,
      "aux_brier/n_step_records": 55.34375,
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.4250916710319539,
      "calib/avg_num_step_conf": 6.91796875,
      "calib/ece": 0.29323770491803275,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.006407992217316472,
      "calib/mean_conf": 0.0485655737704918,
      "calib/mu_c": 0.04433734939759036,
      "calib/mu_w": 0.05074534161490683,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0008196721311475411,
      "calib/std_conf": 0.04543045264962123,
      "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": 2216.0,
      "completions/max_terminated_length": 2216.0,
      "completions/mean_length": 431.98828125,
      "completions/mean_terminated_length": 433.682373046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 52.0,
      "epoch": 0.06613333333333334,
      "grad_norm": 0.028923727571964264,
      "learning_rate": 3.833333333333334e-06,
      "loss": 0.0875,
      "num_tokens": 13144907.0,
      "reward": 0.9612126350402832,
      "reward_std": 0.28925445675849915,
      "rewards/accuracy_reward_step": 0.32421875,
      "rewards/final_brier_reward_step": 0.6495382785797119,
      "rewards/format_reward_step_strict": 0.94921875,
      "step": 62
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.602942894007313e-06,
      "aux_brier/mean_group_std": 0.05251221459921494,
      "aux_brier/mean_r": 0.9662158757182411,
      "aux_brier/n_active_tok": 200.25,
      "aux_brier/n_groups": 11.15625,
      "aux_brier/n_step_records": 50.0625,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.519625,
      "calib/avg_num_step_conf": 6.2734375,
      "calib/ece": 0.44718577075098814,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.001172750000000014,
      "calib/mean_conf": 0.049889328063241103,
      "calib/mu_c": 0.04929599999999999,
      "calib/mu_w": 0.05046875000000001,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00150197628458498,
      "calib/std_conf": 0.03841220682472356,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2723.0,
      "completions/max_terminated_length": 2723.0,
      "completions/mean_length": 371.84375,
      "completions/mean_terminated_length": 371.84375,
      "completions/min_length": 60.0,
      "completions/min_terminated_length": 60.0,
      "epoch": 0.0672,
      "grad_norm": 0.08982036262750626,
      "learning_rate": 3.8055555555555556e-06,
      "loss": 0.0424,
      "num_tokens": 13348739.0,
      "reward": 1.116524338722229,
      "reward_std": 0.27404630184173584,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.5442226529121399,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 63
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.3130418960172197e-06,
      "aux_brier/mean_group_std": 0.042722527588865376,
      "aux_brier/mean_r": 0.9680877043615644,
      "aux_brier/n_active_tok": 178.375,
      "aux_brier/n_groups": 9.40625,
      "aux_brier/n_step_records": 44.59375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.49615580124539327,
      "calib/avg_num_step_conf": 5.578125,
      "calib/ece": 0.44328685258964134,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.001266997077138142,
      "calib/mean_conf": 0.04515936254980079,
      "calib/mu_c": 0.044508196721311476,
      "calib/mu_w": 0.04577519379844962,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.001195219123505976,
      "calib/std_conf": 0.03392302192209186,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2537.0,
      "completions/max_terminated_length": 2537.0,
      "completions/mean_length": 367.1953125,
      "completions/mean_terminated_length": 368.63531494140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 122.0,
      "epoch": 0.06826666666666667,
      "grad_norm": 0.10238279402256012,
      "learning_rate": 3.777777777777778e-06,
      "loss": 0.0644,
      "num_tokens": 13546517.0,
      "reward": 1.0977150201797485,
      "reward_std": 0.2985689342021942,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.5392975807189941,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 64
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.6477803753977014e-06,
      "aux_brier/mean_group_std": 0.04302786409165657,
      "aux_brier/mean_r": 0.976210879805285,
      "aux_brier/n_active_tok": 167.0,
      "aux_brier/n_groups": 9.59375,
      "aux_brier/n_step_records": 41.75,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.48399531269273466,
      "calib/avg_num_step_conf": 5.21875,
      "calib/ece": 0.4312549019607843,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0019476995189342539,
      "calib/mean_conf": 0.04325490196078432,
      "calib/mu_c": 0.042231404958677686,
      "calib/mu_w": 0.04417910447761194,
      "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": 0.02801949121121735,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2826.0,
      "completions/max_terminated_length": 2826.0,
      "completions/mean_length": 313.296875,
      "completions/mean_terminated_length": 313.296875,
      "completions/min_length": 100.0,
      "completions/min_terminated_length": 100.0,
      "epoch": 0.06933333333333333,
      "grad_norm": 0.014597240835428238,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.031,
      "num_tokens": 13731745.0,
      "reward": 1.1108815670013428,
      "reward_std": 0.19665348529815674,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.5607136487960815,
      "rewards/format_reward_step_strict": 0.99609375,
      "step": 65
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 3.5007501467121394e-08,
      "aux_brier/mean_group_std": 0.0469428577393126,
      "aux_brier/mean_r": 0.972645798072567,
      "aux_brier/n_active_tok": 201.375,
      "aux_brier/n_groups": 12.40625,
      "aux_brier/n_step_records": 50.34375,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.45569280343716434,
      "calib/avg_num_step_conf": 6.38671875,
      "calib/ece": 0.34440000000000004,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004108485499462945,
      "calib/mean_conf": 0.0476,
      "calib/mu_c": 0.04510204081632653,
      "calib/mu_w": 0.049210526315789475,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.03163289427162807,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2717.0,
      "completions/max_terminated_length": 2717.0,
      "completions/mean_length": 407.3828125,
      "completions/mean_terminated_length": 408.98040771484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 83.0,
      "epoch": 0.0704,
      "grad_norm": 0.014792166650295258,
      "learning_rate": 3.7222222222222225e-06,
      "loss": 0.0846,
      "num_tokens": 13942387.0,
      "reward": 1.0222983360290527,
      "reward_std": 0.22025427222251892,
      "rewards/accuracy_reward_step": 0.3828125,
      "rewards/final_brier_reward_step": 0.6204434037208557,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 66
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.0238753276901065e-06,
      "aux_brier/mean_group_std": 0.0689542588139005,
      "aux_brier/mean_r": 0.9557478736472529,
      "aux_brier/n_active_tok": 187.5,
      "aux_brier/n_groups": 12.5,
      "aux_brier/n_step_records": 46.875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5230611328500866,
      "calib/avg_num_step_conf": 5.859375,
      "calib/ece": 0.43263999999999997,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0013175957405863054,
      "calib/mean_conf": 0.04536,
      "calib/mu_c": 0.04605042016806722,
      "calib/mu_w": 0.044732824427480916,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.001,
      "calib/std_conf": 0.0296659805164097,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2470.0,
      "completions/max_terminated_length": 2470.0,
      "completions/mean_length": 388.69921875,
      "completions/mean_terminated_length": 388.69921875,
      "completions/min_length": 100.0,
      "completions/min_terminated_length": 100.0,
      "epoch": 0.07146666666666666,
      "grad_norm": 0.0096532441675663,
      "learning_rate": 3.694444444444445e-06,
      "loss": 0.0827,
      "num_tokens": 14146902.0,
      "reward": 1.0910406112670898,
      "reward_std": 0.2472362220287323,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5516625046730042,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 67
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.434876368837216e-06,
      "aux_brier/mean_group_std": 0.03418800083448402,
      "aux_brier/mean_r": 0.9780984264300685,
      "aux_brier/n_active_tok": 177.75,
      "aux_brier/n_groups": 10.90625,
      "aux_brier/n_step_records": 44.4375,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.44980609066056326,
      "calib/avg_num_step_conf": 5.5546875,
      "calib/ece": 0.3773228346456693,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.001331298874690058,
      "calib/mean_conf": 0.04637795275590552,
      "calib/mu_c": 0.045607476635514024,
      "calib/mu_w": 0.04693877551020408,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0012204724409448822,
      "calib/std_conf": 0.037472704845089806,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2140.0,
      "completions/max_terminated_length": 2140.0,
      "completions/mean_length": 343.1953125,
      "completions/mean_terminated_length": 343.1953125,
      "completions/min_length": 104.0,
      "completions/min_terminated_length": 104.0,
      "epoch": 0.07253333333333334,
      "grad_norm": 0.051165852695703506,
      "learning_rate": 3.6666666666666666e-06,
      "loss": 0.0449,
      "num_tokens": 14338848.0,
      "reward": 1.0701727867126465,
      "reward_std": 0.2218499779701233,
      "rewards/accuracy_reward_step": 0.421875,
      "rewards/final_brier_reward_step": 0.608816385269165,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 68
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.0088494057657336e-06,
      "aux_brier/mean_group_std": 0.04627228822255264,
      "aux_brier/mean_r": 0.9687250312122738,
      "aux_brier/n_active_tok": 166.625,
      "aux_brier/n_groups": 10.3125,
      "aux_brier/n_step_records": 41.65625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.48369788106630207,
      "calib/avg_num_step_conf": 5.21875,
      "calib/ece": 0.33550200803212854,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0032727272727272813,
      "calib/mean_conf": 0.046024096385542175,
      "calib/mu_c": 0.04399999999999999,
      "calib/mu_w": 0.04727272727272727,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.029721860264746814,
      "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": 2178.0,
      "completions/max_terminated_length": 2178.0,
      "completions/mean_length": 389.7890625,
      "completions/mean_terminated_length": 392.8582763671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 118.0,
      "epoch": 0.0736,
      "grad_norm": 0.018729669973254204,
      "learning_rate": 3.638888888888889e-06,
      "loss": 0.044,
      "num_tokens": 14543130.0,
      "reward": 1.0191528797149658,
      "reward_std": 0.24353712797164917,
      "rewards/accuracy_reward_step": 0.375,
      "rewards/final_brier_reward_step": 0.6312991976737976,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 69
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 3.4977007354608514e-06,
      "aux_brier/mean_group_std": 0.04300463422366555,
      "aux_brier/mean_r": 0.9763941743900566,
      "aux_brier/n_active_tok": 177.625,
      "aux_brier/n_groups": 11.40625,
      "aux_brier/n_step_records": 44.40625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.47789115646258506,
      "calib/avg_num_step_conf": 5.55078125,
      "calib/ece": 0.37183266932270914,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.002151360544217669,
      "calib/mean_conf": 0.04250996015936255,
      "calib/mu_c": 0.04125000000000001,
      "calib/mu_w": 0.04340136054421768,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.022663195245232134,
      "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": 2213.0,
      "completions/max_terminated_length": 2213.0,
      "completions/mean_length": 393.6953125,
      "completions/mean_terminated_length": 393.6953125,
      "completions/min_length": 96.0,
      "completions/min_terminated_length": 96.0,
      "epoch": 0.07466666666666667,
      "grad_norm": 0.023448629304766655,
      "learning_rate": 3.6111111111111115e-06,
      "loss": 0.0782,
      "num_tokens": 14750908.0,
      "reward": 1.047849178314209,
      "reward_std": 0.22698289155960083,
      "rewards/accuracy_reward_step": 0.40625,
      "rewards/final_brier_reward_step": 0.6054589748382568,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 70
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.71178257294347e-06,
      "aux_brier/mean_group_std": 0.047826320667128996,
      "aux_brier/mean_r": 0.9638519321791721,
      "aux_brier/n_active_tok": 174.75,
      "aux_brier/n_groups": 10.65625,
      "aux_brier/n_step_records": 43.6875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.465642382309049,
      "calib/avg_num_step_conf": 5.4609375,
      "calib/ece": 0.3138339920948616,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.005360195360195365,
      "calib/mean_conf": 0.04584980237154151,
      "calib/mu_c": 0.042417582417582415,
      "calib/mu_w": 0.04777777777777778,
      "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": 0.027658439912256715,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2663.0,
      "completions/max_terminated_length": 2663.0,
      "completions/mean_length": 379.71484375,
      "completions/mean_terminated_length": 379.71484375,
      "completions/min_length": 114.0,
      "completions/min_terminated_length": 114.0,
      "epoch": 0.07573333333333333,
      "grad_norm": 0.013549897819757462,
      "learning_rate": 3.5833333333333335e-06,
      "loss": 0.0927,
      "num_tokens": 14952523.0,
      "reward": 1.0146431922912598,
      "reward_std": 0.24198010563850403,
      "rewards/accuracy_reward_step": 0.35546875,
      "rewards/final_brier_reward_step": 0.6601351499557495,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 71
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.6784378907058795e-06,
      "aux_brier/mean_group_std": 0.045835261917549525,
      "aux_brier/mean_r": 0.9716608193362408,
      "aux_brier/n_active_tok": 162.875,
      "aux_brier/n_groups": 9.25,
      "aux_brier/n_step_records": 40.71875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4948198198198198,
      "calib/avg_num_step_conf": 5.08984375,
      "calib/ece": 0.4011155378486056,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0025225225225225176,
      "calib/mean_conf": 0.04111553784860558,
      "calib/mu_c": 0.04252252252252252,
      "calib/mu_w": 0.04,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.029476687351544087,
      "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": 2815.0,
      "completions/max_terminated_length": 2815.0,
      "completions/mean_length": 361.96484375,
      "completions/mean_terminated_length": 363.38433837890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.0768,
      "grad_norm": 0.010157402604818344,
      "learning_rate": 3.555555555555556e-06,
      "loss": 0.0369,
      "num_tokens": 15149594.0,
      "reward": 1.0662086009979248,
      "reward_std": 0.27851247787475586,
      "rewards/accuracy_reward_step": 0.43359375,
      "rewards/final_brier_reward_step": 0.5773344039916992,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 72
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -8.495429312582736e-06,
      "aux_brier/mean_group_std": 0.047189342901178016,
      "aux_brier/mean_r": 0.9716737835120474,
      "aux_brier/n_active_tok": 155.625,
      "aux_brier/n_groups": 8.40625,
      "aux_brier/n_step_records": 38.90625,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4852047146401985,
      "calib/avg_num_step_conf": 4.86328125,
      "calib/ece": 0.4717716535433071,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0005533498759305172,
      "calib/mean_conf": 0.04003937007874016,
      "calib/mu_c": 0.03976923076923077,
      "calib/mu_w": 0.04032258064516129,
      "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": 0.017512605842156293,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3024.0,
      "completions/max_terminated_length": 3024.0,
      "completions/mean_length": 341.09375,
      "completions/mean_terminated_length": 341.09375,
      "completions/min_length": 104.0,
      "completions/min_terminated_length": 104.0,
      "epoch": 0.07786666666666667,
      "grad_norm": 0.022411437705159187,
      "learning_rate": 3.5277777777777784e-06,
      "loss": 0.0334,
      "num_tokens": 15343946.0,
      "reward": 1.131695032119751,
      "reward_std": 0.2937101721763611,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5189679861068726,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 73
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.871608686509489e-07,
      "aux_brier/mean_group_std": 0.033736425267474904,
      "aux_brier/mean_r": 0.9816442073615657,
      "aux_brier/n_active_tok": 151.625,
      "aux_brier/n_groups": 8.53125,
      "aux_brier/n_step_records": 37.90625,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.47032279314888015,
      "calib/avg_num_step_conf": 4.73828125,
      "calib/ece": 0.4068548387096775,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.009715415019762863,
      "calib/mean_conf": 0.043951612903225815,
      "calib/mu_c": 0.03854545454545453,
      "calib/mu_w": 0.04826086956521739,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0036290322580645163,
      "calib/std_conf": 0.057932365297918294,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2353.0,
      "completions/max_terminated_length": 2353.0,
      "completions/mean_length": 376.48828125,
      "completions/mean_terminated_length": 377.9647216796875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 112.0,
      "epoch": 0.07893333333333333,
      "grad_norm": 0.04105854406952858,
      "learning_rate": 3.5e-06,
      "loss": 0.1344,
      "num_tokens": 15544255.0,
      "reward": 1.05582857131958,
      "reward_std": 0.25634765625,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.5670648217201233,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 74
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -9.998759458708373e-06,
      "aux_brier/mean_group_std": 0.08682547556133395,
      "aux_brier/mean_r": 0.9394467702215956,
      "aux_brier/n_active_tok": 150.125,
      "aux_brier/n_groups": 7.875,
      "aux_brier/n_step_records": 37.53125,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5641666666666667,
      "calib/avg_num_step_conf": 4.69921875,
      "calib/ece": 0.570234375,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003184615384615376,
      "calib/mean_conf": 0.03914062500000001,
      "calib/mu_c": 0.04038461538461538,
      "calib/mu_w": 0.037200000000000004,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.0184360102682054,
      "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": 876.0,
      "completions/max_terminated_length": 876.0,
      "completions/mean_length": 314.4765625,
      "completions/mean_terminated_length": 315.7098083496094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 102.0,
      "epoch": 0.08,
      "grad_norm": 0.018008295446634293,
      "learning_rate": 3.4722222222222224e-06,
      "loss": 0.0035,
      "num_tokens": 15729513.0,
      "reward": 1.2168571949005127,
      "reward_std": 0.2569185495376587,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.4377410113811493,
      "rewards/format_reward_step_strict": 0.99609375,
      "step": 75
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.66402079527073e-06,
      "aux_brier/mean_group_std": 0.06815238686891403,
      "aux_brier/mean_r": 0.9553527514328811,
      "aux_brier/n_active_tok": 158.875,
      "aux_brier/n_groups": 9.40625,
      "aux_brier/n_step_records": 39.71875,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.44999353085780824,
      "calib/avg_num_step_conf": 5.046875,
      "calib/ece": 0.4321686746987952,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0024770345452192963,
      "calib/mean_conf": 0.04172690763052209,
      "calib/mu_c": 0.04042372881355932,
      "calib/mu_w": 0.04290076335877862,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.018558215115376163,
      "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": 3043.0,
      "completions/max_terminated_length": 3043.0,
      "completions/mean_length": 367.40234375,
      "completions/mean_terminated_length": 370.2952880859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 97.0,
      "epoch": 0.08106666666666666,
      "grad_norm": 0.026839526370167732,
      "learning_rate": 3.444444444444445e-06,
      "loss": 0.1172,
      "num_tokens": 15926624.0,
      "reward": 1.081865906715393,
      "reward_std": 0.26003503799438477,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.546213686466217,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 76
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 8.547130481928455e-06,
      "aux_brier/mean_group_std": 0.06601589482379376,
      "aux_brier/mean_r": 0.9545857882905452,
      "aux_brier/n_active_tok": 157.875,
      "aux_brier/n_groups": 8.9375,
      "aux_brier/n_step_records": 39.46875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4649763779527559,
      "calib/avg_num_step_conf": 4.93359375,
      "calib/ece": 0.45936507936507937,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0006878740157480259,
      "calib/mean_conf": 0.03666666666666667,
      "calib/mu_c": 0.036320000000000005,
      "calib/mu_w": 0.03700787401574803,
      "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": 0.016207973738207152,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2683.0,
      "completions/max_terminated_length": 2683.0,
      "completions/mean_length": 352.59375,
      "completions/mean_terminated_length": 352.59375,
      "completions/min_length": 94.0,
      "completions/min_terminated_length": 94.0,
      "epoch": 0.08213333333333334,
      "grad_norm": 0.05786745995283127,
      "learning_rate": 3.416666666666667e-06,
      "loss": 0.0668,
      "num_tokens": 16121552.0,
      "reward": 1.1129639148712158,
      "reward_std": 0.2750511169433594,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.529980480670929,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 77
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.102487066006468e-06,
      "aux_brier/mean_group_std": 0.06814221900666223,
      "aux_brier/mean_r": 0.9545244499354607,
      "aux_brier/n_active_tok": 158.25,
      "aux_brier/n_groups": 8.5,
      "aux_brier/n_step_records": 39.5625,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5554151273575733,
      "calib/avg_num_step_conf": 5.05859375,
      "calib/ece": 0.40772,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0007693304815606858,
      "calib/mean_conf": 0.03732,
      "calib/mu_c": 0.03774774774774774,
      "calib/mu_w": 0.03697841726618705,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0005200000000000001,
      "calib/std_conf": 0.015633860687622875,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2256.0,
      "completions/max_terminated_length": 2256.0,
      "completions/mean_length": 374.73046875,
      "completions/mean_terminated_length": 379.1739196777344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 120.0,
      "epoch": 0.0832,
      "grad_norm": 0.08024071902036667,
      "learning_rate": 3.3888888888888893e-06,
      "loss": -0.0005,
      "num_tokens": 16325507.0,
      "reward": 1.0624713897705078,
      "reward_std": 0.28480181097984314,
      "rewards/accuracy_reward_step": 0.43359375,
      "rewards/final_brier_reward_step": 0.5701979994773865,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 78
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.230446156603616e-06,
      "aux_brier/mean_group_std": 0.0627408457648402,
      "aux_brier/mean_r": 0.9599723760079495,
      "aux_brier/n_active_tok": 159.375,
      "aux_brier/n_groups": 9.34375,
      "aux_brier/n_step_records": 39.84375,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5355419847328245,
      "calib/avg_num_step_conf": 5.01171875,
      "calib/ece": 0.45500000000000007,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0014045801526717597,
      "calib/mean_conf": 0.03328125,
      "calib/mu_c": 0.034,
      "calib/mu_w": 0.03259541984732824,
      "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": 0.015692821557562554,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 854.0,
      "completions/max_terminated_length": 854.0,
      "completions/mean_length": 358.53515625,
      "completions/mean_terminated_length": 359.9411926269531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 91.0,
      "epoch": 0.08426666666666667,
      "grad_norm": 0.1831938475370407,
      "learning_rate": 3.3611111111111117e-06,
      "loss": 0.02,
      "num_tokens": 16523668.0,
      "reward": 1.1222201585769653,
      "reward_std": 0.19442899525165558,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.543567955493927,
      "rewards/format_reward_step_strict": 0.99609375,
      "step": 79
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.605331955951584e-06,
      "aux_brier/mean_group_std": 0.07893908027919577,
      "aux_brier/mean_r": 0.9501262667542172,
      "aux_brier/n_active_tok": 163.0,
      "aux_brier/n_groups": 9.71875,
      "aux_brier/n_step_records": 40.75,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4577464788732394,
      "calib/avg_num_step_conf": 5.09375,
      "calib/ece": 0.5284189723320158,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0008685446009389641,
      "calib/mean_conf": 0.032845849802371534,
      "calib/mu_c": 0.03246478873239436,
      "calib/mu_w": 0.033333333333333326,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.015134572643453475,
      "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": 2172.0,
      "completions/max_terminated_length": 2172.0,
      "completions/mean_length": 335.3359375,
      "completions/mean_terminated_length": 336.6510009765625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 124.0,
      "epoch": 0.08533333333333333,
      "grad_norm": 0.011240994557738304,
      "learning_rate": 3.3333333333333333e-06,
      "loss": 0.0131,
      "num_tokens": 16711674.0,
      "reward": 1.165907382965088,
      "reward_std": 0.26851609349250793,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.46831679344177246,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 80
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.511620979837707e-06,
      "aux_brier/mean_group_std": 0.06015123286817331,
      "aux_brier/mean_r": 0.9625598457159206,
      "aux_brier/n_active_tok": 162.625,
      "aux_brier/n_groups": 10.34375,
      "aux_brier/n_step_records": 40.65625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.4599961442066705,
      "calib/avg_num_step_conf": 5.09765625,
      "calib/ece": 0.501104,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.001850266692371958,
      "calib/mean_conf": 0.030896,
      "calib/mu_c": 0.030030075187969924,
      "calib/mu_w": 0.03188034188034188,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.010532862099163742,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 388.05078125,
      "completions/mean_terminated_length": 389.57257080078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 94.0,
      "epoch": 0.0864,
      "grad_norm": 0.054194867610931396,
      "learning_rate": 3.3055555555555558e-06,
      "loss": 0.0418,
      "num_tokens": 16917263.0,
      "reward": 1.1276578903198242,
      "reward_std": 0.2511928677558899,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.48719385266304016,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 81
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -7.117875442164134e-06,
      "aux_brier/mean_group_std": 0.047760515298038604,
      "aux_brier/mean_r": 0.9708146194599993,
      "aux_brier/n_active_tok": 148.125,
      "aux_brier/n_groups": 8.03125,
      "aux_brier/n_step_records": 37.03125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5342366700201207,
      "calib/avg_num_step_conf": 4.82421875,
      "calib/ece": 0.5280708661417323,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0014411468812877241,
      "calib/mean_conf": 0.030984251968503934,
      "calib/mu_c": 0.03161971830985916,
      "calib/mu_w": 0.030178571428571433,
      "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": 0.009811999926145797,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2004.0,
      "completions/max_terminated_length": 2004.0,
      "completions/mean_length": 318.28125,
      "completions/mean_terminated_length": 319.5294189453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 84.0,
      "epoch": 0.08746666666666666,
      "grad_norm": 0.10533162951469421,
      "learning_rate": 3.277777777777778e-06,
      "loss": 0.0102,
      "num_tokens": 17104295.0,
      "reward": 1.168663740158081,
      "reward_std": 0.22765403985977173,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.47153007984161377,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 82
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.144158094965042e-05,
      "aux_brier/mean_group_std": 0.06297535992917615,
      "aux_brier/mean_r": 0.9501589902161882,
      "aux_brier/n_active_tok": 171.5,
      "aux_brier/n_groups": 10.125,
      "aux_brier/n_step_records": 42.875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5132980399571437,
      "calib/avg_num_step_conf": 5.546875,
      "calib/ece": 0.48182539682539677,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.000162601626016258,
      "calib/mean_conf": 0.03007936507936508,
      "calib/mu_c": 0.030000000000000002,
      "calib/mu_w": 0.03016260162601626,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.010540626750591659,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 1712.0,
      "completions/max_terminated_length": 1712.0,
      "completions/mean_length": 376.55859375,
      "completions/mean_terminated_length": 381.02374267578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 113.0,
      "epoch": 0.08853333333333334,
      "grad_norm": 0.015030056238174438,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.0041,
      "num_tokens": 17307958.0,
      "reward": 1.1205906867980957,
      "reward_std": 0.2422322928905487,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.5058003664016724,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 83
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.0064067055531867e-05,
      "aux_brier/mean_group_std": 0.07812349659141118,
      "aux_brier/mean_r": 0.946900485161512,
      "aux_brier/n_active_tok": 152.5,
      "aux_brier/n_groups": 8.5625,
      "aux_brier/n_step_records": 38.125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4547616068076586,
      "calib/avg_num_step_conf": 4.765625,
      "calib/ece": 0.4519762845849802,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0020410461769490643,
      "calib/mean_conf": 0.030237154150197627,
      "calib/mu_c": 0.02918032786885246,
      "calib/mu_w": 0.031221374045801525,
      "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": 0.013685595952137801,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2304.0,
      "completions/max_terminated_length": 2304.0,
      "completions/mean_length": 336.703125,
      "completions/mean_terminated_length": 336.703125,
      "completions/min_length": 104.0,
      "completions/min_terminated_length": 104.0,
      "epoch": 0.0896,
      "grad_norm": 0.0301885474473238,
      "learning_rate": 3.2222222222222227e-06,
      "loss": 0.0855,
      "num_tokens": 17500074.0,
      "reward": 1.105313777923584,
      "reward_std": 0.22371599078178406,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.5384425520896912,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 84
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.720426093423157e-06,
      "aux_brier/mean_group_std": 0.06952150267362742,
      "aux_brier/mean_r": 0.9518234900073032,
      "aux_brier/n_active_tok": 170.75,
      "aux_brier/n_groups": 10.78125,
      "aux_brier/n_step_records": 42.6875,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5591596638655462,
      "calib/avg_num_step_conf": 5.5546875,
      "calib/ece": 0.45745901639344266,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.001692773109243692,
      "calib/mean_conf": 0.031065573770491805,
      "calib/mu_c": 0.031932773109243695,
      "calib/mu_w": 0.030240000000000003,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.00040983606557377055,
      "calib/std_conf": 0.010999804547382808,
      "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": 2923.0,
      "completions/max_terminated_length": 2923.0,
      "completions/mean_length": 418.2734375,
      "completions/mean_terminated_length": 421.5669250488281,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.09066666666666667,
      "grad_norm": 0.023108787834644318,
      "learning_rate": 3.1944444444444443e-06,
      "loss": 0.1353,
      "num_tokens": 17714976.0,
      "reward": 1.0706396102905273,
      "reward_std": 0.27474746108055115,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5169335603713989,
      "rewards/format_reward_step_strict": 0.953125,
      "step": 85
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.3089714198686337e-06,
      "aux_brier/mean_group_std": 0.0705696843840126,
      "aux_brier/mean_r": 0.9573568341197747,
      "aux_brier/n_active_tok": 166.875,
      "aux_brier/n_groups": 10.09375,
      "aux_brier/n_step_records": 41.71875,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5020595394120951,
      "calib/avg_num_step_conf": 5.265625,
      "calib/ece": 0.39404296875,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 8.831055357922529e-05,
      "calib/mean_conf": 0.03173828125000001,
      "calib/mu_c": 0.03178899082568807,
      "calib/mu_w": 0.031700680272108844,
      "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": 0.009175260734491333,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 940.0,
      "completions/max_terminated_length": 940.0,
      "completions/mean_length": 354.41015625,
      "completions/mean_terminated_length": 355.8000183105469,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 95.0,
      "epoch": 0.09173333333333333,
      "grad_norm": 0.07036297768354416,
      "learning_rate": 3.1666666666666667e-06,
      "loss": 0.0081,
      "num_tokens": 17911217.0,
      "reward": 1.0738774538040161,
      "reward_std": 0.2070370614528656,
      "rewards/accuracy_reward_step": 0.42578125,
      "rewards/final_brier_reward_step": 0.6001975536346436,
      "rewards/format_reward_step_strict": 0.99609375,
      "step": 86
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.883740235801405e-06,
      "aux_brier/mean_group_std": 0.08489868975486516,
      "aux_brier/mean_r": 0.9432929140526566,
      "aux_brier/n_active_tok": 158.625,
      "aux_brier/n_groups": 9.25,
      "aux_brier/n_step_records": 39.65625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4946864340068224,
      "calib/avg_num_step_conf": 5.3046875,
      "calib/ece": 0.5568593625498007,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003730313565993175,
      "calib/mean_conf": 0.03278207171314741,
      "calib/mu_c": 0.03431283783783784,
      "calib/mu_w": 0.030582524271844665,
      "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": 0.029066496733548594,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2376.0,
      "completions/max_terminated_length": 2376.0,
      "completions/mean_length": 346.8515625,
      "completions/mean_terminated_length": 350.9644470214844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 103.0,
      "epoch": 0.0928,
      "grad_norm": 0.034281790256500244,
      "learning_rate": 3.138888888888889e-06,
      "loss": 0.0222,
      "num_tokens": 18105507.0,
      "reward": 1.1783933639526367,
      "reward_std": 0.2476852387189865,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.4401359558105469,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 87
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 9.146717485641886e-07,
      "aux_brier/mean_group_std": 0.07024082438078165,
      "aux_brier/mean_r": 0.9578755974288631,
      "aux_brier/n_active_tok": 167.875,
      "aux_brier/n_groups": 8.59375,
      "aux_brier/n_step_records": 41.96875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.506267716535433,
      "calib/avg_num_step_conf": 5.25390625,
      "calib/ece": 0.4708333333333333,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00018834645669291855,
      "calib/mean_conf": 0.033134920634920635,
      "calib/mu_c": 0.03322834645669292,
      "calib/mu_w": 0.03304,
      "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": 0.011131146901506041,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2309.0,
      "completions/max_terminated_length": 2309.0,
      "completions/mean_length": 361.390625,
      "completions/mean_terminated_length": 364.2362060546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 105.0,
      "epoch": 0.09386666666666667,
      "grad_norm": 0.04726596921682358,
      "learning_rate": 3.1111111111111116e-06,
      "loss": 0.059,
      "num_tokens": 18307871.0,
      "reward": 1.116339921951294,
      "reward_std": 0.22101491689682007,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5200472474098206,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 88
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.687176872129271e-07,
      "aux_brier/mean_group_std": 0.07569657181508732,
      "aux_brier/mean_r": 0.9485480143526361,
      "aux_brier/n_active_tok": 185.25,
      "aux_brier/n_groups": 11.09375,
      "aux_brier/n_step_records": 46.3125,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5373424759080802,
      "calib/avg_num_step_conf": 5.8125,
      "calib/ece": 0.4155078125,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0007499382258462986,
      "calib/mean_conf": 0.0333984375,
      "calib/mu_c": 0.03298245614035088,
      "calib/mu_w": 0.033732394366197176,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0017968749999999999,
      "calib/std_conf": 0.017404130186785945,
      "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": 913.0,
      "completions/max_terminated_length": 913.0,
      "completions/mean_length": 365.80859375,
      "completions/mean_terminated_length": 367.2431640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 104.0,
      "epoch": 0.09493333333333333,
      "grad_norm": 0.046813610941171646,
      "learning_rate": 3.0833333333333336e-06,
      "loss": 0.0114,
      "num_tokens": 18510406.0,
      "reward": 1.0880454778671265,
      "reward_std": 0.1457396298646927,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.5787441730499268,
      "rewards/format_reward_step_strict": 0.99609375,
      "step": 89
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.3200514974730524e-06,
      "aux_brier/mean_group_std": 0.08874093639098773,
      "aux_brier/mean_r": 0.9315546717243935,
      "aux_brier/n_active_tok": 185.5,
      "aux_brier/n_groups": 10.5,
      "aux_brier/n_step_records": 46.375,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5146074444791993,
      "calib/avg_num_step_conf": 5.8125,
      "calib/ece": 0.5155118110236221,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0006230841413825405,
      "calib/mean_conf": 0.03173228346456693,
      "calib/mu_c": 0.03201438848920863,
      "calib/mu_w": 0.03139130434782609,
      "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": 0.008702835829699591,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 355.82421875,
      "completions/mean_terminated_length": 357.2196350097656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 89.0,
      "epoch": 0.096,
      "grad_norm": 0.026251893490552902,
      "learning_rate": 3.055555555555556e-06,
      "loss": 0.0506,
      "num_tokens": 18704817.0,
      "reward": 1.1636964082717896,
      "reward_std": 0.2448050081729889,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.48291015625,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 90
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.9404183573843454e-06,
      "aux_brier/mean_group_std": 0.05313952584804532,
      "aux_brier/mean_r": 0.9705964411583399,
      "aux_brier/n_active_tok": 204.0,
      "aux_brier/n_groups": 13.375,
      "aux_brier/n_step_records": 51.0,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.49686879960317465,
      "calib/avg_num_step_conf": 6.375,
      "calib/ece": 0.47090551181102364,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 3.1001984126990956e-05,
      "calib/mean_conf": 0.033031496062992126,
      "calib/mu_c": 0.033046875,
      "calib/mu_w": 0.03301587301587301,
      "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": 0.009550064523733063,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2646.0,
      "completions/max_terminated_length": 2646.0,
      "completions/mean_length": 398.25,
      "completions/mean_terminated_length": 398.25,
      "completions/min_length": 122.0,
      "completions/min_terminated_length": 122.0,
      "epoch": 0.09706666666666666,
      "grad_norm": 0.04937665909528732,
      "learning_rate": 3.0277777777777776e-06,
      "loss": 0.0691,
      "num_tokens": 18914481.0,
      "reward": 1.1271090507507324,
      "reward_std": 0.16296067833900452,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.5240613222122192,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 91
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -5.775527954110515e-06,
      "aux_brier/mean_group_std": 0.03261796387138986,
      "aux_brier/mean_r": 0.9787621676505022,
      "aux_brier/n_active_tok": 169.375,
      "aux_brier/n_groups": 9.5625,
      "aux_brier/n_step_records": 42.34375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5343565891472868,
      "calib/avg_num_step_conf": 5.296875,
      "calib/ece": 0.4596850393700787,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0010468217054263596,
      "calib/mean_conf": 0.03322834645669291,
      "calib/mu_c": 0.033760000000000005,
      "calib/mu_w": 0.032713178294573646,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0003937007874015748,
      "calib/std_conf": 0.009990695652856174,
      "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": 1947.0,
      "completions/max_terminated_length": 1947.0,
      "completions/mean_length": 330.39453125,
      "completions/mean_terminated_length": 330.39453125,
      "completions/min_length": 63.0,
      "completions/min_terminated_length": 63.0,
      "epoch": 0.09813333333333334,
      "grad_norm": 0.028532471507787704,
      "learning_rate": 3e-06,
      "loss": 0.0251,
      "num_tokens": 19105782.0,
      "reward": 1.1143501996994019,
      "reward_std": 0.2323814481496811,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.5355257987976074,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 92
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.171904975982301e-06,
      "aux_brier/mean_group_std": 0.05268081661406212,
      "aux_brier/mean_r": 0.9699430246107784,
      "aux_brier/n_active_tok": 201.375,
      "aux_brier/n_groups": 11.65625,
      "aux_brier/n_step_records": 50.34375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5058885842472582,
      "calib/avg_num_step_conf": 6.29296875,
      "calib/ece": 0.5011023622047245,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0006168993020937152,
      "calib/mean_conf": 0.03433070866141732,
      "calib/mu_c": 0.03404411764705883,
      "calib/mu_w": 0.034661016949152544,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.009888397016745997,
      "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": 1705.0,
      "completions/max_terminated_length": 1705.0,
      "completions/mean_length": 394.0859375,
      "completions/mean_terminated_length": 395.63140869140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 124.0,
      "epoch": 0.0992,
      "grad_norm": 0.0336824506521225,
      "learning_rate": 2.9722222222222225e-06,
      "loss": 0.0405,
      "num_tokens": 19312444.0,
      "reward": 1.1493513584136963,
      "reward_std": 0.2246081531047821,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.49584296345710754,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 93
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -5.373293926247413e-06,
      "aux_brier/mean_group_std": 0.0731912638488667,
      "aux_brier/mean_r": 0.9554281096642034,
      "aux_brier/n_active_tok": 180.375,
      "aux_brier/n_groups": 10.71875,
      "aux_brier/n_step_records": 45.09375,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.6063914616941667,
      "calib/avg_num_step_conf": 5.73046875,
      "calib/ece": 0.5027734375,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003510396859473717,
      "calib/mean_conf": 0.0323828125,
      "calib/mu_c": 0.034014598540145984,
      "calib/mu_w": 0.030504201680672267,
      "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": 0.009691830559282584,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2024.0,
      "completions/max_terminated_length": 2024.0,
      "completions/mean_length": 331.71875,
      "completions/mean_terminated_length": 333.0196228027344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 89.0,
      "epoch": 0.10026666666666667,
      "grad_norm": 0.07165474444627762,
      "learning_rate": 2.944444444444445e-06,
      "loss": -0.0138,
      "num_tokens": 19506044.0,
      "reward": 1.160183072090149,
      "reward_std": 0.17494770884513855,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.5001074075698853,
      "rewards/format_reward_step_strict": 1.0,
      "step": 94
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.3494379560441416e-06,
      "aux_brier/mean_group_std": 0.07180938440355472,
      "aux_brier/mean_r": 0.9510812979632172,
      "aux_brier/n_active_tok": 186.5,
      "aux_brier/n_groups": 10.625,
      "aux_brier/n_step_records": 46.625,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5774968394437421,
      "calib/avg_num_step_conf": 5.97265625,
      "calib/ece": 0.5188537549407115,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0038249051833122646,
      "calib/mean_conf": 0.03450592885375494,
      "calib/mu_c": 0.03621428571428572,
      "calib/mu_w": 0.032389380530973455,
      "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": 0.013316345732329211,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2617.0,
      "completions/max_terminated_length": 2617.0,
      "completions/mean_length": 368.58984375,
      "completions/mean_terminated_length": 370.0353088378906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 84.0,
      "epoch": 0.10133333333333333,
      "grad_norm": 0.061531733721494675,
      "learning_rate": 2.916666666666667e-06,
      "loss": 0.0277,
      "num_tokens": 19706531.0,
      "reward": 1.1648377180099487,
      "reward_std": 0.24357908964157104,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.4796636700630188,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 95
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.363781245544704e-06,
      "aux_brier/mean_group_std": 0.0684968045927621,
      "aux_brier/mean_r": 0.9433583986571044,
      "aux_brier/n_active_tok": 182.375,
      "aux_brier/n_groups": 9.9375,
      "aux_brier/n_step_records": 45.59375,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.549083269671505,
      "calib/avg_num_step_conf": 5.703125,
      "calib/ece": 0.5654296875,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0015571683218742124,
      "calib/mean_conf": 0.03613281250000001,
      "calib/mu_c": 0.03675324675324675,
      "calib/mu_w": 0.03519607843137254,
      "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": 0.010168787333789795,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 882.0,
      "completions/max_terminated_length": 882.0,
      "completions/mean_length": 338.2265625,
      "completions/mean_terminated_length": 339.5529479980469,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 119.0,
      "epoch": 0.1024,
      "grad_norm": 0.06455075740814209,
      "learning_rate": 2.888888888888889e-06,
      "loss": 0.0046,
      "num_tokens": 19898933.0,
      "reward": 1.21187424659729,
      "reward_std": 0.1398114562034607,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.44124725461006165,
      "rewards/format_reward_step_strict": 1.0,
      "step": 96
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 7.579109583100152e-06,
      "aux_brier/mean_group_std": 0.04659628467337417,
      "aux_brier/mean_r": 0.9751859140923029,
      "aux_brier/n_active_tok": 197.875,
      "aux_brier/n_groups": 12.6875,
      "aux_brier/n_step_records": 49.46875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5492060443078499,
      "calib/avg_num_step_conf": 6.18359375,
      "calib/ece": 0.4189641434262948,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0024913561275451343,
      "calib/mean_conf": 0.03521912350597609,
      "calib/mu_c": 0.036578947368421044,
      "calib/mu_w": 0.03408759124087591,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.010907606945228587,
      "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": 2517.0,
      "completions/max_terminated_length": 2517.0,
      "completions/mean_length": 379.0859375,
      "completions/mean_terminated_length": 379.0859375,
      "completions/min_length": 104.0,
      "completions/min_terminated_length": 104.0,
      "epoch": 0.10346666666666667,
      "grad_norm": 0.014394364319741726,
      "learning_rate": 2.861111111111111e-06,
      "loss": 0.0726,
      "num_tokens": 20101051.0,
      "reward": 1.0771472454071045,
      "reward_std": 0.2732182741165161,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.5664016008377075,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 97
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 4.318549504528368e-06,
      "aux_brier/mean_group_std": 0.09819066026645625,
      "aux_brier/mean_r": 0.9276285348741788,
      "aux_brier/n_active_tok": 198.5,
      "aux_brier/n_groups": 12.28125,
      "aux_brier/n_step_records": 49.625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5613399440630562,
      "calib/avg_num_step_conf": 6.203125,
      "calib/ece": 0.5073809523809524,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005240274599542329,
      "calib/mean_conf": 0.04023809523809524,
      "calib/mu_c": 0.04260869565217391,
      "calib/mu_w": 0.03736842105263158,
      "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": 0.025369156554926893,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2803.0,
      "completions/max_terminated_length": 2803.0,
      "completions/mean_length": 402.28125,
      "completions/mean_terminated_length": 402.28125,
      "completions/min_length": 55.0,
      "completions/min_terminated_length": 55.0,
      "epoch": 0.10453333333333334,
      "grad_norm": 0.024372128769755363,
      "learning_rate": 2.8333333333333335e-06,
      "loss": 0.1009,
      "num_tokens": 20310219.0,
      "reward": 1.153505563735962,
      "reward_std": 0.2584661543369293,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.4890226423740387,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 98
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 7.956959314214807e-06,
      "aux_brier/mean_group_std": 0.047173973295863036,
      "aux_brier/mean_r": 0.9749459528583877,
      "aux_brier/n_active_tok": 201.125,
      "aux_brier/n_groups": 11.21875,
      "aux_brier/n_step_records": 50.28125,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5842321520287622,
      "calib/avg_num_step_conf": 6.28515625,
      "calib/ece": 0.2649212598425197,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002251815980629533,
      "calib/mean_conf": 0.042716535433070864,
      "calib/mu_c": 0.044285714285714275,
      "calib/mu_w": 0.04203389830508474,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0022440944881889765,
      "calib/std_conf": 0.017346127861177077,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2905.0,
      "completions/max_terminated_length": 2905.0,
      "completions/mean_length": 411.3984375,
      "completions/mean_terminated_length": 411.3984375,
      "completions/min_length": 104.0,
      "completions/min_terminated_length": 104.0,
      "epoch": 0.1056,
      "grad_norm": 0.006937370169907808,
      "learning_rate": 2.805555555555556e-06,
      "loss": 0.0425,
      "num_tokens": 20521337.0,
      "reward": 0.9758594036102295,
      "reward_std": 0.22256730496883392,
      "rewards/accuracy_reward_step": 0.30078125,
      "rewards/final_brier_reward_step": 0.7159379124641418,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 99
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 3.2180199117937125e-06,
      "aux_brier/mean_group_std": 0.078943701829868,
      "aux_brier/mean_r": 0.9410495528956415,
      "aux_brier/n_active_tok": 185.25,
      "aux_brier/n_groups": 9.78125,
      "aux_brier/n_step_records": 46.3125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6130044561601707,
      "calib/avg_num_step_conf": 5.83203125,
      "calib/ece": 0.4011417322834646,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005856398669428241,
      "calib/mean_conf": 0.043740157480314956,
      "calib/mu_c": 0.04699115044247788,
      "calib/mu_w": 0.04113475177304964,
      "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": 0.012287060345896567,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1845.0,
      "completions/max_terminated_length": 1845.0,
      "completions/mean_length": 370.84375,
      "completions/mean_terminated_length": 372.2980651855469,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 108.0,
      "epoch": 0.10666666666666667,
      "grad_norm": 0.021300625056028366,
      "learning_rate": 2.7777777777777783e-06,
      "loss": 0.0129,
      "num_tokens": 20723681.0,
      "reward": 1.0831012725830078,
      "reward_std": 0.2674819231033325,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.5902175903320312,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 100
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 8.315309462103571e-07,
      "aux_brier/mean_group_std": 0.05673601983434623,
      "aux_brier/mean_r": 0.9700530041716272,
      "aux_brier/n_active_tok": 218.0,
      "aux_brier/n_groups": 12.5,
      "aux_brier/n_step_records": 54.5,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6083049068485962,
      "calib/avg_num_step_conf": 7.10546875,
      "calib/ece": 0.3651394422310757,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00596759380739964,
      "calib/mean_conf": 0.0452191235059761,
      "calib/mu_c": 0.0487378640776699,
      "calib/mu_w": 0.04277027027027026,
      "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": 0.01406628420809822,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2435.0,
      "completions/max_terminated_length": 2435.0,
      "completions/mean_length": 399.171875,
      "completions/mean_terminated_length": 403.9051513671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 104.0,
      "epoch": 0.10773333333333333,
      "grad_norm": 0.09895070642232895,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.0022,
      "num_tokens": 20932861.0,
      "reward": 1.0434354543685913,
      "reward_std": 0.2181820571422577,
      "rewards/accuracy_reward_step": 0.40234375,
      "rewards/final_brier_reward_step": 0.6112421751022339,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 101
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.102560478199459e-06,
      "aux_brier/mean_group_std": 0.05331271846939198,
      "aux_brier/mean_r": 0.9617336874931215,
      "aux_brier/n_active_tok": 196.625,
      "aux_brier/n_groups": 11.3125,
      "aux_brier/n_step_records": 49.15625,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6790845518118245,
      "calib/avg_num_step_conf": 6.1640625,
      "calib/ece": 0.5193675889328062,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.006489510489510485,
      "calib/mean_conf": 0.0458498023715415,
      "calib/mu_c": 0.048671328671328666,
      "calib/mu_w": 0.04218181818181818,
      "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": 0.014078028249992373,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2670.0,
      "completions/max_terminated_length": 2670.0,
      "completions/mean_length": 342.27734375,
      "completions/mean_terminated_length": 343.61962890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 98.0,
      "epoch": 0.1088,
      "grad_norm": 0.04714421182870865,
      "learning_rate": 2.7222222222222224e-06,
      "loss": -0.0133,
      "num_tokens": 21127180.0,
      "reward": 1.1712284088134766,
      "reward_std": 0.20991604030132294,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.48178908228874207,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 102
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.0155701142554605e-06,
      "aux_brier/mean_group_std": 0.051797072497869974,
      "aux_brier/mean_r": 0.9657128892684819,
      "aux_brier/n_active_tok": 208.25,
      "aux_brier/n_groups": 12.4375,
      "aux_brier/n_step_records": 52.0625,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5553644041288707,
      "calib/avg_num_step_conf": 6.51171875,
      "calib/ece": 0.49925196850393705,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0023715983734751214,
      "calib/mean_conf": 0.04964566929133858,
      "calib/mu_c": 0.05071942446043165,
      "calib/mu_w": 0.04834782608695653,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0008267716535433072,
      "calib/std_conf": 0.012869439818615358,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1714.0,
      "completions/max_terminated_length": 1714.0,
      "completions/mean_length": 410.11328125,
      "completions/mean_terminated_length": 410.11328125,
      "completions/min_length": 118.0,
      "completions/min_terminated_length": 118.0,
      "epoch": 0.10986666666666667,
      "grad_norm": 0.015716806054115295,
      "learning_rate": 2.6944444444444444e-06,
      "loss": 0.0166,
      "num_tokens": 21336721.0,
      "reward": 1.1644841432571411,
      "reward_std": 0.16049611568450928,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.5016871094703674,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 103
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.2022423732392937e-06,
      "aux_brier/mean_group_std": 0.0487820049126365,
      "aux_brier/mean_r": 0.9658913802581619,
      "aux_brier/n_active_tok": 201.375,
      "aux_brier/n_groups": 11.15625,
      "aux_brier/n_step_records": 50.34375,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5270082184517497,
      "calib/avg_num_step_conf": 6.4296875,
      "calib/ece": 0.30792968749999994,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00334305408271475,
      "calib/mean_conf": 0.05144531250000001,
      "calib/mu_c": 0.05358695652173913,
      "calib/mu_w": 0.05024390243902438,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.015732782550373716,
      "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": 1044.0,
      "completions/max_terminated_length": 1044.0,
      "completions/mean_length": 369.2109375,
      "completions/mean_terminated_length": 370.6588439941406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.11093333333333333,
      "grad_norm": 0.03981407731771469,
      "learning_rate": 2.666666666666667e-06,
      "loss": -0.0108,
      "num_tokens": 21537919.0,
      "reward": 1.019655466079712,
      "reward_std": 0.21034060418605804,
      "rewards/accuracy_reward_step": 0.359375,
      "rewards/final_brier_reward_step": 0.6723718643188477,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 104
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.0062088509641427e-06,
      "aux_brier/mean_group_std": 0.04389778124056223,
      "aux_brier/mean_r": 0.9772179865062308,
      "aux_brier/n_active_tok": 202.125,
      "aux_brier/n_groups": 11.5625,
      "aux_brier/n_step_records": 50.53125,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5621921182266011,
      "calib/avg_num_step_conf": 6.33203125,
      "calib/ece": 0.4012890625,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002948275862068965,
      "calib/mean_conf": 0.05183593750000001,
      "calib/mu_c": 0.05344827586206897,
      "calib/mu_w": 0.0505,
      "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": 0.014742199920503513,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1866.0,
      "completions/max_terminated_length": 1866.0,
      "completions/mean_length": 393.2265625,
      "completions/mean_terminated_length": 394.7686462402344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 131.0,
      "epoch": 0.112,
      "grad_norm": 0.025966746732592583,
      "learning_rate": 2.6388888888888893e-06,
      "loss": -0.0021,
      "num_tokens": 21744345.0,
      "reward": 1.098300814628601,
      "reward_std": 0.26817578077316284,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.5885159969329834,
      "rewards/format_reward_step_strict": 0.99609375,
      "step": 105
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 8.061786714863572e-07,
      "aux_brier/mean_group_std": 0.08333754965469,
      "aux_brier/mean_r": 0.9468736608927933,
      "aux_brier/n_active_tok": 198.375,
      "aux_brier/n_groups": 10.96875,
      "aux_brier/n_step_records": 49.59375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5614661654135338,
      "calib/avg_num_step_conf": 6.19921875,
      "calib/ece": 0.4209881422924901,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0017687969924812252,
      "calib/mean_conf": 0.05332015810276681,
      "calib/mu_c": 0.054250000000000013,
      "calib/mu_w": 0.05248120300751879,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.014143682117769804,
      "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": 3029.0,
      "completions/max_terminated_length": 3029.0,
      "completions/mean_length": 390.21484375,
      "completions/mean_terminated_length": 390.21484375,
      "completions/min_length": 99.0,
      "completions/min_terminated_length": 99.0,
      "epoch": 0.11306666666666666,
      "grad_norm": 0.010917603969573975,
      "learning_rate": 2.6111111111111113e-06,
      "loss": 0.086,
      "num_tokens": 21948824.0,
      "reward": 1.1026880741119385,
      "reward_std": 0.19752243161201477,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5670023560523987,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 106
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.976572967609673e-06,
      "aux_brier/mean_group_std": 0.06270684416678649,
      "aux_brier/mean_r": 0.9626215013030841,
      "aux_brier/n_active_tok": 210.25,
      "aux_brier/n_groups": 11.15625,
      "aux_brier/n_step_records": 52.5625,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5606106870229008,
      "calib/avg_num_step_conf": 6.62109375,
      "calib/ece": 0.45664062499999997,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003357557251908394,
      "calib/mean_conf": 0.05507812500000002,
      "calib/mu_c": 0.05671755725190839,
      "calib/mu_w": 0.05336,
      "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": 0.01581119528955275,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1217.0,
      "completions/max_terminated_length": 1217.0,
      "completions/mean_length": 366.03125,
      "completions/mean_terminated_length": 367.4666748046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 107.0,
      "epoch": 0.11413333333333334,
      "grad_norm": 0.10252933204174042,
      "learning_rate": 2.5833333333333337e-06,
      "loss": 0.0131,
      "num_tokens": 22147144.0,
      "reward": 1.1474798917770386,
      "reward_std": 0.22707661986351013,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.5430445671081543,
      "rewards/format_reward_step_strict": 1.0,
      "step": 107
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.7100047952921535e-06,
      "aux_brier/mean_group_std": 0.07736919953269787,
      "aux_brier/mean_r": 0.9485736905461646,
      "aux_brier/n_active_tok": 204.75,
      "aux_brier/n_groups": 10.6875,
      "aux_brier/n_step_records": 51.1875,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5773251139104797,
      "calib/avg_num_step_conf": 6.44140625,
      "calib/ece": 0.5821411764705882,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004966094880729029,
      "calib/mean_conf": 0.060996078431372565,
      "calib/mu_c": 0.06276829268292683,
      "calib/mu_w": 0.0578021978021978,
      "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": 0.017661634068076287,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1509.0,
      "completions/max_terminated_length": 1509.0,
      "completions/mean_length": 399.640625,
      "completions/mean_terminated_length": 401.2078552246094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 106.0,
      "epoch": 0.1152,
      "grad_norm": 0.06794151663780212,
      "learning_rate": 2.5555555555555557e-06,
      "loss": -0.0184,
      "num_tokens": 22352684.0,
      "reward": 1.2485934495925903,
      "reward_std": 0.23719748854637146,
      "rewards/accuracy_reward_step": 0.64453125,
      "rewards/final_brier_reward_step": 0.4318739175796509,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 108
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.611892308490571e-06,
      "aux_brier/mean_group_std": 0.06195867460477537,
      "aux_brier/mean_r": 0.9659657078520748,
      "aux_brier/n_active_tok": 240.75,
      "aux_brier/n_groups": 13.875,
      "aux_brier/n_step_records": 60.1875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.637492182614134,
      "calib/avg_num_step_conf": 7.53125,
      "calib/ece": 0.42561264822134387,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0059987492182614205,
      "calib/mean_conf": 0.06260869565217392,
      "calib/mu_c": 0.06569105691056912,
      "calib/mu_w": 0.0596923076923077,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0010276679841897233,
      "calib/std_conf": 0.02386376525141341,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2190.0,
      "completions/max_terminated_length": 2190.0,
      "completions/mean_length": 445.56640625,
      "completions/mean_terminated_length": 445.56640625,
      "completions/min_length": 140.0,
      "completions/min_terminated_length": 140.0,
      "epoch": 0.11626666666666667,
      "grad_norm": 0.0783458799123764,
      "learning_rate": 2.5277777777777778e-06,
      "loss": 0.091,
      "num_tokens": 22571349.0,
      "reward": 1.112328290939331,
      "reward_std": 0.18900692462921143,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.5665007829666138,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 109
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.9528525646572703e-06,
      "aux_brier/mean_group_std": 0.05912420198305332,
      "aux_brier/mean_r": 0.9646633949108119,
      "aux_brier/n_active_tok": 202.25,
      "aux_brier/n_groups": 11.6875,
      "aux_brier/n_step_records": 50.5625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.491933418693982,
      "calib/avg_num_step_conf": 6.37109375,
      "calib/ece": 0.3733333333333333,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.002245838668373877,
      "calib/mean_conf": 0.06317460317460317,
      "calib/mu_c": 0.06190909090909091,
      "calib/mu_w": 0.06415492957746478,
      "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": 0.02035988873464131,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2042.0,
      "completions/max_terminated_length": 2042.0,
      "completions/mean_length": 387.70703125,
      "completions/mean_terminated_length": 390.75982666015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 66.0,
      "epoch": 0.11733333333333333,
      "grad_norm": 0.12243294715881348,
      "learning_rate": 2.5e-06,
      "loss": -0.0142,
      "num_tokens": 22775522.0,
      "reward": 1.072763442993164,
      "reward_std": 0.2849552035331726,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.6035538911819458,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 110
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 4.18118770684206e-06,
      "aux_brier/mean_group_std": 0.05934831866230545,
      "aux_brier/mean_r": 0.9594280376685048,
      "aux_brier/n_active_tok": 206.125,
      "aux_brier/n_groups": 11.25,
      "aux_brier/n_step_records": 51.53125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5145471464019851,
      "calib/avg_num_step_conf": 6.4609375,
      "calib/ece": 0.418503937007874,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0022208436724565767,
      "calib/mean_conf": 0.06968503937007875,
      "calib/mu_c": 0.06854838709677419,
      "calib/mu_w": 0.07076923076923076,
      "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": 0.02389072919636157,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1435.0,
      "completions/max_terminated_length": 1435.0,
      "completions/mean_length": 399.2109375,
      "completions/mean_terminated_length": 402.3543395996094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 100.0,
      "epoch": 0.1184,
      "grad_norm": 0.0319228358566761,
      "learning_rate": 2.4722222222222226e-06,
      "loss": -0.0364,
      "num_tokens": 22985128.0,
      "reward": 1.1226773262023926,
      "reward_std": 0.2619039714336395,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.5688344240188599,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 111
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.869699991829954e-06,
      "aux_brier/mean_group_std": 0.08900723429146168,
      "aux_brier/mean_r": 0.9368204580656649,
      "aux_brier/n_active_tok": 219.5,
      "aux_brier/n_groups": 14.0,
      "aux_brier/n_step_records": 54.875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6206819913639827,
      "calib/avg_num_step_conf": 6.91015625,
      "calib/ece": 0.42372509960159355,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00807340614681229,
      "calib/mean_conf": 0.07268924302788846,
      "calib/mu_c": 0.0767741935483871,
      "calib/mu_w": 0.06870078740157481,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.001195219123505976,
      "calib/std_conf": 0.03017939463248341,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2916.0,
      "completions/max_terminated_length": 2916.0,
      "completions/mean_length": 435.79296875,
      "completions/mean_terminated_length": 439.2243957519531,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 124.0,
      "epoch": 0.11946666666666667,
      "grad_norm": 0.04229222610592842,
      "learning_rate": 2.4444444444444447e-06,
      "loss": 0.041,
      "num_tokens": 23204611.0,
      "reward": 1.1137549877166748,
      "reward_std": 0.22391203045845032,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.5643951892852783,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 112
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.837179632313557e-06,
      "aux_brier/mean_group_std": 0.07313167256655854,
      "aux_brier/mean_r": 0.9554314023548512,
      "aux_brier/n_active_tok": 212.875,
      "aux_brier/n_groups": 11.59375,
      "aux_brier/n_step_records": 53.21875,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.534562355730239,
      "calib/avg_num_step_conf": 6.7265625,
      "calib/ece": 0.4660236220472441,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0011997005427662466,
      "calib/mean_conf": 0.0780708661417323,
      "calib/mu_c": 0.07751824817518249,
      "calib/mu_w": 0.07871794871794874,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.002362204724409449,
      "calib/std_conf": 0.031470191145683585,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1244.0,
      "completions/max_terminated_length": 1244.0,
      "completions/mean_length": 379.30078125,
      "completions/mean_terminated_length": 380.78826904296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 114.0,
      "epoch": 0.12053333333333334,
      "grad_norm": 0.05542054399847984,
      "learning_rate": 2.4166666666666667e-06,
      "loss": 0.0012,
      "num_tokens": 23406912.0,
      "reward": 1.1644923686981201,
      "reward_std": 0.22591552138328552,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.5329699516296387,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 113
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.435252810910175e-07,
      "aux_brier/mean_group_std": 0.05343525777276535,
      "aux_brier/mean_r": 0.9654501302267348,
      "aux_brier/n_active_tok": 215.875,
      "aux_brier/n_groups": 13.0,
      "aux_brier/n_step_records": 53.96875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4301480051480051,
      "calib/avg_num_step_conf": 6.74609375,
      "calib/ece": 0.5135573122529645,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.01527992277992278,
      "calib/mean_conf": 0.0824901185770751,
      "calib/mu_c": 0.07614864864864866,
      "calib/mu_w": 0.09142857142857144,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.005533596837944663,
      "calib/std_conf": 0.04936478819386015,
      "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": 2761.0,
      "completions/max_terminated_length": 2761.0,
      "completions/mean_length": 398.90625,
      "completions/mean_terminated_length": 400.4706115722656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 134.0,
      "epoch": 0.1216,
      "grad_norm": 0.057506460696458817,
      "learning_rate": 2.388888888888889e-06,
      "loss": 0.0129,
      "num_tokens": 23614056.0,
      "reward": 1.1900386810302734,
      "reward_std": 0.23270046710968018,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.4867175817489624,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 114
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 3.736093342643354e-06,
      "aux_brier/mean_group_std": 0.04719172928977428,
      "aux_brier/mean_r": 0.966516740672227,
      "aux_brier/n_active_tok": 214.5,
      "aux_brier/n_groups": 12.0625,
      "aux_brier/n_step_records": 53.625,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5499689054726369,
      "calib/avg_num_step_conf": 6.90625,
      "calib/ece": 0.44574803149606307,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0024850746268656604,
      "calib/mean_conf": 0.08181102362204724,
      "calib/mu_c": 0.08298507462686568,
      "calib/mu_w": 0.08050000000000002,
      "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": 0.02849001269500677,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1399.0,
      "completions/max_terminated_length": 1399.0,
      "completions/mean_length": 383.9921875,
      "completions/mean_terminated_length": 387.0157470703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 88.0,
      "epoch": 0.12266666666666666,
      "grad_norm": 0.02982662245631218,
      "learning_rate": 2.361111111111111e-06,
      "loss": -0.0267,
      "num_tokens": 23817622.0,
      "reward": 1.1565759181976318,
      "reward_std": 0.26518166065216064,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.5481789112091064,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 115
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.1013238353795316e-06,
      "aux_brier/mean_group_std": 0.03578971010904936,
      "aux_brier/mean_r": 0.97549386885622,
      "aux_brier/n_active_tok": 229.375,
      "aux_brier/n_groups": 13.9375,
      "aux_brier/n_step_records": 57.34375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4629319955406912,
      "calib/avg_num_step_conf": 7.171875,
      "calib/ece": 0.4496588235294118,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004413229282794515,
      "calib/mean_conf": 0.09151764705882354,
      "calib/mu_c": 0.0894927536231884,
      "calib/mu_w": 0.09390598290598291,
      "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": 0.03734637397881659,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1759.0,
      "completions/max_terminated_length": 1759.0,
      "completions/mean_length": 452.3984375,
      "completions/mean_terminated_length": 454.1725769042969,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 76.0,
      "epoch": 0.12373333333333333,
      "grad_norm": 0.08965926617383957,
      "learning_rate": 2.3333333333333336e-06,
      "loss": 0.0048,
      "num_tokens": 24037956.0,
      "reward": 1.168178677558899,
      "reward_std": 0.24060563743114471,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5399023294448853,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 116
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.7152085230474086e-06,
      "aux_brier/mean_group_std": 0.04103590002150434,
      "aux_brier/mean_r": 0.9730136647613208,
      "aux_brier/n_active_tok": 241.875,
      "aux_brier/n_groups": 14.53125,
      "aux_brier/n_step_records": 60.46875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5059576069233663,
      "calib/avg_num_step_conf": 7.55859375,
      "calib/ece": 0.3790909090909091,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0008014549103223245,
      "calib/mean_conf": 0.09126482213438737,
      "calib/mu_c": 0.0908403361344538,
      "calib/mu_w": 0.09164179104477613,
      "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": 0.032388163120866516,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2772.0,
      "completions/max_terminated_length": 2772.0,
      "completions/mean_length": 434.98828125,
      "completions/mean_terminated_length": 434.98828125,
      "completions/min_length": 145.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.1248,
      "grad_norm": 0.07783927023410797,
      "learning_rate": 2.305555555555556e-06,
      "loss": 0.0594,
      "num_tokens": 24255913.0,
      "reward": 1.1086399555206299,
      "reward_std": 0.233639657497406,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5986223220825195,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 117
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.994761994061349e-07,
      "aux_brier/mean_group_std": 0.05193975383776459,
      "aux_brier/mean_r": 0.9674305220500208,
      "aux_brier/n_active_tok": 228.25,
      "aux_brier/n_groups": 13.6875,
      "aux_brier/n_step_records": 57.0625,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.45714642678660666,
      "calib/avg_num_step_conf": 7.13671875,
      "calib/ece": 0.46248031496062986,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.00930534732633681,
      "calib/mean_conf": 0.08649606299212598,
      "calib/mu_c": 0.08224637681159422,
      "calib/mu_w": 0.09155172413793103,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0028346456692913387,
      "calib/std_conf": 0.03446906041992227,
      "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": 2277.0,
      "completions/max_terminated_length": 2277.0,
      "completions/mean_length": 412.62890625,
      "completions/mean_terminated_length": 414.2470703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 152.0,
      "epoch": 0.12586666666666665,
      "grad_norm": 0.04219573363661766,
      "learning_rate": 2.277777777777778e-06,
      "loss": 0.0051,
      "num_tokens": 24465554.0,
      "reward": 1.168454885482788,
      "reward_std": 0.19997341930866241,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5331949591636658,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 118
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.0914254302690773e-06,
      "aux_brier/mean_group_std": 0.06396913642304744,
      "aux_brier/mean_r": 0.9579897961784033,
      "aux_brier/n_active_tok": 234.75,
      "aux_brier/n_groups": 13.46875,
      "aux_brier/n_step_records": 58.6875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5385101010101011,
      "calib/avg_num_step_conf": 7.4140625,
      "calib/ece": 0.4386904761904762,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00046969696969698105,
      "calib/mean_conf": 0.08757936507936508,
      "calib/mu_c": 0.08780303030303031,
      "calib/mu_w": 0.08733333333333333,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.00123015873015873,
      "calib/std_conf": 0.028869122620313853,
      "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": 2284.0,
      "completions/max_terminated_length": 2284.0,
      "completions/mean_length": 460.0234375,
      "completions/mean_terminated_length": 461.8274841308594,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 65.0,
      "epoch": 0.12693333333333334,
      "grad_norm": 0.025566641241312027,
      "learning_rate": 2.25e-06,
      "loss": 0.0355,
      "num_tokens": 24688384.0,
      "reward": 1.1455440521240234,
      "reward_std": 0.2643626928329468,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.5509262084960938,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 119
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.7104528945720574e-06,
      "aux_brier/mean_group_std": 0.0541598135283197,
      "aux_brier/mean_r": 0.9692082576724005,
      "aux_brier/n_active_tok": 219.875,
      "aux_brier/n_groups": 11.40625,
      "aux_brier/n_step_records": 54.96875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5153220762976861,
      "calib/avg_num_step_conf": 6.87109375,
      "calib/ece": 0.4294466403162055,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0015766103814884164,
      "calib/mean_conf": 0.08438735177865614,
      "calib/mu_c": 0.08515384615384615,
      "calib/mu_w": 0.08357723577235773,
      "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": 0.023289723558292966,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2621.0,
      "completions/max_terminated_length": 2621.0,
      "completions/mean_length": 425.43359375,
      "completions/mean_terminated_length": 425.43359375,
      "completions/min_length": 141.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.128,
      "grad_norm": 0.04736745357513428,
      "learning_rate": 2.222222222222222e-06,
      "loss": 0.0703,
      "num_tokens": 24903983.0,
      "reward": 1.1388731002807617,
      "reward_std": 0.22865451872348785,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5554922223091125,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 120
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.1134249522370965e-07,
      "aux_brier/mean_group_std": 0.04610778267001389,
      "aux_brier/mean_r": 0.9691028837137745,
      "aux_brier/n_active_tok": 254.375,
      "aux_brier/n_groups": 15.84375,
      "aux_brier/n_step_records": 63.59375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.41995152442913636,
      "calib/avg_num_step_conf": 8.3359375,
      "calib/ece": 0.37533864541832673,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.010095037632351048,
      "calib/mean_conf": 0.09581673306772909,
      "calib/mu_c": 0.09042735042735044,
      "calib/mu_w": 0.10052238805970148,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0025099601593625495,
      "calib/std_conf": 0.03615524129351776,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3070.0,
      "completions/max_terminated_length": 3070.0,
      "completions/mean_length": 479.9921875,
      "completions/mean_terminated_length": 483.7716369628906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.12906666666666666,
      "grad_norm": 0.0358821265399456,
      "learning_rate": 2.1944444444444445e-06,
      "loss": 0.0478,
      "num_tokens": 25131917.0,
      "reward": 1.096218228340149,
      "reward_std": 0.29095008969306946,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.5958105325698853,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 121
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 3.868681449137323e-06,
      "aux_brier/mean_group_std": 0.05225980236816296,
      "aux_brier/mean_r": 0.9696455510374863,
      "aux_brier/n_active_tok": 226.625,
      "aux_brier/n_groups": 14.78125,
      "aux_brier/n_step_records": 56.65625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5430551079600388,
      "calib/avg_num_step_conf": 7.08203125,
      "calib/ece": 0.49194444444444446,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0014521430873348229,
      "calib/mean_conf": 0.08345238095238097,
      "calib/mu_c": 0.08406896551724138,
      "calib/mu_w": 0.08261682242990656,
      "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": 0.021903144350213826,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2541.0,
      "completions/max_terminated_length": 2541.0,
      "completions/mean_length": 412.9140625,
      "completions/mean_terminated_length": 414.5333557128906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 110.0,
      "epoch": 0.13013333333333332,
      "grad_norm": 0.03914785385131836,
      "learning_rate": 2.166666666666667e-06,
      "loss": 0.0433,
      "num_tokens": 25344967.0,
      "reward": 1.1850626468658447,
      "reward_std": 0.21061810851097107,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.5058754086494446,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 122
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.521688863114772e-07,
      "aux_brier/mean_group_std": 0.0756448575875828,
      "aux_brier/mean_r": 0.9458321603027738,
      "aux_brier/n_active_tok": 249.625,
      "aux_brier/n_groups": 16.1875,
      "aux_brier/n_step_records": 62.40625,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.42751326174149307,
      "calib/avg_num_step_conf": 7.8984375,
      "calib/ece": 0.38771084337349393,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.01668262388407296,
      "calib/mean_conf": 0.09759036144578315,
      "calib/mu_c": 0.08881355932203391,
      "calib/mu_w": 0.10549618320610687,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.005702811244979919,
      "calib/std_conf": 0.05823212850851676,
      "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": 2906.0,
      "completions/max_terminated_length": 2906.0,
      "completions/mean_length": 499.5390625,
      "completions/mean_terminated_length": 505.46246337890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 147.0,
      "epoch": 0.1312,
      "grad_norm": 0.05221722647547722,
      "learning_rate": 2.138888888888889e-06,
      "loss": 0.0486,
      "num_tokens": 25578137.0,
      "reward": 1.0872745513916016,
      "reward_std": 0.2682573199272156,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.5756609439849854,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 123
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -5.05113540327784e-07,
      "aux_brier/mean_group_std": 0.053274477799555124,
      "aux_brier/mean_r": 0.9693263882317068,
      "aux_brier/n_active_tok": 228.75,
      "aux_brier/n_groups": 13.0,
      "aux_brier/n_step_records": 57.1875,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.4933316714445968,
      "calib/avg_num_step_conf": 7.1484375,
      "calib/ece": 0.47050980392156866,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0006693256886451671,
      "calib/mean_conf": 0.0863529411764706,
      "calib/mu_c": 0.08605633802816902,
      "calib/mu_w": 0.08672566371681419,
      "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": 0.025660546911943815,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2265.0,
      "completions/max_terminated_length": 2265.0,
      "completions/mean_length": 442.71875,
      "completions/mean_terminated_length": 442.71875,
      "completions/min_length": 150.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.13226666666666667,
      "grad_norm": 0.010232754051685333,
      "learning_rate": 2.1111111111111114e-06,
      "loss": 0.0051,
      "num_tokens": 25798289.0,
      "reward": 1.1849322319030762,
      "reward_std": 0.20356249809265137,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.5287914276123047,
      "rewards/format_reward_step_strict": 0.99609375,
      "step": 124
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.0275335669422967e-07,
      "aux_brier/mean_group_std": 0.06576867125092614,
      "aux_brier/mean_r": 0.9578147937750323,
      "aux_brier/n_active_tok": 238.875,
      "aux_brier/n_groups": 14.78125,
      "aux_brier/n_step_records": 59.71875,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4444934475806452,
      "calib/avg_num_step_conf": 7.46484375,
      "calib/ece": 0.40813492063492063,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.010098286290322597,
      "calib/mean_conf": 0.089484126984127,
      "calib/mu_c": 0.08435483870967743,
      "calib/mu_w": 0.09445312500000003,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0027777777777777775,
      "calib/std_conf": 0.03736525721121247,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3033.0,
      "completions/max_terminated_length": 3033.0,
      "completions/mean_length": 492.0390625,
      "completions/mean_terminated_length": 492.0390625,
      "completions/min_length": 134.0,
      "completions/min_terminated_length": 134.0,
      "epoch": 0.13333333333333333,
      "grad_norm": 0.007184287067502737,
      "learning_rate": 2.0833333333333334e-06,
      "loss": 0.058,
      "num_tokens": 26029059.0,
      "reward": 1.119678020477295,
      "reward_std": 0.23738686740398407,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.5724620819091797,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 125
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.0497718808409893e-06,
      "aux_brier/mean_group_std": 0.06422724326178852,
      "aux_brier/mean_r": 0.9581111974496881,
      "aux_brier/n_active_tok": 250.625,
      "aux_brier/n_groups": 15.8125,
      "aux_brier/n_step_records": 62.65625,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.6346127763461278,
      "calib/avg_num_step_conf": 8.16796875,
      "calib/ece": 0.5089878542510121,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.011266784212667844,
      "calib/mean_conf": 0.08210526315789474,
      "calib/mu_c": 0.08671232876712329,
      "calib/mu_w": 0.07544554455445544,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.026928099132810256,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2866.0,
      "completions/max_terminated_length": 2866.0,
      "completions/mean_length": 489.84765625,
      "completions/mean_terminated_length": 495.6561584472656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 126.0,
      "epoch": 0.1344,
      "grad_norm": 0.05721435695886612,
      "learning_rate": 2.0555555555555555e-06,
      "loss": 0.0539,
      "num_tokens": 26259924.0,
      "reward": 1.1713693141937256,
      "reward_std": 0.24430416524410248,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.4823523163795471,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 126
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.2410819398567252e-06,
      "aux_brier/mean_group_std": 0.03815124177472212,
      "aux_brier/mean_r": 0.9759511842763717,
      "aux_brier/n_active_tok": 254.125,
      "aux_brier/n_groups": 15.25,
      "aux_brier/n_step_records": 63.53125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5595032328276038,
      "calib/avg_num_step_conf": 8.07421875,
      "calib/ece": 0.42336000000000007,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.006253120798924525,
      "calib/mean_conf": 0.08464,
      "calib/mu_c": 0.08771653543307088,
      "calib/mu_w": 0.08146341463414636,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.029217638508270995,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2709.0,
      "completions/max_terminated_length": 2709.0,
      "completions/mean_length": 472.50390625,
      "completions/mean_terminated_length": 474.3569030761719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.13546666666666668,
      "grad_norm": 0.017782477661967278,
      "learning_rate": 2.027777777777778e-06,
      "loss": 0.0695,
      "num_tokens": 26484557.0,
      "reward": 1.1242926120758057,
      "reward_std": 0.20912383496761322,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.5596703290939331,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 127
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.740578158271333e-07,
      "aux_brier/mean_group_std": 0.05217880205747242,
      "aux_brier/mean_r": 0.9636292925598783,
      "aux_brier/n_active_tok": 226.625,
      "aux_brier/n_groups": 14.4375,
      "aux_brier/n_step_records": 56.65625,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5945245901639344,
      "calib/avg_num_step_conf": 7.14453125,
      "calib/ece": 0.411251012145749,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00912708196721311,
      "calib/mean_conf": 0.08267611336032388,
      "calib/mu_c": 0.08729508196721311,
      "calib/mu_w": 0.078168,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.028494767547860126,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3055.0,
      "completions/max_terminated_length": 3055.0,
      "completions/mean_length": 481.00390625,
      "completions/mean_terminated_length": 486.70751953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 128.0,
      "epoch": 0.13653333333333334,
      "grad_norm": 0.03822065517306328,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0539,
      "num_tokens": 26714358.0,
      "reward": 1.101963996887207,
      "reward_std": 0.2587791085243225,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.5641059279441833,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 128
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -4.896738784038179e-07,
      "aux_brier/mean_group_std": 0.06356348446878757,
      "aux_brier/mean_r": 0.9580698250196816,
      "aux_brier/n_active_tok": 229.125,
      "aux_brier/n_groups": 12.875,
      "aux_brier/n_step_records": 57.28125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.561960933536276,
      "calib/avg_num_step_conf": 7.18359375,
      "calib/ece": 0.4877165354330709,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004614408929477398,
      "calib/mean_conf": 0.08913385826771655,
      "calib/mu_c": 0.09109589041095889,
      "calib/mu_w": 0.08648148148148149,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0010236220472440946,
      "calib/std_conf": 0.029551066757920824,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2464.0,
      "completions/max_terminated_length": 2464.0,
      "completions/mean_length": 433.4140625,
      "completions/mean_terminated_length": 435.1137390136719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 138.0,
      "epoch": 0.1376,
      "grad_norm": 0.013987638987600803,
      "learning_rate": 1.9722222222222224e-06,
      "loss": -0.0069,
      "num_tokens": 26927696.0,
      "reward": 1.1956642866134644,
      "reward_std": 0.1896999180316925,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.5170320272445679,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 129
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.888920857895382e-06,
      "aux_brier/mean_group_std": 0.05988606144609164,
      "aux_brier/mean_r": 0.9603871145268268,
      "aux_brier/n_active_tok": 230.375,
      "aux_brier/n_groups": 12.375,
      "aux_brier/n_step_records": 57.59375,
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5627044025157233,
      "calib/avg_num_step_conf": 7.25,
      "calib/ece": 0.4921484375,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005339622641509431,
      "calib/mean_conf": 0.09378906250000002,
      "calib/mu_c": 0.09599999999999999,
      "calib/mu_w": 0.09066037735849056,
      "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": 0.03325003579503477,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1662.0,
      "completions/max_terminated_length": 1662.0,
      "completions/mean_length": 420.58984375,
      "completions/mean_terminated_length": 422.2392272949219,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.13866666666666666,
      "grad_norm": 0.04971790313720703,
      "learning_rate": 1.944444444444445e-06,
      "loss": 0.0109,
      "num_tokens": 27140655.0,
      "reward": 1.2151026725769043,
      "reward_std": 0.18918028473854065,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.5166605710983276,
      "rewards/format_reward_step_strict": 1.0,
      "step": 130
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.1584866417280182e-06,
      "aux_brier/mean_group_std": 0.051293487482689956,
      "aux_brier/mean_r": 0.9696329716740563,
      "aux_brier/n_active_tok": 233.875,
      "aux_brier/n_groups": 13.59375,
      "aux_brier/n_step_records": 58.46875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6290830381739473,
      "calib/avg_num_step_conf": 7.4296875,
      "calib/ece": 0.30363636363636365,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.010230880230880243,
      "calib/mean_conf": 0.08932806324110673,
      "calib/mu_c": 0.09555555555555556,
      "calib/mu_w": 0.08532467532467532,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0008300395256916996,
      "calib/std_conf": 0.02596636397927713,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2996.0,
      "completions/max_terminated_length": 2996.0,
      "completions/mean_length": 422.24609375,
      "completions/mean_terminated_length": 425.57086181640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.13973333333333332,
      "grad_norm": 0.024130966514348984,
      "learning_rate": 1.916666666666667e-06,
      "loss": -0.0385,
      "num_tokens": 27354958.0,
      "reward": 1.0475884675979614,
      "reward_std": 0.1937212347984314,
      "rewards/accuracy_reward_step": 0.38671875,
      "rewards/final_brier_reward_step": 0.6669164299964905,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 131
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.1103363183061496e-06,
      "aux_brier/mean_group_std": 0.09580266574378875,
      "aux_brier/mean_r": 0.933538275029279,
      "aux_brier/n_active_tok": 243.875,
      "aux_brier/n_groups": 13.8125,
      "aux_brier/n_step_records": 60.96875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.47300931391840484,
      "calib/avg_num_step_conf": 7.97265625,
      "calib/ece": 0.5131225296442689,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.004783549783549779,
      "calib/mean_conf": 0.09557312252964427,
      "calib/mu_c": 0.0937012987012987,
      "calib/mu_w": 0.09848484848484848,
      "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": 0.030414910611129702,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1342.0,
      "completions/max_terminated_length": 1342.0,
      "completions/mean_length": 446.3828125,
      "completions/mean_terminated_length": 449.89764404296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 128.0,
      "epoch": 0.1408,
      "grad_norm": 0.013880199752748013,
      "learning_rate": 1.888888888888889e-06,
      "loss": -0.007,
      "num_tokens": 27574824.0,
      "reward": 1.216127872467041,
      "reward_std": 0.22834108769893646,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.4895116984844208,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 132
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.060480568832901e-06,
      "aux_brier/mean_group_std": 0.04970631518586872,
      "aux_brier/mean_r": 0.9654551346843375,
      "aux_brier/n_active_tok": 260.75,
      "aux_brier/n_groups": 14.0625,
      "aux_brier/n_step_records": 65.1875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4812318265926513,
      "calib/avg_num_step_conf": 8.1484375,
      "calib/ece": 0.27964426877470355,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0029130319851969205,
      "calib/mean_conf": 0.10375494071146246,
      "calib/mu_c": 0.10195876288659794,
      "calib/mu_w": 0.10487179487179486,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.03518066338743322,
      "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": 2876.0,
      "completions/max_terminated_length": 2876.0,
      "completions/mean_length": 529.78125,
      "completions/mean_terminated_length": 529.78125,
      "completions/min_length": 167.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.14186666666666667,
      "grad_norm": 0.043244197964668274,
      "learning_rate": 1.8611111111111113e-06,
      "loss": 0.0394,
      "num_tokens": 27816792.0,
      "reward": 1.038818120956421,
      "reward_std": 0.2598777711391449,
      "rewards/accuracy_reward_step": 0.37890625,
      "rewards/final_brier_reward_step": 0.6708973050117493,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 133
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -9.444852798190517e-07,
      "aux_brier/mean_group_std": 0.05817175404218855,
      "aux_brier/mean_r": 0.9565321587049483,
      "aux_brier/n_active_tok": 263.25,
      "aux_brier/n_groups": 16.46875,
      "aux_brier/n_step_records": 65.8125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.628446348733234,
      "calib/avg_num_step_conf": 8.2265625,
      "calib/ece": 0.41905511811023616,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009413810233482367,
      "calib/mean_conf": 0.10062992125984253,
      "calib/mu_c": 0.10515151515151515,
      "calib/mu_w": 0.09573770491803278,
      "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": 0.037083604873144424,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2462.0,
      "completions/max_terminated_length": 2462.0,
      "completions/mean_length": 544.76953125,
      "completions/mean_terminated_length": 544.76953125,
      "completions/min_length": 170.0,
      "completions/min_terminated_length": 170.0,
      "epoch": 0.14293333333333333,
      "grad_norm": 0.010487113147974014,
      "learning_rate": 1.8333333333333333e-06,
      "loss": 0.0301,
      "num_tokens": 28065205.0,
      "reward": 1.1551158428192139,
      "reward_std": 0.17233869433403015,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.5735882520675659,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 134
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.0978945453341566e-07,
      "aux_brier/mean_group_std": 0.032730762372773364,
      "aux_brier/mean_r": 0.9779772052692718,
      "aux_brier/n_active_tok": 258.25,
      "aux_brier/n_groups": 16.96875,
      "aux_brier/n_step_records": 64.5625,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5892671517671517,
      "calib/avg_num_step_conf": 8.2578125,
      "calib/ece": 0.49424603174603177,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005738045738045733,
      "calib/mean_conf": 0.09702380952380954,
      "calib/mu_c": 0.09939189189189189,
      "calib/mu_w": 0.09365384615384616,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.001984126984126984,
      "calib/std_conf": 0.04208211253028225,
      "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": 2262.0,
      "completions/max_terminated_length": 2262.0,
      "completions/mean_length": 510.83984375,
      "completions/mean_terminated_length": 514.8621826171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.144,
      "grad_norm": 0.04326924681663513,
      "learning_rate": 1.8055555555555557e-06,
      "loss": 0.0158,
      "num_tokens": 28301860.0,
      "reward": 1.1951674222946167,
      "reward_std": 0.25529035925865173,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.5072324275970459,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 135
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 4.7476447551098033e-07,
      "aux_brier/mean_group_std": 0.052697667470998966,
      "aux_brier/mean_r": 0.9702369810499175,
      "aux_brier/n_active_tok": 268.375,
      "aux_brier/n_groups": 16.6875,
      "aux_brier/n_step_records": 67.09375,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.576388450116889,
      "calib/avg_num_step_conf": 8.46484375,
      "calib/ece": 0.37202380952380953,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.003968253968253968,
      "calib/gap": 0.013157894736842146,
      "calib/mean_conf": 0.10019841269841272,
      "calib/mu_c": 0.10714285714285718,
      "calib/mu_w": 0.09398496240601503,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.0,
      "calib/std_conf": 0.062319265007951326,
      "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": 2290.0,
      "completions/max_terminated_length": 2290.0,
      "completions/mean_length": 485.88671875,
      "completions/mean_terminated_length": 487.79217529296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 135.0,
      "epoch": 0.14506666666666668,
      "grad_norm": 0.020810028538107872,
      "learning_rate": 1.777777777777778e-06,
      "loss": 0.0279,
      "num_tokens": 28534735.0,
      "reward": 1.1033775806427002,
      "reward_std": 0.230840265750885,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.6010105609893799,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 136
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.4710560704497833e-06,
      "aux_brier/mean_group_std": 0.05615136213528867,
      "aux_brier/mean_r": 0.9564882836260133,
      "aux_brier/n_active_tok": 291.125,
      "aux_brier/n_groups": 19.15625,
      "aux_brier/n_step_records": 72.78125,
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.5463025323275862,
      "calib/avg_num_step_conf": 9.296875,
      "calib/ece": 0.37434426229508194,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.003116918103448274,
      "calib/mean_conf": 0.10327868852459017,
      "calib/mu_c": 0.1049137931034483,
      "calib/mu_w": 0.10179687500000002,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0011065573770491805,
      "calib/std_conf": 0.029391161349853923,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2996.0,
      "completions/max_terminated_length": 2996.0,
      "completions/mean_length": 536.6875,
      "completions/mean_terminated_length": 543.0513916015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 181.0,
      "epoch": 0.14613333333333334,
      "grad_norm": 0.04827215522527695,
      "learning_rate": 1.75e-06,
      "loss": 0.1091,
      "num_tokens": 28779111.0,
      "reward": 1.0727916955947876,
      "reward_std": 0.3305455446243286,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.5802293419837952,
      "rewards/format_reward_step_strict": 0.94921875,
      "step": 137
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.542300720584592e-07,
      "aux_brier/mean_group_std": 0.0390463759611963,
      "aux_brier/mean_r": 0.9705610088967543,
      "aux_brier/n_active_tok": 261.25,
      "aux_brier/n_groups": 15.625,
      "aux_brier/n_step_records": 65.3125,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5334576138147566,
      "calib/avg_num_step_conf": 8.1640625,
      "calib/ece": 0.5104330708661418,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.002059131344845644,
      "calib/mean_conf": 0.10720472440944882,
      "calib/mu_c": 0.10641025641025642,
      "calib/mu_w": 0.10846938775510206,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.001732283464566929,
      "calib/std_conf": 0.041588315777102806,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2517.0,
      "completions/max_terminated_length": 2517.0,
      "completions/mean_length": 484.7890625,
      "completions/mean_terminated_length": 484.7890625,
      "completions/min_length": 64.0,
      "completions/min_terminated_length": 64.0,
      "epoch": 0.1472,
      "grad_norm": 0.01851765625178814,
      "learning_rate": 1.7222222222222224e-06,
      "loss": 0.0265,
      "num_tokens": 29007553.0,
      "reward": 1.230314016342163,
      "reward_std": 0.25490623712539673,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.49938085675239563,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 138
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.8302141957526885e-07,
      "aux_brier/mean_group_std": 0.05321893990740268,
      "aux_brier/mean_r": 0.9606402633369969,
      "aux_brier/n_active_tok": 243.5,
      "aux_brier/n_groups": 13.625,
      "aux_brier/n_step_records": 60.875,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5765964240102172,
      "calib/avg_num_step_conf": 7.72265625,
      "calib/ece": 0.4759683794466404,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005374201787994873,
      "calib/mean_conf": 0.09715415019762846,
      "calib/mu_c": 0.09944827586206895,
      "calib/mu_w": 0.09407407407407407,
      "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": 0.02985809575283303,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2362.0,
      "completions/max_terminated_length": 2362.0,
      "completions/mean_length": 465.15625,
      "completions/mean_terminated_length": 466.98040771484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.14826666666666666,
      "grad_norm": 0.01863935962319374,
      "learning_rate": 1.6944444444444446e-06,
      "loss": 0.0281,
      "num_tokens": 29229729.0,
      "reward": 1.1916272640228271,
      "reward_std": 0.23998242616653442,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.5243219137191772,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 139
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -4.0164620990212896e-07,
      "aux_brier/mean_group_std": 0.046515244450848044,
      "aux_brier/mean_r": 0.9689238602553203,
      "aux_brier/n_active_tok": 242.875,
      "aux_brier/n_groups": 12.96875,
      "aux_brier/n_step_records": 60.71875,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5593640305595705,
      "calib/avg_num_step_conf": 7.625,
      "calib/ece": 0.5574409448818898,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005654208823731868,
      "calib/mean_conf": 0.10003937007874016,
      "calib/mu_c": 0.10197604790419162,
      "calib/mu_w": 0.09632183908045976,
      "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": 0.03087309599467006,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2788.0,
      "completions/max_terminated_length": 2788.0,
      "completions/mean_length": 482.44921875,
      "completions/mean_terminated_length": 484.3412170410156,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 138.0,
      "epoch": 0.14933333333333335,
      "grad_norm": 0.06730857491493225,
      "learning_rate": 1.6666666666666667e-06,
      "loss": 0.0164,
      "num_tokens": 29458252.0,
      "reward": 1.2639412879943848,
      "reward_std": 0.2333708554506302,
      "rewards/accuracy_reward_step": 0.65234375,
      "rewards/final_brier_reward_step": 0.4620152711868286,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 140
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.325099744242113e-07,
      "aux_brier/mean_group_std": 0.07243225738584982,
      "aux_brier/mean_r": 0.9495889869856867,
      "aux_brier/n_active_tok": 264.5,
      "aux_brier/n_groups": 15.0,
      "aux_brier/n_step_records": 66.125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6613786952089704,
      "calib/avg_num_step_conf": 8.265625,
      "calib/ece": 0.4766007905138339,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.003952569169960474,
      "calib/gap": 0.0061525229357797945,
      "calib/mean_conf": 0.10047430830039526,
      "calib/mu_c": 0.103125,
      "calib/mu_w": 0.0969724770642202,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003952569169960474,
      "calib/std_conf": 0.06619957497815487,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 557.9453125,
      "completions/mean_terminated_length": 560.1333618164062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.1504,
      "grad_norm": 0.012635590508580208,
      "learning_rate": 1.638888888888889e-06,
      "loss": 0.0347,
      "num_tokens": 29708182.0,
      "reward": 1.1885128021240234,
      "reward_std": 0.2409987598657608,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.5274890661239624,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 141
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.078029411993313e-07,
      "aux_brier/mean_group_std": 0.06055486257432703,
      "aux_brier/mean_r": 0.958171473881109,
      "aux_brier/n_active_tok": 267.75,
      "aux_brier/n_groups": 14.625,
      "aux_brier/n_step_records": 66.9375,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.61821910147269,
      "calib/avg_num_step_conf": 8.5703125,
      "calib/ece": 0.42677165354330704,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.013240539364941267,
      "calib/mean_conf": 0.09685039370078741,
      "calib/mu_c": 0.1031578947368421,
      "calib/mu_w": 0.08991735537190083,
      "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": 0.03550980318013948,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2141.0,
      "completions/max_terminated_length": 2141.0,
      "completions/mean_length": 510.6875,
      "completions/mean_terminated_length": 512.6902465820312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.15146666666666667,
      "grad_norm": 0.030646586790680885,
      "learning_rate": 1.6111111111111113e-06,
      "loss": 0.0078,
      "num_tokens": 29944078.0,
      "reward": 1.1579464673995972,
      "reward_std": 0.1530137062072754,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.5692859292030334,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 142
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -4.110612904228983e-07,
      "aux_brier/mean_group_std": 0.05591120441773031,
      "aux_brier/mean_r": 0.9650633850167166,
      "aux_brier/n_active_tok": 269.125,
      "aux_brier/n_groups": 16.46875,
      "aux_brier/n_step_records": 67.28125,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5596239989140763,
      "calib/avg_num_step_conf": 8.48046875,
      "calib/ece": 0.47085714285714286,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004455002036106978,
      "calib/mean_conf": 0.09648979591836734,
      "calib/mu_c": 0.09841726618705036,
      "calib/mu_w": 0.09396226415094339,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.029800782381853592,
      "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": 2547.0,
      "completions/max_terminated_length": 2547.0,
      "completions/mean_length": 536.37890625,
      "completions/mean_terminated_length": 540.6023559570312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 206.0,
      "epoch": 0.15253333333333333,
      "grad_norm": 0.09321467578411102,
      "learning_rate": 1.5833333333333333e-06,
      "loss": 0.0439,
      "num_tokens": 30188727.0,
      "reward": 1.1492786407470703,
      "reward_std": 0.23780958354473114,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.5111774206161499,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 143
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.2350600124833022e-06,
      "aux_brier/mean_group_std": 0.027071194279506893,
      "aux_brier/mean_r": 0.9816140675573792,
      "aux_brier/n_active_tok": 273.75,
      "aux_brier/n_groups": 14.78125,
      "aux_brier/n_step_records": 68.4375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6468660111044218,
      "calib/avg_num_step_conf": 8.890625,
      "calib/ece": 0.50632,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0175650545186969,
      "calib/mean_conf": 0.09768,
      "calib/mu_c": 0.10463576158940398,
      "calib/mu_w": 0.08707070707070708,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.032781360557487545,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2009.0,
      "completions/max_terminated_length": 2009.0,
      "completions/mean_length": 510.34765625,
      "completions/mean_terminated_length": 516.3992309570312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.1536,
      "grad_norm": 0.15854761004447937,
      "learning_rate": 1.5555555555555558e-06,
      "loss": -0.0021,
      "num_tokens": 30423504.0,
      "reward": 1.206978440284729,
      "reward_std": 0.23236022889614105,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.4997890591621399,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 144
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -8.570846455890901e-07,
      "aux_brier/mean_group_std": 0.07063431696107034,
      "aux_brier/mean_r": 0.9479332649163424,
      "aux_brier/n_active_tok": 289.875,
      "aux_brier/n_groups": 18.125,
      "aux_brier/n_step_records": 72.46875,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6206608569353668,
      "calib/avg_num_step_conf": 9.05859375,
      "calib/ece": 0.5703187250996016,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.014122004357298504,
      "calib/mean_conf": 0.10697211155378486,
      "calib/mu_c": 0.11152941176470589,
      "calib/mu_w": 0.09740740740740739,
      "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": 0.04086216286136444,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2327.0,
      "completions/max_terminated_length": 2327.0,
      "completions/mean_length": 530.8046875,
      "completions/mean_terminated_length": 530.8046875,
      "completions/min_length": 163.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.15466666666666667,
      "grad_norm": 0.04638217017054558,
      "learning_rate": 1.527777777777778e-06,
      "loss": 0.054,
      "num_tokens": 30662094.0,
      "reward": 1.2672154903411865,
      "reward_std": 0.2803875803947449,
      "rewards/accuracy_reward_step": 0.6640625,
      "rewards/final_brier_reward_step": 0.45167461037635803,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 145
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.7337525574090407e-06,
      "aux_brier/mean_group_std": 0.02803442662753144,
      "aux_brier/mean_r": 0.9840496662095859,
      "aux_brier/n_active_tok": 277.875,
      "aux_brier/n_groups": 15.40625,
      "aux_brier/n_step_records": 69.46875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5574900602795947,
      "calib/avg_num_step_conf": 8.91015625,
      "calib/ece": 0.3627091633466136,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": 0.0005027574708221377,
      "calib/mean_conf": 0.0952988047808765,
      "calib/mu_c": 0.09557522123893808,
      "calib/mu_w": 0.09507246376811594,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003904382470119522,
      "calib/std_conf": 0.06506057825744377,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1659.0,
      "completions/max_terminated_length": 1659.0,
      "completions/mean_length": 530.1796875,
      "completions/mean_terminated_length": 540.7410278320312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 188.0,
      "epoch": 0.15573333333333333,
      "grad_norm": 0.11035989969968796,
      "learning_rate": 1.5e-06,
      "loss": -0.0713,
      "num_tokens": 30905036.0,
      "reward": 1.0813231468200684,
      "reward_std": 0.2637237310409546,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.6065425872802734,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 146
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.50355611464137e-07,
      "aux_brier/mean_group_std": 0.05485977103305224,
      "aux_brier/mean_r": 0.9593606115870782,
      "aux_brier/n_active_tok": 285.75,
      "aux_brier/n_groups": 15.0625,
      "aux_brier/n_step_records": 71.4375,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6319531839919457,
      "calib/avg_num_step_conf": 9.0859375,
      "calib/ece": 0.3577470355731225,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.012301157815252997,
      "calib/mean_conf": 0.10549407114624505,
      "calib/mu_c": 0.11215517241379314,
      "calib/mu_w": 0.09985401459854014,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0023715415019762843,
      "calib/std_conf": 0.042298732500783094,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2304.0,
      "completions/max_terminated_length": 2304.0,
      "completions/mean_length": 534.6953125,
      "completions/mean_terminated_length": 538.905517578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 183.0,
      "epoch": 0.1568,
      "grad_norm": 0.1080685704946518,
      "learning_rate": 1.4722222222222225e-06,
      "loss": 0.0109,
      "num_tokens": 31145598.0,
      "reward": 1.1032731533050537,
      "reward_std": 0.19219431281089783,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6240301132202148,
      "rewards/format_reward_step_strict": 0.98828125,
      "step": 147
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -5.543715002609417e-07,
      "aux_brier/mean_group_std": 0.05581918718485776,
      "aux_brier/mean_r": 0.9629090772270584,
      "aux_brier/n_active_tok": 267.125,
      "aux_brier/n_groups": 14.59375,
      "aux_brier/n_step_records": 66.78125,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.6817950581395349,
      "calib/avg_num_step_conf": 8.7421875,
      "calib/ece": 0.5501626016260164,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0040650406504065045,
      "calib/gap": 0.007065406976744196,
      "calib/mean_conf": 0.10796747967479675,
      "calib/mu_c": 0.1104375,
      "calib/mu_w": 0.1033720930232558,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003861788617886179,
      "calib/std_conf": 0.06123338466647912,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2803.0,
      "completions/max_terminated_length": 2803.0,
      "completions/mean_length": 537.64453125,
      "completions/mean_terminated_length": 544.019775390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.15786666666666666,
      "grad_norm": 0.07648499310016632,
      "learning_rate": 1.4444444444444445e-06,
      "loss": 0.0359,
      "num_tokens": 31388347.0,
      "reward": 1.2202636003494263,
      "reward_std": 0.24338200688362122,
      "rewards/accuracy_reward_step": 0.625,
      "rewards/final_brier_reward_step": 0.45917969942092896,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 148
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.798103858059342e-07,
      "aux_brier/mean_group_std": 0.05612567180075495,
      "aux_brier/mean_r": 0.9629851097100218,
      "aux_brier/n_active_tok": 284.0,
      "aux_brier/n_groups": 15.40625,
      "aux_brier/n_step_records": 71.0,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.523456308152775,
      "calib/avg_num_step_conf": 8.91796875,
      "calib/ece": 0.45616,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.007068774806428521,
      "calib/mean_conf": 0.10784000000000002,
      "calib/mu_c": 0.11092198581560284,
      "calib/mu_w": 0.10385321100917431,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.041077176144423556,
      "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": 2486.0,
      "completions/max_terminated_length": 2486.0,
      "completions/mean_length": 580.59375,
      "completions/mean_terminated_length": 582.87060546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 115.0,
      "epoch": 0.15893333333333334,
      "grad_norm": 0.01760271191596985,
      "learning_rate": 1.4166666666666667e-06,
      "loss": 0.0146,
      "num_tokens": 31641435.0,
      "reward": 1.172803521156311,
      "reward_std": 0.2619955241680145,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.5349640846252441,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 149
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -9.429146688444412e-08,
      "aux_brier/mean_group_std": 0.052168910145743194,
      "aux_brier/mean_r": 0.962446100782362,
      "aux_brier/n_active_tok": 291.0,
      "aux_brier/n_groups": 18.375,
      "aux_brier/n_step_records": 72.75,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5845388188453882,
      "calib/avg_num_step_conf": 9.26171875,
      "calib/ece": 0.44153846153846155,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.004048582995951417,
      "calib/gap": 0.017595885865958838,
      "calib/mean_conf": 0.11603238866396763,
      "calib/mu_c": 0.12386861313868612,
      "calib/mu_w": 0.10627272727272728,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0014574898785425102,
      "calib/std_conf": 0.07325049200179348,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2841.0,
      "completions/max_terminated_length": 2841.0,
      "completions/mean_length": 544.83203125,
      "completions/mean_terminated_length": 549.1220703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.16,
      "grad_norm": 0.15550203621387482,
      "learning_rate": 1.3888888888888892e-06,
      "loss": 0.0602,
      "num_tokens": 31885872.0,
      "reward": 1.1478254795074463,
      "reward_std": 0.28443899750709534,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.5366144776344299,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 150
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.6749030640770286e-07,
      "aux_brier/mean_group_std": 0.05182812839019988,
      "aux_brier/mean_r": 0.9620563896566551,
      "aux_brier/n_active_tok": 292.25,
      "aux_brier/n_groups": 16.5,
      "aux_brier/n_step_records": 73.0625,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.6068881257429666,
      "calib/avg_num_step_conf": 9.55078125,
      "calib/ece": 0.34688259109311737,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.004048582995951417,
      "calib/gap": 0.011942940166424554,
      "calib/mean_conf": 0.1182995951417004,
      "calib/mu_c": 0.12477876106194694,
      "calib/mu_w": 0.11283582089552238,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.003846153846153846,
      "calib/std_conf": 0.07323872092514362,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 2542.0,
      "completions/max_terminated_length": 2542.0,
      "completions/mean_length": 576.828125,
      "completions/mean_terminated_length": 585.984130859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.16106666666666666,
      "grad_norm": 0.03265540674328804,
      "learning_rate": 1.3611111111111112e-06,
      "loss": 0.0357,
      "num_tokens": 32140564.0,
      "reward": 1.0755085945129395,
      "reward_std": 0.2857625484466553,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.6145347356796265,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 151
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.598158972411113e-07,
      "aux_brier/mean_group_std": 0.021333223741434024,
      "aux_brier/mean_r": 0.9815279845409678,
      "aux_brier/n_active_tok": 320.0,
      "aux_brier/n_groups": 21.0625,
      "aux_brier/n_step_records": 80.0,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5063945578231293,
      "calib/avg_num_step_conf": 10.24609375,
      "calib/ece": 0.4658367346938776,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.004081632653061225,
      "calib/gap": -0.006619047619047622,
      "calib/mean_conf": 0.11612244897959186,
      "calib/mu_c": 0.11328571428571428,
      "calib/mu_w": 0.1199047619047619,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00526530612244898,
      "calib/std_conf": 0.06644980694465628,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2940.0,
      "completions/max_terminated_length": 2940.0,
      "completions/mean_length": 596.078125,
      "completions/mean_terminated_length": 605.5397338867188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 218.0,
      "epoch": 0.16213333333333332,
      "grad_norm": 0.18766255676746368,
      "learning_rate": 1.3333333333333334e-06,
      "loss": 0.0151,
      "num_tokens": 32398552.0,
      "reward": 1.1546235084533691,
      "reward_std": 0.2653184235095978,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.5169316530227661,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 152
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.1766621638441563e-07,
      "aux_brier/mean_group_std": 0.06026336398708991,
      "aux_brier/mean_r": 0.9515601117150679,
      "aux_brier/n_active_tok": 303.75,
      "aux_brier/n_groups": 19.0625,
      "aux_brier/n_step_records": 75.9375,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5809523809523809,
      "calib/avg_num_step_conf": 9.4921875,
      "calib/ece": 0.44047808764940244,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.00796812749003984,
      "calib/gap": 0.010137065637065637,
      "calib/mean_conf": 0.12430278884462152,
      "calib/mu_c": 0.12878571428571428,
      "calib/mu_w": 0.11864864864864864,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0035059760956175296,
      "calib/std_conf": 0.08171842112270698,
      "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": 2612.0,
      "completions/max_terminated_length": 2612.0,
      "completions/mean_length": 582.7734375,
      "completions/mean_terminated_length": 582.7734375,
      "completions/min_length": 239.0,
      "completions/min_terminated_length": 239.0,
      "epoch": 0.1632,
      "grad_norm": 0.0220626313239336,
      "learning_rate": 1.3055555555555556e-06,
      "loss": 0.0362,
      "num_tokens": 32655062.0,
      "reward": 1.166546106338501,
      "reward_std": 0.21586893498897552,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.5411847829818726,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 153
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.138314249393151e-07,
      "aux_brier/mean_group_std": 0.05190131323072046,
      "aux_brier/mean_r": 0.9613522500884691,
      "aux_brier/n_active_tok": 289.625,
      "aux_brier/n_groups": 16.09375,
      "aux_brier/n_step_records": 72.40625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.614176245210728,
      "calib/avg_num_step_conf": 9.17578125,
      "calib/ece": 0.338804780876494,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.016377394636015313,
      "calib/mean_conf": 0.12334661354581673,
      "calib/mu_c": 0.1321551724137931,
      "calib/mu_w": 0.11577777777777779,
      "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": 0.044165067257084134,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 3031.0,
      "completions/max_terminated_length": 3031.0,
      "completions/mean_length": 534.1328125,
      "completions/mean_terminated_length": 536.2274780273438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 172.0,
      "epoch": 0.16426666666666667,
      "grad_norm": 0.04481395334005356,
      "learning_rate": 1.2777777777777779e-06,
      "loss": 0.05,
      "num_tokens": 32896240.0,
      "reward": 1.0980093479156494,
      "reward_std": 0.2588661313056946,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6264125108718872,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 154
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.8488905962076316e-07,
      "aux_brier/mean_group_std": 0.055795254315706595,
      "aux_brier/mean_r": 0.9584929775784383,
      "aux_brier/n_active_tok": 272.0,
      "aux_brier/n_groups": 15.875,
      "aux_brier/n_step_records": 68.0,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5912463882322039,
      "calib/avg_num_step_conf": 8.78515625,
      "calib/ece": 0.3091164658634538,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.010975177304964542,
      "calib/mean_conf": 0.12461847389558234,
      "calib/mu_c": 0.13083333333333333,
      "calib/mu_w": 0.11985815602836879,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04246372205593495,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2867.0,
      "completions/max_terminated_length": 2867.0,
      "completions/mean_length": 517.8515625,
      "completions/mean_terminated_length": 521.9291381835938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 185.0,
      "epoch": 0.16533333333333333,
      "grad_norm": 0.021404221653938293,
      "learning_rate": 1.25e-06,
      "loss": 0.0495,
      "num_tokens": 33136026.0,
      "reward": 1.0664126873016357,
      "reward_std": 0.29070472717285156,
      "rewards/accuracy_reward_step": 0.421875,
      "rewards/final_brier_reward_step": 0.6406508088111877,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 155
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 7.492172946974018e-08,
      "aux_brier/mean_group_std": 0.06266845824162358,
      "aux_brier/mean_r": 0.9578044682141771,
      "aux_brier/n_active_tok": 294.875,
      "aux_brier/n_groups": 16.59375,
      "aux_brier/n_step_records": 73.71875,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5657978995901639,
      "calib/avg_num_step_conf": 9.54296875,
      "calib/ece": 0.38716,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.008,
      "calib/gap": -0.0013780737704918444,
      "calib/mean_conf": 0.14036,
      "calib/mu_c": 0.1396875,
      "calib/mu_w": 0.14106557377049184,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0077599999999999995,
      "calib/std_conf": 0.08532450058453316,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2693.0,
      "completions/max_terminated_length": 2693.0,
      "completions/mean_length": 563.61328125,
      "completions/mean_terminated_length": 570.2964477539062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 148.0,
      "epoch": 0.1664,
      "grad_norm": 0.2925337851047516,
      "learning_rate": 1.2222222222222223e-06,
      "loss": -0.0149,
      "num_tokens": 33385071.0,
      "reward": 1.1328585147857666,
      "reward_std": 0.26367974281311035,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.5861214995384216,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 156
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.6625511922718772e-07,
      "aux_brier/mean_group_std": 0.06699555604094448,
      "aux_brier/mean_r": 0.9519206889973116,
      "aux_brier/n_active_tok": 300.875,
      "aux_brier/n_groups": 17.1875,
      "aux_brier/n_step_records": 75.21875,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.6571286472148542,
      "calib/avg_num_step_conf": 9.58203125,
      "calib/ece": 0.44301204819277107,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.02359549071618039,
      "calib/mean_conf": 0.1393172690763052,
      "calib/mu_c": 0.14917241379310348,
      "calib/mu_w": 0.12557692307692309,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04394154275501317,
      "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": 2596.0,
      "completions/max_terminated_length": 2596.0,
      "completions/mean_length": 580.53515625,
      "completions/mean_terminated_length": 585.1063232421875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 154.0,
      "epoch": 0.16746666666666668,
      "grad_norm": 0.0340881273150444,
      "learning_rate": 1.1944444444444446e-06,
      "loss": 0.0318,
      "num_tokens": 33637416.0,
      "reward": 1.1834672689437866,
      "reward_std": 0.2716543674468994,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.5463691353797913,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 157
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.893006712800794e-07,
      "aux_brier/mean_group_std": 0.04659970857217336,
      "aux_brier/mean_r": 0.9556051760123047,
      "aux_brier/n_active_tok": 283.25,
      "aux_brier/n_groups": 15.65625,
      "aux_brier/n_step_records": 70.8125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5285946605231061,
      "calib/avg_num_step_conf": 8.921875,
      "calib/ece": 0.48756972111553787,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": -0.010868681393142676,
      "calib/mean_conf": 0.15235059760956174,
      "calib/mu_c": 0.14828025477707008,
      "calib/mu_w": 0.15914893617021275,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.007211155378486055,
      "calib/std_conf": 0.08224715321997639,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3013.0,
      "completions/max_terminated_length": 3013.0,
      "completions/mean_length": 526.14453125,
      "completions/mean_terminated_length": 530.2874145507812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 50.0,
      "epoch": 0.16853333333333334,
      "grad_norm": 0.029545556753873825,
      "learning_rate": 1.1666666666666668e-06,
      "loss": 0.0203,
      "num_tokens": 33877349.0,
      "reward": 1.2334338426589966,
      "reward_std": 0.2396588772535324,
      "rewards/accuracy_reward_step": 0.61328125,
      "rewards/final_brier_reward_step": 0.519672691822052,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 158
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.2961515244723785e-07,
      "aux_brier/mean_group_std": 0.04257286923089113,
      "aux_brier/mean_r": 0.9653046586147206,
      "aux_brier/n_active_tok": 292.125,
      "aux_brier/n_groups": 18.9375,
      "aux_brier/n_step_records": 73.03125,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5898283534647171,
      "calib/avg_num_step_conf": 9.2265625,
      "calib/ece": 0.3743426294820717,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.012827082008900176,
      "calib/mean_conf": 0.14358565737051793,
      "calib/mu_c": 0.14976923076923077,
      "calib/mu_w": 0.1369421487603306,
      "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": 0.0396082866076552,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2583.0,
      "completions/max_terminated_length": 2583.0,
      "completions/mean_length": 543.1484375,
      "completions/mean_terminated_length": 547.4251708984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 227.0,
      "epoch": 0.1696,
      "grad_norm": 0.1250731647014618,
      "learning_rate": 1.138888888888889e-06,
      "loss": 0.0708,
      "num_tokens": 34121179.0,
      "reward": 1.1458985805511475,
      "reward_std": 0.21349376440048218,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.5992199182510376,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 159
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.480219699256935e-06,
      "aux_brier/mean_group_std": 0.0334162566209208,
      "aux_brier/mean_r": 0.9741950480593927,
      "aux_brier/n_active_tok": 286.5,
      "aux_brier/n_groups": 14.5625,
      "aux_brier/n_step_records": 71.625,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6212211221122113,
      "calib/avg_num_step_conf": 9.30078125,
      "calib/ece": 0.4492828685258964,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.00398406374501992,
      "calib/gap": 0.007781518151815159,
      "calib/mean_conf": 0.15613545816733068,
      "calib/mu_c": 0.15926666666666667,
      "calib/mu_w": 0.1514851485148515,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003904382470119522,
      "calib/std_conf": 0.06574584029143392,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2403.0,
      "completions/max_terminated_length": 2403.0,
      "completions/mean_length": 520.37890625,
      "completions/mean_terminated_length": 526.5494384765625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 205.0,
      "epoch": 0.17066666666666666,
      "grad_norm": 0.1191963478922844,
      "learning_rate": 1.111111111111111e-06,
      "loss": -0.0061,
      "num_tokens": 34359236.0,
      "reward": 1.2144297361373901,
      "reward_std": 0.2789672613143921,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.5530316233634949,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 160
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.7483956492037045e-06,
      "aux_brier/mean_group_std": 0.0687977784760483,
      "aux_brier/mean_r": 0.9512190988550542,
      "aux_brier/n_active_tok": 285.5,
      "aux_brier/n_groups": 16.28125,
      "aux_brier/n_step_records": 71.375,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5962960080927555,
      "calib/avg_num_step_conf": 9.25,
      "calib/ece": 0.5571428571428573,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.019979768111431012,
      "calib/mean_conf": 0.1611111111111111,
      "calib/mu_c": 0.16674033149171272,
      "calib/mu_w": 0.1467605633802817,
      "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": 0.07475592265353499,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2405.0,
      "completions/max_terminated_length": 2405.0,
      "completions/mean_length": 533.8671875,
      "completions/mean_terminated_length": 538.0708618164062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 184.0,
      "epoch": 0.17173333333333332,
      "grad_norm": 0.1318727284669876,
      "learning_rate": 1.0833333333333335e-06,
      "loss": 0.0067,
      "num_tokens": 34599826.0,
      "reward": 1.3139612674713135,
      "reward_std": 0.2085951864719391,
      "rewards/accuracy_reward_step": 0.70703125,
      "rewards/final_brier_reward_step": 0.47459492087364197,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 161
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.298125269944995e-07,
      "aux_brier/mean_group_std": 0.049664201148746846,
      "aux_brier/mean_r": 0.959903875455974,
      "aux_brier/n_active_tok": 281.25,
      "aux_brier/n_groups": 15.09375,
      "aux_brier/n_step_records": 70.3125,
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.44551948051948054,
      "calib/avg_num_step_conf": 8.91796875,
      "calib/ece": 0.4522047244094488,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.00795584415584416,
      "calib/mean_conf": 0.15897637795275588,
      "calib/mu_c": 0.15584415584415584,
      "calib/mu_w": 0.1638,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.002440944881889764,
      "calib/std_conf": 0.04499054398757696,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2582.0,
      "completions/max_terminated_length": 2582.0,
      "completions/mean_length": 526.84765625,
      "completions/mean_terminated_length": 528.9137573242188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 166.0,
      "epoch": 0.1728,
      "grad_norm": 0.08773786574602127,
      "learning_rate": 1.0555555555555557e-06,
      "loss": 0.0125,
      "num_tokens": 34838843.0,
      "reward": 1.2393226623535156,
      "reward_std": 0.24485798180103302,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.5510406494140625,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 162
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 6.975275894799404e-07,
      "aux_brier/mean_group_std": 0.04379480709244599,
      "aux_brier/mean_r": 0.9603404263289442,
      "aux_brier/n_active_tok": 314.875,
      "aux_brier/n_groups": 19.46875,
      "aux_brier/n_step_records": 78.71875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5445731389901093,
      "calib/avg_num_step_conf": 10.2578125,
      "calib/ece": 0.373413654618474,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.004016064257028112,
      "calib/gap": 0.01171655387818843,
      "calib/mean_conf": 0.1727710843373494,
      "calib/mu_c": 0.17808823529411763,
      "calib/mu_w": 0.1663716814159292,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.06473262747229393,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2452.0,
      "completions/max_terminated_length": 2452.0,
      "completions/mean_length": 596.85546875,
      "completions/mean_terminated_length": 601.5551147460938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.17386666666666667,
      "grad_norm": 0.021641548722982407,
      "learning_rate": 1.0277777777777777e-06,
      "loss": 0.0425,
      "num_tokens": 35096470.0,
      "reward": 1.1708632707595825,
      "reward_std": 0.2582431137561798,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.5975155830383301,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 163
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.549956473936303e-07,
      "aux_brier/mean_group_std": 0.030194568687748117,
      "aux_brier/mean_r": 0.9703900914496881,
      "aux_brier/n_active_tok": 324.375,
      "aux_brier/n_groups": 18.4375,
      "aux_brier/n_step_records": 81.09375,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5226155358898722,
      "calib/avg_num_step_conf": 10.13671875,
      "calib/ece": 0.3902016129032258,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.004032258064516129,
      "calib/gap": -0.008477876106194687,
      "calib/mean_conf": 0.17786290322580645,
      "calib/mu_c": 0.174,
      "calib/mu_w": 0.18247787610619468,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.01185483870967742,
      "calib/std_conf": 0.07882113629639674,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2622.0,
      "completions/max_terminated_length": 2622.0,
      "completions/mean_length": 631.1875,
      "completions/mean_terminated_length": 633.6627807617188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 237.0,
      "epoch": 0.17493333333333333,
      "grad_norm": 0.2804113030433655,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0745,
      "num_tokens": 35364190.0,
      "reward": 1.158782958984375,
      "reward_std": 0.24492324888706207,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.5882566571235657,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 164
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.6713932611267523e-06,
      "aux_brier/mean_group_std": 0.043575130170754296,
      "aux_brier/mean_r": 0.9642686975868887,
      "aux_brier/n_active_tok": 307.125,
      "aux_brier/n_groups": 16.46875,
      "aux_brier/n_step_records": 76.78125,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5588864402385273,
      "calib/avg_num_step_conf": 10.171875,
      "calib/ece": 0.2985140562248996,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009804252009333686,
      "calib/mean_conf": 0.16734939759036147,
      "calib/mu_c": 0.1725862068965517,
      "calib/mu_w": 0.16278195488721803,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.039568485817860495,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2695.0,
      "completions/max_terminated_length": 2695.0,
      "completions/mean_length": 568.49609375,
      "completions/mean_terminated_length": 577.5198974609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 205.0,
      "epoch": 0.176,
      "grad_norm": 0.23183128237724304,
      "learning_rate": 9.722222222222224e-07,
      "loss": 0.0162,
      "num_tokens": 35615301.0,
      "reward": 1.0983335971832275,
      "reward_std": 0.27246785163879395,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6433343887329102,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 165
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.2855281397022855e-07,
      "aux_brier/mean_group_std": 0.06651987752337103,
      "aux_brier/mean_r": 0.9438013246607929,
      "aux_brier/n_active_tok": 316.125,
      "aux_brier/n_groups": 17.4375,
      "aux_brier/n_step_records": 79.03125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5278820558323664,
      "calib/avg_num_step_conf": 10.2265625,
      "calib/ece": 0.4563095238095239,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.000769230769230772,
      "calib/mean_conf": 0.18257936507936506,
      "calib/mu_c": 0.18285714285714283,
      "calib/mu_w": 0.18208791208791206,
      "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": 0.0483137650329631,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 2914.0,
      "completions/max_terminated_length": 2914.0,
      "completions/mean_length": 600.9453125,
      "completions/mean_terminated_length": 605.6771850585938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 182.0,
      "epoch": 0.17706666666666668,
      "grad_norm": 0.0220769215375185,
      "learning_rate": 9.444444444444445e-07,
      "loss": 0.0317,
      "num_tokens": 35875327.0,
      "reward": 1.2528619766235352,
      "reward_std": 0.2590295672416687,
      "rewards/accuracy_reward_step": 0.62890625,
      "rewards/final_brier_reward_step": 0.5426980257034302,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 166
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -8.33835688496265e-07,
      "aux_brier/mean_group_std": 0.05013251534438689,
      "aux_brier/mean_r": 0.9487601246756485,
      "aux_brier/n_active_tok": 307.75,
      "aux_brier/n_groups": 19.25,
      "aux_brier/n_step_records": 76.9375,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5766081052705256,
      "calib/avg_num_step_conf": 9.81640625,
      "calib/ece": 0.4459677419354839,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.004032258064516129,
      "calib/gap": 0.01416112549870513,
      "calib/mean_conf": 0.18709677419354842,
      "calib/mu_c": 0.19229299363057323,
      "calib/mu_w": 0.1781318681318681,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.07162034980173725,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01171875,
      "completions/max_length": 2687.0,
      "completions/max_terminated_length": 2687.0,
      "completions/mean_length": 593.51953125,
      "completions/mean_terminated_length": 600.5573120117188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 209.0,
      "epoch": 0.17813333333333334,
      "grad_norm": 0.12341222912073135,
      "learning_rate": 9.166666666666666e-07,
      "loss": 0.0711,
      "num_tokens": 36132876.0,
      "reward": 1.2357680797576904,
      "reward_std": 0.26579955220222473,
      "rewards/accuracy_reward_step": 0.61328125,
      "rewards/final_brier_reward_step": 0.5524476766586304,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 167
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.599532062663723e-07,
      "aux_brier/mean_group_std": 0.07778371042239708,
      "aux_brier/mean_r": 0.9417164956729162,
      "aux_brier/n_active_tok": 324.0,
      "aux_brier/n_groups": 19.15625,
      "aux_brier/n_step_records": 81.0,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5481107800650055,
      "calib/avg_num_step_conf": 10.50390625,
      "calib/ece": 0.39934959349593496,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.007831798483206975,
      "calib/mean_conf": 0.17788617886178865,
      "calib/mu_c": 0.18119718309859156,
      "calib/mu_w": 0.1733653846153846,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04079000420596301,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2611.0,
      "completions/max_terminated_length": 2611.0,
      "completions/mean_length": 607.4453125,
      "completions/mean_terminated_length": 617.0873413085938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 190.0,
      "epoch": 0.1792,
      "grad_norm": 0.1288532018661499,
      "learning_rate": 8.88888888888889e-07,
      "loss": 0.0463,
      "num_tokens": 36393054.0,
      "reward": 1.1789710521697998,
      "reward_std": 0.30341869592666626,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.5752593278884888,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 168
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.4110437178138646e-06,
      "aux_brier/mean_group_std": 0.04138852539772683,
      "aux_brier/mean_r": 0.9618776345813835,
      "aux_brier/n_active_tok": 306.5,
      "aux_brier/n_groups": 16.84375,
      "aux_brier/n_step_records": 76.625,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.6126653439153439,
      "calib/avg_num_step_conf": 10.2265625,
      "calib/ece": 0.3870564516129032,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.016166666666666635,
      "calib/mean_conf": 0.17745967741935487,
      "calib/mu_c": 0.1845,
      "calib/mu_w": 0.16833333333333336,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04478891337358236,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2643.0,
      "completions/max_terminated_length": 2643.0,
      "completions/mean_length": 574.2890625,
      "completions/mean_terminated_length": 588.072021484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 131.0,
      "epoch": 0.18026666666666666,
      "grad_norm": 0.30349060893058777,
      "learning_rate": 8.611111111111112e-07,
      "loss": -0.0289,
      "num_tokens": 36644256.0,
      "reward": 1.1761419773101807,
      "reward_std": 0.251577228307724,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.5873804688453674,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 169
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.2160936344873807e-06,
      "aux_brier/mean_group_std": 0.04908099203046632,
      "aux_brier/mean_r": 0.9580694977165491,
      "aux_brier/n_active_tok": 314.875,
      "aux_brier/n_groups": 16.09375,
      "aux_brier/n_step_records": 78.71875,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5563486616334935,
      "calib/avg_num_step_conf": 10.328125,
      "calib/ece": 0.4449799196787149,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.004921070693205243,
      "calib/mean_conf": 0.18072289156626506,
      "calib/mu_c": 0.1825806451612903,
      "calib/mu_w": 0.17765957446808506,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0016064257028112448,
      "calib/std_conf": 0.03979212448155175,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2064.0,
      "completions/max_terminated_length": 2064.0,
      "completions/mean_length": 578.5546875,
      "completions/mean_terminated_length": 585.4150390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.18133333333333335,
      "grad_norm": 0.14562365412712097,
      "learning_rate": 8.333333333333333e-07,
      "loss": 0.0099,
      "num_tokens": 36896518.0,
      "reward": 1.2218576669692993,
      "reward_std": 0.2752302885055542,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.5436804890632629,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 170
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.1814434455748035e-06,
      "aux_brier/mean_group_std": 0.053518122634929895,
      "aux_brier/mean_r": 0.9501083085068752,
      "aux_brier/n_active_tok": 315.125,
      "aux_brier/n_groups": 16.90625,
      "aux_brier/n_step_records": 78.78125,
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6047500000000001,
      "calib/avg_num_step_conf": 10.27734375,
      "calib/ece": 0.32648221343873524,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.018338749999999987,
      "calib/mean_conf": 0.18015810276679842,
      "calib/mu_c": 0.18921875,
      "calib/mu_w": 0.17088,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00035573122529644266,
      "calib/std_conf": 0.04962283341769699,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2884.0,
      "completions/max_terminated_length": 2884.0,
      "completions/mean_length": 573.12109375,
      "completions/mean_terminated_length": 577.6338500976562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 188.0,
      "epoch": 0.1824,
      "grad_norm": 0.024209948256611824,
      "learning_rate": 8.055555555555557e-07,
      "loss": 0.0182,
      "num_tokens": 37150133.0,
      "reward": 1.1526150703430176,
      "reward_std": 0.277605801820755,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.6417101621627808,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 171
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.2334762645727082e-06,
      "aux_brier/mean_group_std": 0.040059631960310045,
      "aux_brier/mean_r": 0.9610557823416042,
      "aux_brier/n_active_tok": 304.875,
      "aux_brier/n_groups": 16.4375,
      "aux_brier/n_step_records": 76.21875,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.4816964285714286,
      "calib/avg_num_step_conf": 9.546875,
      "calib/ece": 0.4575298804780876,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.0013612637362637459,
      "calib/mean_conf": 0.18286852589641436,
      "calib/mu_c": 0.18237499999999998,
      "calib/mu_w": 0.18373626373626373,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.001474103585657371,
      "calib/std_conf": 0.0393743228034895,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2581.0,
      "completions/max_terminated_length": 2581.0,
      "completions/mean_length": 536.265625,
      "completions/mean_terminated_length": 542.62451171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 190.0,
      "epoch": 0.18346666666666667,
      "grad_norm": 0.10507678985595703,
      "learning_rate": 7.777777777777779e-07,
      "loss": -0.0064,
      "num_tokens": 37390769.0,
      "reward": 1.2439298629760742,
      "reward_std": 0.26565784215927124,
      "rewards/accuracy_reward_step": 0.625,
      "rewards/final_brier_reward_step": 0.5382195711135864,
      "rewards/format_reward_step_strict": 0.96875,
      "step": 172
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 8.35707126611851e-07,
      "aux_brier/mean_group_std": 0.06786904352989624,
      "aux_brier/mean_r": 0.9430748218994833,
      "aux_brier/n_active_tok": 309.625,
      "aux_brier/n_groups": 16.53125,
      "aux_brier/n_step_records": 77.40625,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5970543557228717,
      "calib/avg_num_step_conf": 9.9453125,
      "calib/ece": 0.40015999999999996,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.004,
      "calib/gap": 0.018617000198137446,
      "calib/mean_conf": 0.18783999999999998,
      "calib/mu_c": 0.19551020408163264,
      "calib/mu_w": 0.1768932038834952,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.06652619333766212,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2432.0,
      "completions/max_terminated_length": 2432.0,
      "completions/mean_length": 580.0234375,
      "completions/mean_terminated_length": 584.590576171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.18453333333333333,
      "grad_norm": 0.2762625217437744,
      "learning_rate": 7.5e-07,
      "loss": 0.0359,
      "num_tokens": 37642415.0,
      "reward": 1.2034776210784912,
      "reward_std": 0.2693626284599304,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.5795355439186096,
      "rewards/format_reward_step_strict": 0.9609375,
      "step": 173
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.6533995772792807e-07,
      "aux_brier/mean_group_std": 0.0577192916816698,
      "aux_brier/mean_r": 0.9514894683358318,
      "aux_brier/n_active_tok": 303.375,
      "aux_brier/n_groups": 16.4375,
      "aux_brier/n_step_records": 75.84375,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.6169354838709677,
      "calib/avg_num_step_conf": 9.5703125,
      "calib/ece": 0.3134016393442623,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.014459677419354833,
      "calib/mean_conf": 0.17840163934426229,
      "calib/mu_c": 0.18575,
      "calib/mu_w": 0.17129032258064517,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.037996798794425114,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2874.0,
      "completions/max_terminated_length": 2874.0,
      "completions/mean_length": 640.51171875,
      "completions/mean_terminated_length": 645.5551147460938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 207.0,
      "epoch": 0.1856,
      "grad_norm": 0.16314463317394257,
      "learning_rate": 7.222222222222222e-07,
      "loss": 0.0737,
      "num_tokens": 37910618.0,
      "reward": 1.0991191864013672,
      "reward_std": 0.3627810478210449,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6230390667915344,
      "rewards/format_reward_step_strict": 0.94921875,
      "step": 174
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.430361870335961e-07,
      "aux_brier/mean_group_std": 0.04861527252577536,
      "aux_brier/mean_r": 0.9552268439595475,
      "aux_brier/n_active_tok": 329.875,
      "aux_brier/n_groups": 20.9375,
      "aux_brier/n_step_records": 82.46875,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5253977968176254,
      "calib/avg_num_step_conf": 10.55078125,
      "calib/ece": 0.2934979423868313,
      "calib/final_conf_rate": 0.94921875,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.00411522633744856,
      "calib/gap": -0.004075887392900823,
      "calib/mean_conf": 0.1911111111111111,
      "calib/mu_c": 0.18894736842105264,
      "calib/mu_w": 0.19302325581395346,
      "calib/nonempty_final_conf_rate": 0.94921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.007736625514403293,
      "calib/std_conf": 0.07984552988098526,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2722.0,
      "completions/max_terminated_length": 2722.0,
      "completions/mean_length": 665.56640625,
      "completions/mean_terminated_length": 676.1309814453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.18666666666666668,
      "grad_norm": 0.12036768347024918,
      "learning_rate": 6.944444444444446e-07,
      "loss": 0.0299,
      "num_tokens": 38186827.0,
      "reward": 1.0777885913848877,
      "reward_std": 0.2837650775909424,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.6314672231674194,
      "rewards/format_reward_step_strict": 0.94921875,
      "step": 175
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.174230180455709e-07,
      "aux_brier/mean_group_std": 0.08334401902031244,
      "aux_brier/mean_r": 0.9388947683025433,
      "aux_brier/n_active_tok": 320.75,
      "aux_brier/n_groups": 18.15625,
      "aux_brier/n_step_records": 80.1875,
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5531506147540983,
      "calib/avg_num_step_conf": 10.12109375,
      "calib/ece": 0.30643999999999993,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.00463627049180329,
      "calib/mean_conf": 0.18819999999999998,
      "calib/mu_c": 0.19057377049180327,
      "calib/mu_w": 0.18593749999999998,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0033200000000000005,
      "calib/std_conf": 0.05180308871100255,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2902.0,
      "completions/max_terminated_length": 2902.0,
      "completions/mean_length": 605.5,
      "completions/mean_terminated_length": 607.8745727539062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 195.0,
      "epoch": 0.18773333333333334,
      "grad_norm": 0.08937592804431915,
      "learning_rate": 6.666666666666667e-07,
      "loss": 0.0775,
      "num_tokens": 38445899.0,
      "reward": 1.1259515285491943,
      "reward_std": 0.24563716351985931,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.6444308757781982,
      "rewards/format_reward_step_strict": 0.9765625,
      "step": 176
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -1.7630749749031338e-07,
      "aux_brier/mean_group_std": 0.04622896489964074,
      "aux_brier/mean_r": 0.9582977720744049,
      "aux_brier/n_active_tok": 326.75,
      "aux_brier/n_groups": 19.3125,
      "aux_brier/n_step_records": 81.6875,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5487773321257601,
      "calib/avg_num_step_conf": 10.62890625,
      "calib/ece": 0.34775100401606424,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0013222926639927246,
      "calib/mean_conf": 0.18967871485943777,
      "calib/mu_c": 0.19030534351145037,
      "calib/mu_w": 0.18898305084745765,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.005662650602409638,
      "calib/std_conf": 0.06390172855568649,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 3043.0,
      "completions/max_terminated_length": 3043.0,
      "completions/mean_length": 629.140625,
      "completions/mean_terminated_length": 631.60791015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 234.0,
      "epoch": 0.1888,
      "grad_norm": 0.054024118930101395,
      "learning_rate": 6.388888888888889e-07,
      "loss": 0.0862,
      "num_tokens": 38710791.0,
      "reward": 1.147057056427002,
      "reward_std": 0.24264121055603027,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.6116660833358765,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 177
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 2.4876302640031156e-07,
      "aux_brier/mean_group_std": 0.05794493050943693,
      "aux_brier/mean_r": 0.9485298124462436,
      "aux_brier/n_active_tok": 322.0,
      "aux_brier/n_groups": 18.09375,
      "aux_brier/n_step_records": 80.5,
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5144155844155844,
      "calib/avg_num_step_conf": 10.0625,
      "calib/ece": 0.4217322834645669,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002633766233766238,
      "calib/mean_conf": 0.1851968503937008,
      "calib/mu_c": 0.18623376623376622,
      "calib/mu_w": 0.18359999999999999,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00031496062992125983,
      "calib/std_conf": 0.03993795175118499,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2944.0,
      "completions/max_terminated_length": 2944.0,
      "completions/mean_length": 581.05078125,
      "completions/mean_terminated_length": 581.05078125,
      "completions/min_length": 158.0,
      "completions/min_terminated_length": 158.0,
      "epoch": 0.18986666666666666,
      "grad_norm": 0.2664334774017334,
      "learning_rate": 6.111111111111112e-07,
      "loss": 0.0883,
      "num_tokens": 38965612.0,
      "reward": 1.2366094589233398,
      "reward_std": 0.2786281108856201,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.5714383125305176,
      "rewards/format_reward_step_strict": 0.984375,
      "step": 178
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -9.297074625297386e-07,
      "aux_brier/mean_group_std": 0.051580070457991366,
      "aux_brier/mean_r": 0.9524757367183885,
      "aux_brier/n_active_tok": 329.875,
      "aux_brier/n_groups": 19.0,
      "aux_brier/n_step_records": 82.46875,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5602791426333404,
      "calib/avg_num_step_conf": 10.68359375,
      "calib/ece": 0.43336065573770494,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.00819672131147541,
      "calib/gap": 0.0008388520971302493,
      "calib/mean_conf": 0.19385245901639342,
      "calib/mu_c": 0.19417218543046358,
      "calib/mu_w": 0.19333333333333333,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.004180327868852459,
      "calib/std_conf": 0.0820730054991174,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2795.0,
      "completions/max_terminated_length": 2795.0,
      "completions/mean_length": 611.109375,
      "completions/mean_terminated_length": 623.2828979492188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 269.0,
      "epoch": 0.19093333333333334,
      "grad_norm": 0.21540477871894836,
      "learning_rate": 5.833333333333334e-07,
      "loss": 0.0806,
      "num_tokens": 39228320.0,
      "reward": 1.2010421752929688,
      "reward_std": 0.32985225319862366,
      "rewards/accuracy_reward_step": 0.58984375,
      "rewards/final_brier_reward_step": 0.5463562607765198,
      "rewards/format_reward_step_strict": 0.94921875,
      "step": 179
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -8.247282946294288e-07,
      "aux_brier/mean_group_std": 0.06468195060806523,
      "aux_brier/mean_r": 0.9402379308001811,
      "aux_brier/n_active_tok": 326.0,
      "aux_brier/n_groups": 17.84375,
      "aux_brier/n_step_records": 81.5,
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5629285337552743,
      "calib/avg_num_step_conf": 10.30078125,
      "calib/ece": 0.4245275590551181,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.003937007874015748,
      "calib/gap": 0.007372099156118139,
      "calib/mean_conf": 0.20531496062992127,
      "calib/mu_c": 0.2081012658227848,
      "calib/mu_w": 0.20072916666666665,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00389763779527559,
      "calib/std_conf": 0.07317664968857292,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2416.0,
      "completions/max_terminated_length": 2416.0,
      "completions/mean_length": 645.94140625,
      "completions/mean_terminated_length": 648.4745483398438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.192,
      "grad_norm": 0.016230851411819458,
      "learning_rate": 5.555555555555555e-07,
      "loss": -0.0057,
      "num_tokens": 39497537.0,
      "reward": 1.2594654560089111,
      "reward_std": 0.2163536101579666,
      "rewards/accuracy_reward_step": 0.6171875,
      "rewards/final_brier_reward_step": 0.5847371220588684,
      "rewards/format_reward_step_strict": 0.9921875,
      "step": 180
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.806177327701164e-07,
      "aux_brier/mean_group_std": 0.041868387831200275,
      "aux_brier/mean_r": 0.9589451704188524,
      "aux_brier/n_active_tok": 335.0,
      "aux_brier/n_groups": 19.46875,
      "aux_brier/n_step_records": 83.75,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5820322580645162,
      "calib/avg_num_step_conf": 10.97265625,
      "calib/ece": 0.3192369477911647,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.0035651612903226104,
      "calib/mean_conf": 0.19385542168674696,
      "calib/mu_c": 0.19564516129032258,
      "calib/mu_w": 0.19207999999999997,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.00755020080321285,
      "calib/std_conf": 0.07320814785508771,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2800.0,
      "completions/max_terminated_length": 2800.0,
      "completions/mean_length": 583.48046875,
      "completions/mean_terminated_length": 592.7421264648438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 194.0,
      "epoch": 0.19306666666666666,
      "grad_norm": 0.1458035260438919,
      "learning_rate": 5.277777777777779e-07,
      "loss": 0.041,
      "num_tokens": 39753172.0,
      "reward": 1.1297149658203125,
      "reward_std": 0.24853432178497314,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.6360472440719604,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 181
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.120180851745591e-07,
      "aux_brier/mean_group_std": 0.06463662238827718,
      "aux_brier/mean_r": 0.9461888610953979,
      "aux_brier/n_active_tok": 359.625,
      "aux_brier/n_groups": 23.40625,
      "aux_brier/n_step_records": 89.90625,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5073534661240808,
      "calib/avg_num_step_conf": 11.47265625,
      "calib/ece": 0.46689516129032255,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.008064516129032258,
      "calib/gap": -0.013689583779538828,
      "calib/mean_conf": 0.2015725806451613,
      "calib/mu_c": 0.1967701863354037,
      "calib/mu_w": 0.21045977011494252,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.009637096774193547,
      "calib/std_conf": 0.09607386887409627,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2745.0,
      "completions/max_terminated_length": 2745.0,
      "completions/mean_length": 640.58984375,
      "completions/mean_terminated_length": 643.1019897460938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 237.0,
      "epoch": 0.19413333333333332,
      "grad_norm": 0.12417051941156387,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0344,
      "num_tokens": 40023323.0,
      "reward": 1.2451366186141968,
      "reward_std": 0.2409558892250061,
      "rewards/accuracy_reward_step": 0.62890625,
      "rewards/final_brier_reward_step": 0.5352339744567871,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 182
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -3.3778141284024343e-07,
      "aux_brier/mean_group_std": 0.04932587647033141,
      "aux_brier/mean_r": 0.9535072317581955,
      "aux_brier/n_active_tok": 334.0,
      "aux_brier/n_groups": 21.1875,
      "aux_brier/n_step_records": 83.5,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.4950993555316864,
      "calib/avg_num_step_conf": 10.5390625,
      "calib/ece": 0.34751020408163263,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.013406954887218075,
      "calib/mean_conf": 0.19951020408163267,
      "calib/mu_c": 0.2056390977443609,
      "calib/mu_w": 0.19223214285714282,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.0020816326530612244,
      "calib/std_conf": 0.09193478665835655,
      "calib/step_conf_rate": 0.98828125,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 2796.0,
      "completions/max_terminated_length": 2796.0,
      "completions/mean_length": 670.6796875,
      "completions/mean_terminated_length": 675.9606323242188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 210.0,
      "epoch": 0.1952,
      "grad_norm": 0.03099486045539379,
      "learning_rate": 4.7222222222222226e-07,
      "loss": 0.0785,
      "num_tokens": 40301697.0,
      "reward": 1.1382991075515747,
      "reward_std": 0.3700042963027954,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.5922585725784302,
      "rewards/format_reward_step_strict": 0.94140625,
      "step": 183
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -8.974200826905587e-09,
      "aux_brier/mean_group_std": 0.0670935194535348,
      "aux_brier/mean_r": 0.9483376751504038,
      "aux_brier/n_active_tok": 307.125,
      "aux_brier/n_groups": 15.09375,
      "aux_brier/n_step_records": 76.78125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5974809761217528,
      "calib/avg_num_step_conf": 9.8203125,
      "calib/ece": 0.3993625498007969,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.012492784046182148,
      "calib/mean_conf": 0.19027888446215135,
      "calib/mu_c": 0.1954054054054054,
      "calib/mu_w": 0.18291262135922326,
      "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": 0.04007366116384052,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2572.0,
      "completions/max_terminated_length": 2572.0,
      "completions/mean_length": 581.9609375,
      "completions/mean_terminated_length": 586.5433349609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 243.0,
      "epoch": 0.19626666666666667,
      "grad_norm": 0.07316029071807861,
      "learning_rate": 4.444444444444445e-07,
      "loss": 0.0459,
      "num_tokens": 40555959.0,
      "reward": 1.2161612510681152,
      "reward_std": 0.2546941339969635,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.5912078022956848,
      "rewards/format_reward_step_strict": 0.98046875,
      "step": 184
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -6.893121549111925e-07,
      "aux_brier/mean_group_std": 0.07092038911043529,
      "aux_brier/mean_r": 0.937299235638441,
      "aux_brier/n_active_tok": 338.875,
      "aux_brier/n_groups": 22.5,
      "aux_brier/n_step_records": 84.71875,
      "calib/answer_extract_rate": 0.94140625,
      "calib/auroc": 0.5733463598205241,
      "calib/avg_num_step_conf": 11.4140625,
      "calib/ece": 0.4256016597510373,
      "calib/final_conf_rate": 0.94140625,
      "calib/format_rate": 0.93359375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.005728759588941951,
      "calib/mean_conf": 0.19647302904564312,
      "calib/mu_c": 0.19870748299319727,
      "calib/mu_w": 0.19297872340425531,
      "calib/nonempty_final_conf_rate": 0.94140625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.006058091286307054,
      "calib/std_conf": 0.07369619352868514,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.02734375,
      "completions/max_length": 2751.0,
      "completions/max_terminated_length": 2751.0,
      "completions/mean_length": 613.64453125,
      "completions/mean_terminated_length": 630.8955688476562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 244.0,
      "epoch": 0.19733333333333333,
      "grad_norm": 0.04127538576722145,
      "learning_rate": 4.1666666666666667e-07,
      "loss": -0.019,
      "num_tokens": 40819972.0,
      "reward": 1.1787972450256348,
      "reward_std": 0.32682564854621887,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.5511261820793152,
      "rewards/format_reward_step_strict": 0.93359375,
      "step": 185
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.4500213896262437e-07,
      "aux_brier/mean_group_std": 0.05271762792050771,
      "aux_brier/mean_r": 0.9556347104159724,
      "aux_brier/n_active_tok": 328.5,
      "aux_brier/n_groups": 16.65625,
      "aux_brier/n_step_records": 82.125,
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5622324452111687,
      "calib/avg_num_step_conf": 10.26953125,
      "calib/ece": 0.3754761904761904,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.007936507936507936,
      "calib/gap": -0.0025876940770558354,
      "calib/mean_conf": 0.20404761904761903,
      "calib/mu_c": 0.20290780141843973,
      "calib/mu_w": 0.20549549549549556,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.009999999999999998,
      "calib/std_conf": 0.08374354816587623,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2870.0,
      "completions/max_terminated_length": 2870.0,
      "completions/mean_length": 641.70703125,
      "completions/mean_terminated_length": 641.70703125,
      "completions/min_length": 223.0,
      "completions/min_terminated_length": 223.0,
      "epoch": 0.1984,
      "grad_norm": 0.04874926805496216,
      "learning_rate": 3.8888888888888895e-07,
      "loss": 0.0724,
      "num_tokens": 41089289.0,
      "reward": 1.187770128250122,
      "reward_std": 0.24495184421539307,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.6026433706283569,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 186
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.899584302193617e-07,
      "aux_brier/mean_group_std": 0.05150954006876684,
      "aux_brier/mean_r": 0.9477049048969173,
      "aux_brier/n_active_tok": 402.875,
      "aux_brier/n_groups": 30.0,
      "aux_brier/n_step_records": 100.71875,
      "calib/answer_extract_rate": 0.93359375,
      "calib/auroc": 0.474323386537127,
      "calib/avg_num_step_conf": 12.73046875,
      "calib/ece": 0.3715336099585062,
      "calib/final_conf_rate": 0.94140625,
      "calib/format_rate": 0.93359375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": -0.01385443442054135,
      "calib/mean_conf": 0.19601825726141078,
      "calib/mu_c": 0.18969465648854958,
      "calib/mu_w": 0.20354909090909093,
      "calib/nonempty_final_conf_rate": 0.94140625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.011991701244813278,
      "calib/std_conf": 0.0640742333899886,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 3062.0,
      "completions/max_terminated_length": 3062.0,
      "completions/mean_length": 724.9140625,
      "completions/mean_terminated_length": 730.6220703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 185.0,
      "epoch": 0.19946666666666665,
      "grad_norm": 0.09774050116539001,
      "learning_rate": 3.611111111111111e-07,
      "loss": 0.0867,
      "num_tokens": 41376411.0,
      "reward": 1.1225202083587646,
      "reward_std": 0.29422616958618164,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.5760183334350586,
      "rewards/format_reward_step_strict": 0.93359375,
      "step": 187
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.1324720969874136e-06,
      "aux_brier/mean_group_std": 0.058332879115439454,
      "aux_brier/mean_r": 0.9388424391540364,
      "aux_brier/n_active_tok": 356.0,
      "aux_brier/n_groups": 22.71875,
      "aux_brier/n_step_records": 89.0,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.45626342642320084,
      "calib/avg_num_step_conf": 11.13671875,
      "calib/ece": 0.42212,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.008,
      "calib/gap": -0.018939312567132133,
      "calib/mean_conf": 0.21124,
      "calib/mu_c": 0.2038157894736842,
      "calib/mu_w": 0.22275510204081633,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.01268,
      "calib/std_conf": 0.10904523098237721,
      "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": 3027.0,
      "completions/max_terminated_length": 3027.0,
      "completions/mean_length": 675.4609375,
      "completions/mean_terminated_length": 678.10986328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 203.0,
      "epoch": 0.20053333333333334,
      "grad_norm": 0.018280059099197388,
      "learning_rate": 3.3333333333333335e-07,
      "loss": 0.0502,
      "num_tokens": 41653401.0,
      "reward": 1.2164448499679565,
      "reward_std": 0.2764585018157959,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.5610917806625366,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 188
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 3.425059591499746e-08,
      "aux_brier/mean_group_std": 0.06300626618321063,
      "aux_brier/mean_r": 0.9338068102297048,
      "aux_brier/n_active_tok": 313.375,
      "aux_brier/n_groups": 18.375,
      "aux_brier/n_step_records": 78.34375,
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.5577608665843959,
      "calib/avg_num_step_conf": 10.26953125,
      "calib/ece": 0.39285714285714285,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.009125188536953183,
      "calib/mean_conf": 0.19081632653061228,
      "calib/mu_c": 0.1946153846153846,
      "calib/mu_w": 0.18549019607843142,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.04165145633800878,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2559.0,
      "completions/max_terminated_length": 2559.0,
      "completions/mean_length": 604.39453125,
      "completions/mean_terminated_length": 611.561279296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 208.0,
      "epoch": 0.2016,
      "grad_norm": 0.06164390221238136,
      "learning_rate": 3.055555555555556e-07,
      "loss": 0.0641,
      "num_tokens": 41915894.0,
      "reward": 1.1790344715118408,
      "reward_std": 0.2340812385082245,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.5755125284194946,
      "rewards/format_reward_step_strict": 0.953125,
      "step": 189
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -4.969700202472893e-07,
      "aux_brier/mean_group_std": 0.06483934127158777,
      "aux_brier/mean_r": 0.9460970959418037,
      "aux_brier/n_active_tok": 371.5,
      "aux_brier/n_groups": 21.375,
      "aux_brier/n_step_records": 92.875,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5802218974222696,
      "calib/avg_num_step_conf": 11.765625,
      "calib/ece": 0.3767741935483871,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.014863141110815797,
      "calib/mean_conf": 0.20266129032258068,
      "calib/mu_c": 0.20901408450704226,
      "calib/mu_w": 0.19415094339622646,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0034274193548387093,
      "calib/std_conf": 0.07059778017811186,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2904.0,
      "completions/max_terminated_length": 2904.0,
      "completions/mean_length": 666.2421875,
      "completions/mean_terminated_length": 674.142333984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 227.0,
      "epoch": 0.20266666666666666,
      "grad_norm": 0.11814999580383301,
      "learning_rate": 2.7777777777777776e-07,
      "loss": 0.0473,
      "num_tokens": 42192060.0,
      "reward": 1.181680679321289,
      "reward_std": 0.3146505057811737,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.5939102172851562,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 190
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.2416882844433772e-07,
      "aux_brier/mean_group_std": 0.04537874532034743,
      "aux_brier/mean_r": 0.9516971164881491,
      "aux_brier/n_active_tok": 377.75,
      "aux_brier/n_groups": 21.96875,
      "aux_brier/n_step_records": 94.4375,
      "calib/answer_extract_rate": 0.94140625,
      "calib/auroc": 0.6035618279569892,
      "calib/avg_num_step_conf": 12.23046875,
      "calib/ece": 0.30159836065573764,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.9375,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.020198924731182788,
      "calib/mean_conf": 0.2065983606557377,
      "calib/mu_c": 0.21653225806451615,
      "calib/mu_w": 0.19633333333333336,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.06420627105831789,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2954.0,
      "completions/max_terminated_length": 2954.0,
      "completions/mean_length": 644.0234375,
      "completions/mean_terminated_length": 651.6600952148438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 195.0,
      "epoch": 0.20373333333333332,
      "grad_norm": 0.046372488141059875,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0722,
      "num_tokens": 42461098.0,
      "reward": 1.1078917980194092,
      "reward_std": 0.2727097272872925,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.6190671920776367,
      "rewards/format_reward_step_strict": 0.9375,
      "step": 191
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 3.9225027770317666e-07,
      "aux_brier/mean_group_std": 0.05678803464707159,
      "aux_brier/mean_r": 0.9381862612549897,
      "aux_brier/n_active_tok": 346.125,
      "aux_brier/n_groups": 21.96875,
      "aux_brier/n_step_records": 86.53125,
      "calib/answer_extract_rate": 0.93359375,
      "calib/auroc": 0.4678402903811252,
      "calib/avg_num_step_conf": 11.12109375,
      "calib/ece": 0.4224166666666666,
      "calib/final_conf_rate": 0.9375,
      "calib/format_rate": 0.921875,
      "calib/frac_conf_gt_0.9": 0.008333333333333333,
      "calib/gap": -0.014548094373865783,
      "calib/mean_conf": 0.201,
      "calib/mu_c": 0.19524137931034483,
      "calib/mu_w": 0.2097894736842106,
      "calib/nonempty_final_conf_rate": 0.9375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.009625,
      "calib/std_conf": 0.09888039913619551,
      "calib/step_conf_rate": 0.9921875,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 2940.0,
      "completions/max_terminated_length": 2940.0,
      "completions/mean_length": 661.5234375,
      "completions/mean_terminated_length": 672.0238647460938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 185.0,
      "epoch": 0.2048,
      "grad_norm": 0.034542765468358994,
      "learning_rate": 2.2222222222222224e-07,
      "loss": 0.0716,
      "num_tokens": 42735424.0,
      "reward": 1.16123366355896,
      "reward_std": 0.3240078389644623,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.5355597734451294,
      "rewards/format_reward_step_strict": 0.921875,
      "step": 192
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 4.24852955260846e-07,
      "aux_brier/mean_group_std": 0.0412554376989771,
      "aux_brier/mean_r": 0.9550871661785169,
      "aux_brier/n_active_tok": 336.375,
      "aux_brier/n_groups": 18.59375,
      "aux_brier/n_step_records": 84.09375,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5379738729508197,
      "calib/avg_num_step_conf": 10.9296875,
      "calib/ece": 0.31508,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.004,
      "calib/gap": 0.005493084016393424,
      "calib/mean_conf": 0.20395999999999997,
      "calib/mu_c": 0.206640625,
      "calib/mu_w": 0.20114754098360657,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0035199999999999997,
      "calib/std_conf": 0.08556587170128054,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2364.0,
      "completions/max_terminated_length": 2364.0,
      "completions/mean_length": 600.140625,
      "completions/mean_terminated_length": 612.0956420898438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 280.0,
      "epoch": 0.20586666666666667,
      "grad_norm": 0.12756557762622833,
      "learning_rate": 1.9444444444444447e-07,
      "loss": -0.0083,
      "num_tokens": 42994772.0,
      "reward": 1.1384763717651367,
      "reward_std": 0.34489667415618896,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.6242179870605469,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 193
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 3.371264476192959e-07,
      "aux_brier/mean_group_std": 0.07162886939300656,
      "aux_brier/mean_r": 0.9427285563837255,
      "aux_brier/n_active_tok": 319.25,
      "aux_brier/n_groups": 17.6875,
      "aux_brier/n_step_records": 79.8125,
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.647510593220339,
      "calib/avg_num_step_conf": 10.54296875,
      "calib/ece": 0.33292682926829276,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.021532044491525487,
      "calib/mean_conf": 0.18959349593495936,
      "calib/mu_c": 0.19992187500000003,
      "calib/mu_w": 0.17838983050847454,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0010975609756097562,
      "calib/std_conf": 0.04981099845870625,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.01953125,
      "completions/max_length": 2583.0,
      "completions/max_terminated_length": 2583.0,
      "completions/mean_length": 580.15234375,
      "completions/mean_terminated_length": 591.7091674804688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 221.0,
      "epoch": 0.20693333333333333,
      "grad_norm": 0.038445133715867996,
      "learning_rate": 1.6666666666666668e-07,
      "loss": -0.0021,
      "num_tokens": 43249235.0,
      "reward": 1.133557677268982,
      "reward_std": 0.27232956886291504,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.6201683282852173,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 194
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -4.460890020985264e-07,
      "aux_brier/mean_group_std": 0.05744066556824241,
      "aux_brier/mean_r": 0.9387691897144076,
      "aux_brier/n_active_tok": 331.875,
      "aux_brier/n_groups": 20.84375,
      "aux_brier/n_step_records": 82.96875,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.47642195767195766,
      "calib/avg_num_step_conf": 10.64453125,
      "calib/ece": 0.35194331983805666,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.008097165991902834,
      "calib/gap": -0.0006633597883598241,
      "calib/mean_conf": 0.20222672064777322,
      "calib/mu_c": 0.2019259259259259,
      "calib/mu_w": 0.20258928571428572,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.003805668016194332,
      "calib/std_conf": 0.08245905139190195,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2957.0,
      "completions/max_terminated_length": 2957.0,
      "completions/mean_length": 595.90625,
      "completions/mean_terminated_length": 602.9723510742188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 214.0,
      "epoch": 0.208,
      "grad_norm": 0.10510091483592987,
      "learning_rate": 1.3888888888888888e-07,
      "loss": 0.0362,
      "num_tokens": 43507771.0,
      "reward": 1.1608781814575195,
      "reward_std": 0.2738897204399109,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.6044504046440125,
      "rewards/format_reward_step_strict": 0.96484375,
      "step": 195
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.751512754947896e-07,
      "aux_brier/mean_group_std": 0.05371166723777047,
      "aux_brier/mean_r": 0.9496463667116218,
      "aux_brier/n_active_tok": 331.5,
      "aux_brier/n_groups": 20.09375,
      "aux_brier/n_step_records": 82.875,
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5146788990825688,
      "calib/avg_num_step_conf": 10.359375,
      "calib/ece": 0.36634538152610435,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.002650720838794224,
      "calib/mean_conf": 0.19919678714859435,
      "calib/mu_c": 0.20035714285714284,
      "calib/mu_w": 0.19770642201834862,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.001646586345381526,
      "calib/std_conf": 0.04424463561656084,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2733.0,
      "completions/max_terminated_length": 2733.0,
      "completions/mean_length": 556.12890625,
      "completions/mean_terminated_length": 558.309814453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 201.0,
      "epoch": 0.20906666666666668,
      "grad_norm": 0.050895657390356064,
      "learning_rate": 1.1111111111111112e-07,
      "loss": 0.0711,
      "num_tokens": 43752684.0,
      "reward": 1.1843090057373047,
      "reward_std": 0.23462121188640594,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6044234037399292,
      "rewards/format_reward_step_strict": 0.97265625,
      "step": 196
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 5.12161457244531e-07,
      "aux_brier/mean_group_std": 0.08229145749638564,
      "aux_brier/mean_r": 0.932691426152198,
      "aux_brier/n_active_tok": 332.25,
      "aux_brier/n_groups": 18.75,
      "aux_brier/n_step_records": 83.0625,
      "calib/answer_extract_rate": 0.9453125,
      "calib/auroc": 0.61247947454844,
      "calib/avg_num_step_conf": 10.6015625,
      "calib/ece": 0.3320247933884297,
      "calib/final_conf_rate": 0.9453125,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.0,
      "calib/gap": 0.013883415435139546,
      "calib/mean_conf": 0.19929752066115705,
      "calib/mu_c": 0.20595238095238097,
      "calib/mu_w": 0.19206896551724142,
      "calib/nonempty_final_conf_rate": 0.9453125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.005330578512396694,
      "calib/std_conf": 0.06305515683725031,
      "calib/step_conf_rate": 0.99609375,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.015625,
      "completions/max_length": 2876.0,
      "completions/max_terminated_length": 2876.0,
      "completions/mean_length": 621.7421875,
      "completions/mean_terminated_length": 631.6111450195312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 147.0,
      "epoch": 0.21013333333333334,
      "grad_norm": 0.022467566654086113,
      "learning_rate": 8.333333333333334e-08,
      "loss": 0.1037,
      "num_tokens": 44016906.0,
      "reward": 1.1161017417907715,
      "reward_std": 0.27242499589920044,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.6128445267677307,
      "rewards/format_reward_step_strict": 0.94140625,
      "step": 197
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 7.942384199211006e-07,
      "aux_brier/mean_group_std": 0.0473581039447935,
      "aux_brier/mean_r": 0.9577410146574017,
      "aux_brier/n_active_tok": 338.625,
      "aux_brier/n_groups": 21.90625,
      "aux_brier/n_step_records": 84.65625,
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5420021470746108,
      "calib/avg_num_step_conf": 10.890625,
      "calib/ece": 0.36963414634146347,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.0040650406504065045,
      "calib/gap": 0.015100644122383239,
      "calib/mean_conf": 0.19134146341463415,
      "calib/mu_c": 0.19797101449275362,
      "calib/mu_w": 0.18287037037037038,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.0,
      "calib/std_conf": 0.07009451566501977,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2917.0,
      "completions/max_terminated_length": 2917.0,
      "completions/mean_length": 603.48828125,
      "completions/mean_terminated_length": 613.0675048828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 172.0,
      "epoch": 0.2112,
      "grad_norm": 0.07373571395874023,
      "learning_rate": 5.555555555555556e-08,
      "loss": 0.0393,
      "num_tokens": 44276783.0,
      "reward": 1.1614110469818115,
      "reward_std": 0.2225344032049179,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.5909566283226013,
      "rewards/format_reward_step_strict": 0.94921875,
      "step": 198
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": 1.823768354392019e-08,
      "aux_brier/mean_group_std": 0.05707974939527955,
      "aux_brier/mean_r": 0.9469600337955227,
      "aux_brier/n_active_tok": 377.625,
      "aux_brier/n_groups": 22.5,
      "aux_brier/n_step_records": 94.40625,
      "calib/answer_extract_rate": 0.94140625,
      "calib/auroc": 0.4645641882483988,
      "calib/avg_num_step_conf": 12.3828125,
      "calib/ece": 0.3563485477178423,
      "calib/final_conf_rate": 0.94140625,
      "calib/format_rate": 0.9296875,
      "calib/frac_conf_gt_0.9": 0.008298755186721992,
      "calib/gap": -0.02200779727095517,
      "calib/mean_conf": 0.21933609958506223,
      "calib/mu_c": 0.2094736842105263,
      "calib/mu_w": 0.23148148148148148,
      "calib/nonempty_final_conf_rate": 0.94140625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.011908713692946056,
      "calib/std_conf": 0.08807517008094146,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.03515625,
      "completions/max_length": 2848.0,
      "completions/max_terminated_length": 2848.0,
      "completions/mean_length": 656.34765625,
      "completions/mean_terminated_length": 680.26318359375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 184.0,
      "epoch": 0.21226666666666666,
      "grad_norm": 0.13624730706214905,
      "learning_rate": 2.777777777777778e-08,
      "loss": 0.0144,
      "num_tokens": 44549008.0,
      "reward": 1.1288948059082031,
      "reward_std": 0.35054564476013184,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.5780788660049438,
      "rewards/format_reward_step_strict": 0.9296875,
      "step": 199
    },
    {
      "aux_brier/lambda": 0.050000000000000024,
      "aux_brier/loss": -2.408442279396017e-07,
      "aux_brier/mean_group_std": 0.05124705039918557,
      "aux_brier/mean_r": 0.9541607324915391,
      "aux_brier/n_active_tok": 334.25,
      "aux_brier/n_groups": 20.625,
      "aux_brier/n_step_records": 83.5625,
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5763718263718264,
      "calib/avg_num_step_conf": 10.86328125,
      "calib/ece": 0.41603238866396763,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.008097165991902834,
      "calib/gap": 0.0011322686322685904,
      "calib/mean_conf": 0.20582995951417005,
      "calib/mu_c": 0.20628378378378376,
      "calib/mu_w": 0.20515151515151517,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.011336032388663968,
      "calib/std_conf": 0.09908089481737702,
      "calib/step_conf_rate": 1.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2744.0,
      "completions/max_terminated_length": 2744.0,
      "completions/mean_length": 611.85546875,
      "completions/mean_terminated_length": 621.5675048828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 143.0,
      "epoch": 0.21333333333333335,
      "grad_norm": 0.10868275910615921,
      "learning_rate": 0.0,
      "loss": 0.004,
      "num_tokens": 44813691.0,
      "reward": 1.199366807937622,
      "reward_std": 0.2432486116886139,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.5709050893783569,
      "rewards/format_reward_step_strict": 0.95703125,
      "step": 200
    },
    {
      "epoch": 0.21333333333333335,
      "step": 200,
      "total_flos": 0.0,
      "train_loss": 0.03981497859931551,
      "train_runtime": 16137.1786,
      "train_samples_per_second": 3.173,
      "train_steps_per_second": 0.012
    }
  ],
  "logging_steps": 1,
  "max_steps": 200,
  "num_input_tokens_seen": 44813691,
  "num_train_epochs": 1,
  "save_steps": 40,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}