{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.21333333333333335,
  "eval_steps": 500,
  "global_step": 200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.38076182006817844,
      "calib/avg_num_step_conf": 5.23046875,
      "calib/ece": 0.2003187250996017,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.2948207171314741,
      "calib/gap": -0.026059730250481805,
      "calib/mean_conf": 0.8737051792828686,
      "calib/mu_c": 0.865606936416185,
      "calib/mu_w": 0.8916666666666668,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.19239043824701207,
      "calib/std_conf": 0.09027744273295583,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7959393232205367,
      "calib/step_q_c_n": 857.0,
      "calib/step_q_gap": -0.006446568895645877,
      "calib/step_q_w": 0.8023858921161826,
      "calib/step_q_w_n": 482.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2492.0,
      "completions/max_terminated_length": 2492.0,
      "completions/mean_length": 474.94921875,
      "completions/mean_terminated_length": 478.68896484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 138.0,
      "epoch": 0.0010666666666666667,
      "grad_norm": 0.007058217190206051,
      "kl": 0.000291675329208374,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0332,
      "num_tokens": 229171.0,
      "reward": 0.5959470272064209,
      "reward_std": 0.34700775146484375,
      "rewards/accuracy_reward_step": 0.67578125,
      "rewards/final_brier_reward_step": 0.7142800688743591,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.14714524149894714,
      "step": 1
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.44343065693430656,
      "calib/avg_num_step_conf": 5.05859375,
      "calib/ece": 0.3349411764705883,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.2823529411764706,
      "calib/gap": 0.002352468143016151,
      "calib/mean_conf": 0.8721960784313726,
      "calib/mu_c": 0.8732846715328467,
      "calib/mu_w": 0.8709322033898306,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3349411764705883,
      "calib/std_conf": 0.07627016470309335,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7954391371340525,
      "calib/step_q_c_n": 649.0,
      "calib/step_q_gap": 0.011011892552009073,
      "calib/step_q_w": 0.7844272445820434,
      "calib/step_q_w_n": 646.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1966.0,
      "completions/max_terminated_length": 1966.0,
      "completions/mean_length": 492.9765625,
      "completions/mean_terminated_length": 494.9098205566406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.0021333333333333334,
      "grad_norm": 0.00732036679983139,
      "kl": 0.00037539005279541016,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0149,
      "num_tokens": 458661.0,
      "reward": 0.5325330495834351,
      "reward_std": 0.35858410596847534,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.6320762038230896,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.12673982977867126,
      "step": 2
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.43875927889713684,
      "calib/avg_num_step_conf": 5.11328125,
      "calib/ece": 0.2455859374999999,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.36328125,
      "calib/gap": -0.008799045599151345,
      "calib/mean_conf": 0.8862109374999999,
      "calib/mu_c": 0.883048780487805,
      "calib/mu_w": 0.8918478260869563,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2455859374999999,
      "calib/std_conf": 0.04548457024498631,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.8016937669376695,
      "calib/step_q_c_n": 738.0,
      "calib/step_q_gap": 0.04516136763819478,
      "calib/step_q_w": 0.7565323992994747,
      "calib/step_q_w_n": 571.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1381.0,
      "completions/max_terminated_length": 1381.0,
      "completions/mean_length": 491.5390625,
      "completions/mean_terminated_length": 493.4667053222656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 191.0,
      "epoch": 0.0032,
      "grad_norm": 0.007544770836830139,
      "kl": 0.0012790262699127197,
      "learning_rate": 7.5e-07,
      "loss": 0.0274,
      "num_tokens": 689751.0,
      "reward": 0.5081400275230408,
      "reward_std": 0.36799156665802,
      "rewards/accuracy_reward_step": 0.640625,
      "rewards/final_brier_reward_step": 0.6991492509841919,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": -0.009431745857000351,
      "step": 3
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.46746398290480456,
      "calib/avg_num_step_conf": 4.87109375,
      "calib/ece": 0.23257936507936522,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.2777777777777778,
      "calib/gap": 0.0011642655269872293,
      "calib/mean_conf": 0.8794047619047619,
      "calib/mu_c": 0.8798159509202456,
      "calib/mu_w": 0.8786516853932583,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.23257936507936522,
      "calib/std_conf": 0.05107614585828655,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7943381389252948,
      "calib/step_q_c_n": 763.0,
      "calib/step_q_gap": 0.013862932313724596,
      "calib/step_q_w": 0.7804752066115702,
      "calib/step_q_w_n": 484.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1832.0,
      "completions/max_terminated_length": 1832.0,
      "completions/mean_length": 481.61328125,
      "completions/mean_terminated_length": 487.3241271972656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.004266666666666667,
      "grad_norm": 0.006945696193724871,
      "kl": 0.00028836727142333984,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0012,
      "num_tokens": 919212.0,
      "reward": 0.5910017490386963,
      "reward_std": 0.32807308435440063,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.7042097449302673,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.15357500314712524,
      "step": 4
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.4042075163398694,
      "calib/avg_num_step_conf": 4.78125,
      "calib/ece": 0.32217213114754095,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.27049180327868855,
      "calib/gap": -0.01393246187363828,
      "calib/mean_conf": 0.8795491803278689,
      "calib/mu_c": 0.8733823529411765,
      "calib/mu_w": 0.8873148148148148,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.96875,
      "calib/nonempty_step_conf_rate": 0.9609375,
      "calib/pce": 0.32217213114754095,
      "calib/std_conf": 0.04728923948911935,
      "calib/step_conf_rate": 0.9609375,
      "calib/step_q_c": 0.7967194928684628,
      "calib/step_q_c_n": 631.0,
      "calib/step_q_gap": 0.011660470946034507,
      "calib/step_q_w": 0.7850590219224283,
      "calib/step_q_w_n": 593.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 3052.0,
      "completions/max_terminated_length": 3052.0,
      "completions/mean_length": 522.75,
      "completions/mean_terminated_length": 524.800048828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.005333333333333333,
      "grad_norm": 0.006729537155479193,
      "kl": 0.000265657901763916,
      "learning_rate": 1.25e-06,
      "loss": 0.0257,
      "num_tokens": 1159724.0,
      "reward": 0.46391651034355164,
      "reward_std": 0.3454357981681824,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6057500243186951,
      "rewards/format_reward_step": 0.9453125,
      "rewards/step_correlation_reward": 0.026770520955324173,
      "step": 5
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.4620373299588738,
      "calib/avg_num_step_conf": 5.12109375,
      "calib/ece": 0.31055118110236213,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.32677165354330706,
      "calib/gap": -0.005713381841189413,
      "calib/mean_conf": 0.8814173228346457,
      "calib/mu_c": 0.8789655172413793,
      "calib/mu_w": 0.8846788990825687,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.31055118110236213,
      "calib/std_conf": 0.041254432535867074,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7995592286501377,
      "calib/step_q_c_n": 726.0,
      "calib/step_q_gap": -0.006611711520802421,
      "calib/step_q_w": 0.8061709401709402,
      "calib/step_q_w_n": 585.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2127.0,
      "completions/max_terminated_length": 2127.0,
      "completions/mean_length": 440.34765625,
      "completions/mean_terminated_length": 440.34765625,
      "completions/min_length": 167.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.0064,
      "grad_norm": 0.007685510441660881,
      "kl": 0.0007127821445465088,
      "learning_rate": 1.5e-06,
      "loss": 0.0159,
      "num_tokens": 1378405.0,
      "reward": 0.5301942825317383,
      "reward_std": 0.31458911299705505,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.6489687561988831,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.09970106184482574,
      "step": 6
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.48179140447181684,
      "calib/avg_num_step_conf": 5.40234375,
      "calib/ece": 0.2674900398406375,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.2788844621513944,
      "calib/gap": -0.005026107912705968,
      "calib/mean_conf": 0.8785657370517928,
      "calib/mu_c": 0.8766233766233766,
      "calib/mu_w": 0.8816494845360826,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.26625498007968135,
      "calib/std_conf": 0.0434692575958097,
      "calib/step_conf_rate": 0.984375,
      "calib/step_q_c": 0.7888520710059173,
      "calib/step_q_c_n": 845.0,
      "calib/step_q_gap": -0.012225995908580867,
      "calib/step_q_w": 0.8010780669144981,
      "calib/step_q_w_n": 538.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2899.0,
      "completions/max_terminated_length": 2899.0,
      "completions/mean_length": 559.07421875,
      "completions/mean_terminated_length": 559.07421875,
      "completions/min_length": 190.0,
      "completions/min_terminated_length": 190.0,
      "epoch": 0.007466666666666667,
      "grad_norm": 0.006892060860991478,
      "kl": 0.00038546323776245117,
      "learning_rate": 1.75e-06,
      "loss": 0.0366,
      "num_tokens": 1628952.0,
      "reward": 0.527165412902832,
      "reward_std": 0.33386415243148804,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.6698726415634155,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.07039565593004227,
      "step": 7
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.4465880102040817,
      "calib/avg_num_step_conf": 4.6484375,
      "calib/ece": 0.3209126984126984,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.27380952380952384,
      "calib/gap": -0.005874999999999964,
      "calib/mean_conf": 0.876468253968254,
      "calib/mu_c": 0.8738571428571429,
      "calib/mu_w": 0.8797321428571429,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3209126984126984,
      "calib/std_conf": 0.04910131839757406,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7993548387096774,
      "calib/step_q_c_n": 620.0,
      "calib/step_q_gap": 0.03326711941143179,
      "calib/step_q_w": 0.7660877192982456,
      "calib/step_q_w_n": 570.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2512.0,
      "completions/max_terminated_length": 2512.0,
      "completions/mean_length": 515.6953125,
      "completions/mean_terminated_length": 517.7176513671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.008533333333333334,
      "grad_norm": 0.006904206704348326,
      "kl": 0.00042426586151123047,
      "learning_rate": 2.0000000000000003e-06,
      "loss": -0.0211,
      "num_tokens": 1867482.0,
      "reward": 0.49654126167297363,
      "reward_std": 0.30775249004364014,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6335644721984863,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.05404934659600258,
      "step": 8
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5013470400567175,
      "calib/avg_num_step_conf": 4.91796875,
      "calib/ece": 0.23780487804878048,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.21544715447154472,
      "calib/gap": 0.014260191421481827,
      "calib/mean_conf": 0.8678861788617886,
      "calib/mu_c": 0.8731612903225807,
      "calib/mu_w": 0.8589010989010989,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.23780487804878048,
      "calib/std_conf": 0.07653313255108424,
      "calib/step_conf_rate": 0.96484375,
      "calib/step_q_c": 0.7960090361445783,
      "calib/step_q_c_n": 664.0,
      "calib/step_q_gap": 0.10142080085046057,
      "calib/step_q_w": 0.6945882352941177,
      "calib/step_q_w_n": 595.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2378.0,
      "completions/max_terminated_length": 2378.0,
      "completions/mean_length": 491.83203125,
      "completions/mean_terminated_length": 501.6294860839844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 183.0,
      "epoch": 0.0096,
      "grad_norm": 0.007615099661052227,
      "kl": 0.0003460049629211426,
      "learning_rate": 2.25e-06,
      "loss": 0.0029,
      "num_tokens": 2100927.0,
      "reward": 0.5577809810638428,
      "reward_std": 0.35230207443237305,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.6672629117965698,
      "rewards/format_reward_step": 0.94140625,
      "rewards/step_correlation_reward": 0.13892391324043274,
      "step": 9
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.48801742919389984,
      "calib/avg_num_step_conf": 5.17578125,
      "calib/ece": 0.2692063492063492,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.2777777777777778,
      "calib/gap": -0.00218657159833624,
      "calib/mean_conf": 0.8763492063492063,
      "calib/mu_c": 0.8754901960784313,
      "calib/mu_w": 0.8776767676767675,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2692063492063492,
      "calib/std_conf": 0.04302275467708966,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.786162310866575,
      "calib/step_q_c_n": 727.0,
      "calib/step_q_gap": -0.007467271073224313,
      "calib/step_q_w": 0.7936295819397993,
      "calib/step_q_w_n": 598.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2478.0,
      "completions/max_terminated_length": 2478.0,
      "completions/mean_length": 523.890625,
      "completions/mean_terminated_length": 525.9451293945312,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 160.0,
      "epoch": 0.010666666666666666,
      "grad_norm": 0.0069535886868834496,
      "kl": 0.0003247857093811035,
      "learning_rate": 2.5e-06,
      "loss": 0.0169,
      "num_tokens": 2341843.0,
      "reward": 0.5617609620094299,
      "reward_std": 0.3579865097999573,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.6715867519378662,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.1363101601600647,
      "step": 10
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5209415584415583,
      "calib/avg_num_step_conf": 5.5078125,
      "calib/ece": 0.29157480314960627,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.3031496062992126,
      "calib/gap": -0.009902597402597246,
      "calib/mean_conf": 0.871496062992126,
      "calib/mu_c": 0.8675974025974027,
      "calib/mu_w": 0.8775,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2783858267716535,
      "calib/std_conf": 0.0895621786401363,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7814958775029447,
      "calib/step_q_c_n": 849.0,
      "calib/step_q_gap": -0.019627117149461748,
      "calib/step_q_w": 0.8011229946524064,
      "calib/step_q_w_n": 561.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2507.0,
      "completions/max_terminated_length": 2507.0,
      "completions/mean_length": 518.60546875,
      "completions/mean_terminated_length": 518.60546875,
      "completions/min_length": 179.0,
      "completions/min_terminated_length": 179.0,
      "epoch": 0.011733333333333333,
      "grad_norm": 0.006948419380933046,
      "kl": 0.0003502368927001953,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.0181,
      "num_tokens": 2579086.0,
      "reward": 0.5813484191894531,
      "reward_std": 0.3301823139190674,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.6719730496406555,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.17275497317314148,
      "step": 11
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.48346796276013143,
      "calib/avg_num_step_conf": 5.63671875,
      "calib/ece": 0.22937007874015738,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.2874015748031496,
      "calib/gap": -0.005866648411828934,
      "calib/mean_conf": 0.8737795275590551,
      "calib/mu_c": 0.8717469879518073,
      "calib/mu_w": 0.8776136363636362,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.22480314960629913,
      "calib/std_conf": 0.05236048780560937,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7873589164785555,
      "calib/step_q_c_n": 886.0,
      "calib/step_q_gap": 0.007251196550368766,
      "calib/step_q_w": 0.7801077199281867,
      "calib/step_q_w_n": 557.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2107.0,
      "completions/max_terminated_length": 2107.0,
      "completions/mean_length": 456.671875,
      "completions/mean_terminated_length": 458.4627685546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.0128,
      "grad_norm": 0.007615354843437672,
      "kl": 0.0005735158920288086,
      "learning_rate": 3e-06,
      "loss": -0.0368,
      "num_tokens": 2800170.0,
      "reward": 0.575323760509491,
      "reward_std": 0.30451565980911255,
      "rewards/accuracy_reward_step": 0.6484375,
      "rewards/final_brier_reward_step": 0.7140507698059082,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.10847172141075134,
      "step": 12
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5732429335370512,
      "calib/avg_num_step_conf": 4.75,
      "calib/ece": 0.26554687500000007,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.265625,
      "calib/gap": 0.02371530430353963,
      "calib/mean_conf": 0.867109375,
      "calib/mu_c": 0.8765584415584416,
      "calib/mu_w": 0.852843137254902,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.26554687500000007,
      "calib/std_conf": 0.06147474511626197,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.782394366197183,
      "calib/step_q_c_n": 710.0,
      "calib/step_q_gap": 0.019232310861214685,
      "calib/step_q_w": 0.7631620553359684,
      "calib/step_q_w_n": 506.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1232.0,
      "completions/max_terminated_length": 1232.0,
      "completions/mean_length": 458.69921875,
      "completions/mean_terminated_length": 460.4980773925781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 133.0,
      "epoch": 0.013866666666666666,
      "grad_norm": 0.007240649778395891,
      "kl": 0.0006138086318969727,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.0001,
      "num_tokens": 3022189.0,
      "reward": 0.6061074137687683,
      "reward_std": 0.30182117223739624,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.6973890662193298,
      "rewards/format_reward_step": 1.0,
      "rewards/step_correlation_reward": 0.1945132613182068,
      "step": 13
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.36378424100461965,
      "calib/avg_num_step_conf": 5.3515625,
      "calib/ece": 0.32348,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.308,
      "calib/gap": -0.01454811633808295,
      "calib/mean_conf": 0.8799600000000001,
      "calib/mu_c": 0.8736170212765958,
      "calib/mu_w": 0.8881651376146787,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.31972,
      "calib/std_conf": 0.07140587090708998,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7922131147540982,
      "calib/step_q_c_n": 732.0,
      "calib/step_q_gap": 0.02354540315535214,
      "calib/step_q_w": 0.768667711598746,
      "calib/step_q_w_n": 638.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2403.0,
      "completions/max_terminated_length": 2403.0,
      "completions/mean_length": 560.66796875,
      "completions/mean_terminated_length": 560.66796875,
      "completions/min_length": 133.0,
      "completions/min_terminated_length": 133.0,
      "epoch": 0.014933333333333333,
      "grad_norm": 0.006452104542404413,
      "kl": 0.002147197723388672,
      "learning_rate": 3.5e-06,
      "loss": 0.0202,
      "num_tokens": 3271120.0,
      "reward": 0.5175330638885498,
      "reward_std": 0.34320229291915894,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.6269644498825073,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.1026330292224884,
      "step": 14
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.505960135487233,
      "calib/avg_num_step_conf": 4.97265625,
      "calib/ece": 0.2719367588932806,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.2885375494071146,
      "calib/gap": 0.003715476810838947,
      "calib/mean_conf": 0.8727272727272728,
      "calib/mu_c": 0.8742105263157894,
      "calib/mu_w": 0.8704950495049505,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2719367588932806,
      "calib/std_conf": 0.0561343033018714,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7717847769028872,
      "calib/step_q_c_n": 762.0,
      "calib/step_q_gap": 0.0021761663353725424,
      "calib/step_q_w": 0.7696086105675146,
      "calib/step_q_w_n": 511.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1694.0,
      "completions/max_terminated_length": 1694.0,
      "completions/mean_length": 456.48046875,
      "completions/mean_terminated_length": 458.2706298828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 154.0,
      "epoch": 0.016,
      "grad_norm": 0.007334915455430746,
      "kl": 0.0011022090911865234,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.0079,
      "num_tokens": 3495859.0,
      "reward": 0.5454412698745728,
      "reward_std": 0.3382645547389984,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.675797700881958,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.09945990145206451,
      "step": 15
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.488537710255019,
      "calib/avg_num_step_conf": 6.421875,
      "calib/ece": 0.2714859437751005,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.321285140562249,
      "calib/gap": 0.0017986977753660094,
      "calib/mean_conf": 0.8790361445783132,
      "calib/mu_c": 0.8797368421052629,
      "calib/mu_w": 0.8779381443298969,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2700401606425704,
      "calib/std_conf": 0.05849411032373637,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7629069767441861,
      "calib/step_q_c_n": 946.0,
      "calib/step_q_gap": 0.04061471313387088,
      "calib/step_q_w": 0.7222922636103152,
      "calib/step_q_w_n": 698.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2725.0,
      "completions/max_terminated_length": 2725.0,
      "completions/mean_length": 652.53125,
      "completions/mean_terminated_length": 652.53125,
      "completions/min_length": 193.0,
      "completions/min_terminated_length": 193.0,
      "epoch": 0.017066666666666667,
      "grad_norm": 0.00565234012901783,
      "kl": 0.001227736473083496,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0352,
      "num_tokens": 3771755.0,
      "reward": 0.5504010319709778,
      "reward_std": 0.34064918756484985,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.6676726937294006,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.12062937021255493,
      "step": 16
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5512839059674502,
      "calib/avg_num_step_conf": 5.56640625,
      "calib/ece": 0.1828740157480315,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.25196850393700787,
      "calib/gap": 0.0065439421338154835,
      "calib/mean_conf": 0.8718503937007873,
      "calib/mu_c": 0.8738857142857144,
      "calib/mu_w": 0.8673417721518989,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.1828740157480315,
      "calib/std_conf": 0.04479306378042393,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7513369477911647,
      "calib/step_q_c_n": 996.0,
      "calib/step_q_gap": -0.010970744516527575,
      "calib/step_q_w": 0.7623076923076922,
      "calib/step_q_w_n": 429.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2339.0,
      "completions/max_terminated_length": 2339.0,
      "completions/mean_length": 499.140625,
      "completions/mean_terminated_length": 499.140625,
      "completions/min_length": 83.0,
      "completions/min_terminated_length": 83.0,
      "epoch": 0.018133333333333335,
      "grad_norm": 0.006812790408730507,
      "kl": 0.003714442253112793,
      "learning_rate": 4.25e-06,
      "loss": 0.0366,
      "num_tokens": 4003063.0,
      "reward": 0.6302497982978821,
      "reward_std": 0.3418027460575104,
      "rewards/accuracy_reward_step": 0.68359375,
      "rewards/final_brier_reward_step": 0.7393969297409058,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.1875089406967163,
      "step": 17
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4798570705793069,
      "calib/avg_num_step_conf": 4.9921875,
      "calib/ece": 0.3447619047619048,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.2619047619047619,
      "calib/gap": -0.006670882873766715,
      "calib/mean_conf": 0.8685714285714285,
      "calib/mu_c": 0.8654477611940299,
      "calib/mu_w": 0.8721186440677966,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.34079365079365087,
      "calib/std_conf": 0.07573150448112766,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7421671826625388,
      "calib/step_q_c_n": 646.0,
      "calib/step_q_gap": 0.03927161304228555,
      "calib/step_q_w": 0.7028955696202532,
      "calib/step_q_w_n": 632.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 3067.0,
      "completions/max_terminated_length": 3067.0,
      "completions/mean_length": 477.2890625,
      "completions/mean_terminated_length": 479.16082763671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.0192,
      "grad_norm": 0.00782696157693863,
      "kl": 0.002428770065307617,
      "learning_rate": 4.5e-06,
      "loss": 0.0213,
      "num_tokens": 4235969.0,
      "reward": 0.5036895275115967,
      "reward_std": 0.3225781321525574,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.6186789274215698,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.0871376022696495,
      "step": 18
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5229340645643774,
      "calib/avg_num_step_conf": 4.29296875,
      "calib/ece": 0.3119215686274509,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.25882352941176473,
      "calib/gap": 0.010689268353483716,
      "calib/mean_conf": 0.8687843137254903,
      "calib/mu_c": 0.8735211267605634,
      "calib/mu_w": 0.8628318584070797,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3119215686274509,
      "calib/std_conf": 0.0641848781059075,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7364781906300485,
      "calib/step_q_c_n": 619.0,
      "calib/step_q_gap": 0.006269857296715031,
      "calib/step_q_w": 0.7302083333333335,
      "calib/step_q_w_n": 480.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1194.0,
      "completions/max_terminated_length": 1194.0,
      "completions/mean_length": 450.98828125,
      "completions/mean_terminated_length": 452.75689697265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 181.0,
      "epoch": 0.020266666666666665,
      "grad_norm": 0.007010597735643387,
      "kl": 0.003600597381591797,
      "learning_rate": 4.75e-06,
      "loss": 0.0047,
      "num_tokens": 4456182.0,
      "reward": 0.5692561864852905,
      "reward_std": 0.31815648078918457,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.6545273661613464,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.1730474829673767,
      "step": 19
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.3784722222222222,
      "calib/avg_num_step_conf": 4.7578125,
      "calib/ece": 0.2991699604743082,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.20948616600790515,
      "calib/gap": -0.02455147808358804,
      "calib/mean_conf": 0.8644664031620554,
      "calib/mu_c": 0.8538888888888889,
      "calib/mu_w": 0.878440366972477,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2972332015810276,
      "calib/std_conf": 0.05866273148199263,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7301494768310912,
      "calib/step_q_c_n": 669.0,
      "calib/step_q_gap": -0.004549976720821247,
      "calib/step_q_w": 0.7346994535519125,
      "calib/step_q_w_n": 549.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3061.0,
      "completions/max_terminated_length": 3061.0,
      "completions/mean_length": 450.96875,
      "completions/mean_terminated_length": 450.96875,
      "completions/min_length": 164.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.021333333333333333,
      "grad_norm": 0.007373226340860128,
      "kl": 0.0052947998046875,
      "learning_rate": 5e-06,
      "loss": 0.0243,
      "num_tokens": 4676502.0,
      "reward": 0.5320701599121094,
      "reward_std": 0.37527555227279663,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.6444605588912964,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.10952349007129669,
      "step": 20
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.43324428227015055,
      "calib/avg_num_step_conf": 5.33984375,
      "calib/ece": 0.27456349206349207,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.23809523809523808,
      "calib/gap": -0.011166351729979707,
      "calib/mean_conf": 0.8654365079365078,
      "calib/mu_c": 0.8608724832214764,
      "calib/mu_w": 0.8720388349514561,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.27436507936507937,
      "calib/std_conf": 0.05632431542806724,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.6966994382022472,
      "calib/step_q_c_n": 712.0,
      "calib/step_q_gap": -0.009270027446607765,
      "calib/step_q_w": 0.705969465648855,
      "calib/step_q_w_n": 655.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2947.0,
      "completions/max_terminated_length": 2947.0,
      "completions/mean_length": 480.9375,
      "completions/mean_terminated_length": 482.82354736328125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 33.0,
      "epoch": 0.0224,
      "grad_norm": 0.007370475679636002,
      "kl": 0.006046295166015625,
      "learning_rate": 4.9722222222222224e-06,
      "loss": -0.0003,
      "num_tokens": 4902582.0,
      "reward": 0.5867160558700562,
      "reward_std": 0.38447660207748413,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.6640527248382568,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.19609802961349487,
      "step": 21
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5082028085336213,
      "calib/avg_num_step_conf": 5.26953125,
      "calib/ece": 0.2203557312252964,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.1541501976284585,
      "calib/gap": 0.005776397515528053,
      "calib/mean_conf": 0.8567193675889329,
      "calib/mu_c": 0.8588198757763976,
      "calib/mu_w": 0.8530434782608696,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2203557312252964,
      "calib/std_conf": 0.05548838411002395,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7116451233842537,
      "calib/step_q_c_n": 851.0,
      "calib/step_q_gap": 0.022106970773811963,
      "calib/step_q_w": 0.6895381526104417,
      "calib/step_q_w_n": 498.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2967.0,
      "completions/max_terminated_length": 2967.0,
      "completions/mean_length": 452.34765625,
      "completions/mean_terminated_length": 452.34765625,
      "completions/min_length": 157.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.023466666666666667,
      "grad_norm": 0.007838176563382149,
      "kl": 0.008509159088134766,
      "learning_rate": 4.944444444444445e-06,
      "loss": 0.0488,
      "num_tokens": 5120199.0,
      "reward": 0.6486678123474121,
      "reward_std": 0.32256704568862915,
      "rewards/accuracy_reward_step": 0.62890625,
      "rewards/final_brier_reward_step": 0.7099843621253967,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.26469504833221436,
      "step": 22
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5278980576518619,
      "calib/avg_num_step_conf": 4.8046875,
      "calib/ece": 0.3182745098039215,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.17254901960784313,
      "calib/gap": 0.004140170728689885,
      "calib/mean_conf": 0.8555294117647059,
      "calib/mu_c": 0.8574452554744525,
      "calib/mu_w": 0.8533050847457626,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3182745098039215,
      "calib/std_conf": 0.05701405156735107,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.6932137518684603,
      "calib/step_q_c_n": 669.0,
      "calib/step_q_gap": 0.0032137518684602906,
      "calib/step_q_w": 0.6900000000000001,
      "calib/step_q_w_n": 561.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1771.0,
      "completions/max_terminated_length": 1771.0,
      "completions/mean_length": 461.6796875,
      "completions/mean_terminated_length": 461.6796875,
      "completions/min_length": 149.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.024533333333333334,
      "grad_norm": 0.008555812761187553,
      "kl": 0.009517669677734375,
      "learning_rate": 4.9166666666666665e-06,
      "loss": 0.0203,
      "num_tokens": 5342325.0,
      "reward": 0.5603914260864258,
      "reward_std": 0.3658100962638855,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.6425222754478455,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.17279182374477386,
      "step": 23
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.569568674219837,
      "calib/avg_num_step_conf": 5.328125,
      "calib/ece": 0.3778455284552845,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.14227642276422764,
      "calib/gap": 0.011920095408467613,
      "calib/mean_conf": 0.849390243902439,
      "calib/mu_c": 0.8556410256410256,
      "calib/mu_w": 0.843720930232558,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.3758130081300812,
      "calib/std_conf": 0.06699133958666083,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.6887786259541985,
      "calib/step_q_c_n": 655.0,
      "calib/step_q_gap": 0.026282152047287344,
      "calib/step_q_w": 0.6624964739069111,
      "calib/step_q_w_n": 709.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2857.0,
      "completions/max_terminated_length": 2857.0,
      "completions/mean_length": 557.6875,
      "completions/mean_terminated_length": 562.0787353515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 158.0,
      "epoch": 0.0256,
      "grad_norm": 0.006697987671941519,
      "kl": 0.007785320281982422,
      "learning_rate": 4.888888888888889e-06,
      "loss": 0.0085,
      "num_tokens": 5589605.0,
      "reward": 0.5254364013671875,
      "reward_std": 0.34020552039146423,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.5884214639663696,
      "rewards/format_reward_step": 0.9609375,
      "rewards/step_correlation_reward": 0.17807629704475403,
      "step": 24
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.489,
      "calib/avg_num_step_conf": 5.1640625,
      "calib/ece": 0.24513725490196078,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.14901960784313725,
      "calib/gap": -0.0033225806451614437,
      "calib/mean_conf": 0.8529803921568627,
      "calib/mu_c": 0.8516774193548386,
      "calib/mu_w": 0.8550000000000001,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.24513725490196078,
      "calib/std_conf": 0.057234807110221414,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.6781736909323116,
      "calib/step_q_c_n": 783.0,
      "calib/step_q_gap": -0.011993284948950023,
      "calib/step_q_w": 0.6901669758812616,
      "calib/step_q_w_n": 539.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1672.0,
      "completions/max_terminated_length": 1672.0,
      "completions/mean_length": 433.8203125,
      "completions/mean_terminated_length": 435.5216064453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.02666666666666667,
      "grad_norm": 0.007355384062975645,
      "kl": 0.012590408325195312,
      "learning_rate": 4.861111111111111e-06,
      "loss": 0.0228,
      "num_tokens": 5803887.0,
      "reward": 0.583240270614624,
      "reward_std": 0.32405388355255127,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.6939566135406494,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.15221142768859863,
      "step": 25
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.49590108401084015,
      "calib/avg_num_step_conf": 4.4609375,
      "calib/ece": 0.2056692913385827,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.10236220472440945,
      "calib/gap": 0.0048238482384825865,
      "calib/mean_conf": 0.8475590551181104,
      "calib/mu_c": 0.849268292682927,
      "calib/mu_w": 0.8444444444444444,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.20377952755905512,
      "calib/std_conf": 0.06647735432820169,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.6948640915593705,
      "calib/step_q_c_n": 699.0,
      "calib/step_q_gap": 0.03551871909887372,
      "calib/step_q_w": 0.6593453724604967,
      "calib/step_q_w_n": 443.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2930.0,
      "completions/max_terminated_length": 2930.0,
      "completions/mean_length": 453.71875,
      "completions/mean_terminated_length": 453.71875,
      "completions/min_length": 174.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.027733333333333332,
      "grad_norm": 0.007992461323738098,
      "kl": 0.012160301208496094,
      "learning_rate": 4.833333333333333e-06,
      "loss": 0.0226,
      "num_tokens": 6025279.0,
      "reward": 0.6579269170761108,
      "reward_std": 0.3294398784637451,
      "rewards/accuracy_reward_step": 0.640625,
      "rewards/final_brier_reward_step": 0.7225586175918579,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.26673269271850586,
      "step": 26
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.490538990825688,
      "calib/avg_num_step_conf": 5.29296875,
      "calib/ece": 0.2670355731225296,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.05928853754940711,
      "calib/gap": 0.00446929153924569,
      "calib/mean_conf": 0.8327272727272728,
      "calib/mu_c": 0.8346527777777778,
      "calib/mu_w": 0.8301834862385321,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.26529644268774705,
      "calib/std_conf": 0.07480193033531987,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.6901335311572699,
      "calib/step_q_c_n": 674.0,
      "calib/step_q_gap": 0.04452413321307025,
      "calib/step_q_w": 0.6456093979441997,
      "calib/step_q_w_n": 681.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2437.0,
      "completions/max_terminated_length": 2437.0,
      "completions/mean_length": 445.82421875,
      "completions/mean_terminated_length": 447.57257080078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.0288,
      "grad_norm": 0.007899126969277859,
      "kl": 0.01741313934326172,
      "learning_rate": 4.805555555555556e-06,
      "loss": -0.0038,
      "num_tokens": 6244626.0,
      "reward": 0.6128519177436829,
      "reward_std": 0.28570079803466797,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.673927366733551,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.24162021279335022,
      "step": 27
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4634847210541071,
      "calib/avg_num_step_conf": 4.00390625,
      "calib/ece": 0.18398406374501997,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.05179282868525897,
      "calib/gap": 0.0001976450798991669,
      "calib/mean_conf": 0.8373705179282868,
      "calib/mu_c": 0.8374390243902439,
      "calib/mu_w": 0.8372413793103447,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.18398406374501997,
      "calib/std_conf": 0.0569498465952437,
      "calib/step_conf_rate": 0.98046875,
      "calib/step_q_c": 0.6852743902439024,
      "calib/step_q_c_n": 656.0,
      "calib/step_q_gap": 0.009068428184281774,
      "calib/step_q_w": 0.6762059620596206,
      "calib/step_q_w_n": 369.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2803.0,
      "completions/max_terminated_length": 2803.0,
      "completions/mean_length": 498.1484375,
      "completions/mean_terminated_length": 500.10198974609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 137.0,
      "epoch": 0.029866666666666666,
      "grad_norm": 0.007141755428165197,
      "kl": 0.012704849243164062,
      "learning_rate": 4.777777777777778e-06,
      "loss": 0.0374,
      "num_tokens": 6479096.0,
      "reward": 0.6511706113815308,
      "reward_std": 0.30695879459381104,
      "rewards/accuracy_reward_step": 0.640625,
      "rewards/final_brier_reward_step": 0.7094171047210693,
      "rewards/format_reward_step": 0.96484375,
      "rewards/step_correlation_reward": 0.2718302011489868,
      "step": 28
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.3990751156105487,
      "calib/avg_num_step_conf": 4.64453125,
      "calib/ece": 0.3420948616600791,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.08695652173913043,
      "calib/gap": -0.02126546681664787,
      "calib/mean_conf": 0.8400395256916995,
      "calib/mu_c": 0.8294488188976378,
      "calib/mu_w": 0.8507142857142856,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.3400790513833992,
      "calib/std_conf": 0.05838063748347232,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.6675422138836773,
      "calib/step_q_c_n": 533.0,
      "calib/step_q_gap": -0.006482176360225078,
      "calib/step_q_w": 0.6740243902439024,
      "calib/step_q_w_n": 656.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2375.0,
      "completions/max_terminated_length": 2375.0,
      "completions/mean_length": 530.0546875,
      "completions/mean_terminated_length": 530.0546875,
      "completions/min_length": 162.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.030933333333333334,
      "grad_norm": 0.0068199350498616695,
      "kl": 0.014654159545898438,
      "learning_rate": 4.75e-06,
      "loss": -0.0016,
      "num_tokens": 6721918.0,
      "reward": 0.5134754180908203,
      "reward_std": 0.3352017402648926,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.6030503511428833,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.12936919927597046,
      "step": 29
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5166521954849914,
      "calib/avg_num_step_conf": 4.421875,
      "calib/ece": 0.295529411764706,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.07450980392156863,
      "calib/gap": 0.007873976680724315,
      "calib/mean_conf": 0.8367058823529411,
      "calib/mu_c": 0.8402877697841725,
      "calib/mu_w": 0.8324137931034482,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2935686274509805,
      "calib/std_conf": 0.07647716412148525,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.6702232854864434,
      "calib/step_q_c_n": 627.0,
      "calib/step_q_gap": 0.02962922608050278,
      "calib/step_q_w": 0.6405940594059406,
      "calib/step_q_w_n": 505.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2170.0,
      "completions/max_terminated_length": 2170.0,
      "completions/mean_length": 512.81640625,
      "completions/mean_terminated_length": 514.8274536132812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.032,
      "grad_norm": 0.0070888507179915905,
      "kl": 0.014923095703125,
      "learning_rate": 4.722222222222222e-06,
      "loss": 0.0171,
      "num_tokens": 6960183.0,
      "reward": 0.5999123454093933,
      "reward_std": 0.3764262795448303,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.6610507965087891,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.23174266517162323,
      "step": 30
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5482947394915682,
      "calib/avg_num_step_conf": 5.03125,
      "calib/ece": 0.3725296442687748,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.05533596837944664,
      "calib/gap": 0.014529952177195882,
      "calib/mean_conf": 0.8254940711462451,
      "calib/mu_c": 0.8333620689655171,
      "calib/mu_w": 0.8188321167883212,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3697628458498024,
      "calib/std_conf": 0.09492379388062085,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.679337899543379,
      "calib/step_q_c_n": 438.0,
      "calib/step_q_gap": 0.07640848777867304,
      "calib/step_q_w": 0.6029294117647059,
      "calib/step_q_w_n": 850.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2686.0,
      "completions/max_terminated_length": 2686.0,
      "completions/mean_length": 518.25,
      "completions/mean_terminated_length": 520.2824096679688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.03306666666666667,
      "grad_norm": 0.0068392446264624596,
      "kl": 0.017084121704101562,
      "learning_rate": 4.694444444444445e-06,
      "loss": 0.0333,
      "num_tokens": 7198767.0,
      "reward": 0.5243625640869141,
      "reward_std": 0.30371779203414917,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6080308556556702,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.15241290628910065,
      "step": 31
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.48380566801619435,
      "calib/avg_num_step_conf": 3.69140625,
      "calib/ece": 0.29552,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.044,
      "calib/gap": 0.007709016130068913,
      "calib/mean_conf": 0.82752,
      "calib/mu_c": 0.8311278195488723,
      "calib/mu_w": 0.8234188034188034,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.29552,
      "calib/std_conf": 0.07800672791496897,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.684712389380531,
      "calib/step_q_c_n": 452.0,
      "calib/step_q_gap": 0.03771441777809692,
      "calib/step_q_w": 0.6469979716024341,
      "calib/step_q_w_n": 493.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2279.0,
      "completions/max_terminated_length": 2279.0,
      "completions/mean_length": 433.984375,
      "completions/mean_terminated_length": 440.873046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.034133333333333335,
      "grad_norm": 0.009558629244565964,
      "kl": 0.040973663330078125,
      "learning_rate": 4.666666666666667e-06,
      "loss": 0.0403,
      "num_tokens": 7416571.0,
      "reward": 0.5686406493186951,
      "reward_std": 0.33148401975631714,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.6421367526054382,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.19670704007148743,
      "step": 32
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.48044094488188976,
      "calib/avg_num_step_conf": 3.984375,
      "calib/ece": 0.35210317460317464,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.05952380952380952,
      "calib/gap": -0.005273070866141594,
      "calib/mean_conf": 0.8428174603174603,
      "calib/mu_c": 0.84016,
      "calib/mu_w": 0.8454330708661416,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.34944444444444445,
      "calib/std_conf": 0.05107182940358978,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7085355648535564,
      "calib/step_q_c_n": 478.0,
      "calib/step_q_gap": 0.009974679244700302,
      "calib/step_q_w": 0.6985608856088561,
      "calib/step_q_w_n": 542.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2483.0,
      "completions/max_terminated_length": 2483.0,
      "completions/mean_length": 466.28125,
      "completions/mean_terminated_length": 468.1098327636719,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 120.0,
      "epoch": 0.0352,
      "grad_norm": 0.007306600920855999,
      "kl": 0.019073486328125,
      "learning_rate": 4.638888888888889e-06,
      "loss": 0.0535,
      "num_tokens": 7642811.0,
      "reward": 0.5501465797424316,
      "reward_std": 0.3264395594596863,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.6147527694702148,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.19100919365882874,
      "step": 33
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.429690027826312,
      "calib/avg_num_step_conf": 4.3203125,
      "calib/ece": 0.2474803149606299,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.07086614173228346,
      "calib/gap": -0.01012618908949714,
      "calib/mean_conf": 0.8367716535433071,
      "calib/mu_c": 0.8327450980392157,
      "calib/mu_w": 0.8428712871287128,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.24094488188976376,
      "calib/std_conf": 0.08078241300924664,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.6851067073170731,
      "calib/step_q_c_n": 656.0,
      "calib/step_q_gap": -0.019315514905149156,
      "calib/step_q_w": 0.7044222222222223,
      "calib/step_q_w_n": 450.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1928.0,
      "completions/max_terminated_length": 1928.0,
      "completions/mean_length": 422.76953125,
      "completions/mean_terminated_length": 424.427490234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 136.0,
      "epoch": 0.03626666666666667,
      "grad_norm": 0.09493603557348251,
      "kl": 0.18951797485351562,
      "learning_rate": 4.611111111111112e-06,
      "loss": 0.0335,
      "num_tokens": 7856152.0,
      "reward": 0.672420859336853,
      "reward_std": 0.3296073079109192,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.6849804520606995,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.34267374873161316,
      "step": 34
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4924242424242424,
      "calib/avg_num_step_conf": 3.703125,
      "calib/ece": 0.32976190476190465,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.05952380952380952,
      "calib/gap": 0.000606060606060721,
      "calib/mean_conf": 0.8453174603174602,
      "calib/mu_c": 0.8456060606060607,
      "calib/mu_w": 0.845,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.3256349206349205,
      "calib/std_conf": 0.05436044471033665,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7239529914529914,
      "calib/step_q_c_n": 468.0,
      "calib/step_q_gap": 0.02534882478632483,
      "calib/step_q_w": 0.6986041666666666,
      "calib/step_q_w_n": 480.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2316.0,
      "completions/max_terminated_length": 2316.0,
      "completions/mean_length": 506.8046875,
      "completions/mean_terminated_length": 508.79217529296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 171.0,
      "epoch": 0.037333333333333336,
      "grad_norm": 0.0072205690667033195,
      "kl": 0.01589202880859375,
      "learning_rate": 4.583333333333333e-06,
      "loss": 0.0687,
      "num_tokens": 8095150.0,
      "reward": 0.5845861434936523,
      "reward_std": 0.3286598026752472,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.634475827217102,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.23469656705856323,
      "step": 35
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5039518900343642,
      "calib/avg_num_step_conf": 3.546875,
      "calib/ece": 0.07807086614173235,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.06299212598425197,
      "calib/gap": 0.0015498281786939794,
      "calib/mean_conf": 0.8418503937007873,
      "calib/mu_c": 0.8422164948453608,
      "calib/mu_w": 0.8406666666666668,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.07807086614173235,
      "calib/std_conf": 0.049780238108552796,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7303614457831324,
      "calib/step_q_c_n": 664.0,
      "calib/step_q_gap": 0.056386035947066815,
      "calib/step_q_w": 0.6739754098360656,
      "calib/step_q_w_n": 244.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2283.0,
      "completions/max_terminated_length": 2283.0,
      "completions/mean_length": 439.39453125,
      "completions/mean_terminated_length": 439.39453125,
      "completions/min_length": 137.0,
      "completions/min_terminated_length": 137.0,
      "epoch": 0.0384,
      "grad_norm": 0.007099926937371492,
      "kl": 0.022426605224609375,
      "learning_rate": 4.555555555555556e-06,
      "loss": -0.0302,
      "num_tokens": 8310347.0,
      "reward": 0.7966442108154297,
      "reward_std": 0.2431754469871521,
      "rewards/accuracy_reward_step": 0.7578125,
      "rewards/final_brier_reward_step": 0.8040753602981567,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.439994215965271,
      "step": 36
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.497625643055006,
      "calib/avg_num_step_conf": 3.6875,
      "calib/ece": 0.3941295546558704,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0931174089068826,
      "calib/gap": -0.007568922305764447,
      "calib/mean_conf": 0.844251012145749,
      "calib/mu_c": 0.8401754385964912,
      "calib/mu_w": 0.8477443609022557,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.3884210526315789,
      "calib/std_conf": 0.06419972748460089,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7431695331695333,
      "calib/step_q_c_n": 407.0,
      "calib/step_q_gap": 0.05086040840230799,
      "calib/step_q_w": 0.6923091247672253,
      "calib/step_q_w_n": 537.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2723.0,
      "completions/max_terminated_length": 2723.0,
      "completions/mean_length": 489.390625,
      "completions/mean_terminated_length": 493.24407958984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 131.0,
      "epoch": 0.039466666666666664,
      "grad_norm": 0.009426533244550228,
      "kl": 0.030477523803710938,
      "learning_rate": 4.527777777777778e-06,
      "loss": 0.0313,
      "num_tokens": 8542727.0,
      "reward": 0.4761776626110077,
      "reward_std": 0.3279336094856262,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.5761339664459229,
      "rewards/format_reward_step": 0.96484375,
      "rewards/step_correlation_reward": 0.0941900834441185,
      "step": 37
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.45890324214792294,
      "calib/avg_num_step_conf": 3.52734375,
      "calib/ece": 0.2967588932806325,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.10276679841897234,
      "calib/gap": -0.00219478216818636,
      "calib/mean_conf": 0.8432411067193675,
      "calib/mu_c": 0.8422695035460993,
      "calib/mu_w": 0.8444642857142857,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2913438735177866,
      "calib/std_conf": 0.07007455917688288,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7394324853228962,
      "calib/step_q_c_n": 511.0,
      "calib/step_q_gap": 1.9220016773791038e-05,
      "calib/step_q_w": 0.7394132653061224,
      "calib/step_q_w_n": 392.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2769.0,
      "completions/max_terminated_length": 2769.0,
      "completions/mean_length": 449.6328125,
      "completions/mean_terminated_length": 451.3961181640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 133.0,
      "epoch": 0.04053333333333333,
      "grad_norm": 0.007991362363100052,
      "kl": 0.023359298706054688,
      "learning_rate": 4.5e-06,
      "loss": 0.0814,
      "num_tokens": 8764721.0,
      "reward": 0.5837390422821045,
      "reward_std": 0.35104483366012573,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.6577367186546326,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.20192888379096985,
      "step": 38
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.3888250319284803,
      "calib/avg_num_step_conf": 3.73828125,
      "calib/ece": 0.28853754940711457,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.1383399209486166,
      "calib/gap": -0.023364623243933624,
      "calib/mean_conf": 0.8511462450592886,
      "calib/mu_c": 0.8411724137931035,
      "calib/mu_w": 0.8645370370370371,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.28328063241106716,
      "calib/std_conf": 0.06406418437084797,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7339884393063584,
      "calib/step_q_c_n": 519.0,
      "calib/step_q_gap": -0.00384261092195215,
      "calib/step_q_w": 0.7378310502283105,
      "calib/step_q_w_n": 438.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2717.0,
      "completions/max_terminated_length": 2717.0,
      "completions/mean_length": 462.96484375,
      "completions/mean_terminated_length": 462.96484375,
      "completions/min_length": 147.0,
      "completions/min_terminated_length": 147.0,
      "epoch": 0.0416,
      "grad_norm": 0.006961394567042589,
      "kl": 0.02230072021484375,
      "learning_rate": 4.472222222222223e-06,
      "loss": -0.0099,
      "num_tokens": 8989328.0,
      "reward": 0.5831207036972046,
      "reward_std": 0.32755035161972046,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.6547492742538452,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.1997734010219574,
      "step": 39
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5232810465620931,
      "calib/avg_num_step_conf": 3.46875,
      "calib/ece": 0.35177165354330703,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.1062992125984252,
      "calib/gap": 0.01598425196850395,
      "calib/mean_conf": 0.8517716535433071,
      "calib/mu_c": 0.8597637795275591,
      "calib/mu_w": 0.8437795275590552,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.35177165354330703,
      "calib/std_conf": 0.0795979749847648,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7475586854460095,
      "calib/step_q_c_n": 426.0,
      "calib/step_q_gap": 0.0009136638009876741,
      "calib/step_q_w": 0.7466450216450218,
      "calib/step_q_w_n": 462.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1453.0,
      "completions/max_terminated_length": 1453.0,
      "completions/mean_length": 469.97265625,
      "completions/mean_terminated_length": 469.97265625,
      "completions/min_length": 142.0,
      "completions/min_terminated_length": 142.0,
      "epoch": 0.042666666666666665,
      "grad_norm": 0.008221627213060856,
      "kl": 0.03347587585449219,
      "learning_rate": 4.444444444444444e-06,
      "loss": 0.0195,
      "num_tokens": 9216401.0,
      "reward": 0.5157410502433777,
      "reward_std": 0.3499770760536194,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.6230074167251587,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.11081837117671967,
      "step": 40
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5269940314704287,
      "calib/avg_num_step_conf": 3.5625,
      "calib/ece": 0.08657370517928295,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.0796812749003984,
      "calib/gap": 0.003040332790739786,
      "calib/mean_conf": 0.8432270916334661,
      "calib/mu_c": 0.8439175257731959,
      "calib/mu_w": 0.8408771929824561,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.07844621513944232,
      "calib/std_conf": 0.0825462819782565,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7594736842105264,
      "calib/step_q_c_n": 722.0,
      "calib/step_q_gap": 0.006631578947368433,
      "calib/step_q_w": 0.752842105263158,
      "calib/step_q_w_n": 190.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2519.0,
      "completions/max_terminated_length": 2519.0,
      "completions/mean_length": 450.29296875,
      "completions/mean_terminated_length": 450.29296875,
      "completions/min_length": 130.0,
      "completions/min_terminated_length": 130.0,
      "epoch": 0.04373333333333333,
      "grad_norm": 0.0071234856732189655,
      "kl": 0.02417755126953125,
      "learning_rate": 4.416666666666667e-06,
      "loss": 0.0529,
      "num_tokens": 9438924.0,
      "reward": 0.7837193012237549,
      "reward_std": 0.34599435329437256,
      "rewards/accuracy_reward_step": 0.7578125,
      "rewards/final_brier_reward_step": 0.7970812320709229,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.4234822392463684,
      "step": 41
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5055284857571214,
      "calib/avg_num_step_conf": 3.77734375,
      "calib/ece": 0.31543307086614175,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.1220472440944882,
      "calib/gap": -0.001366816591704123,
      "calib/mean_conf": 0.858740157480315,
      "calib/mu_c": 0.8581159420289856,
      "calib/mu_w": 0.8594827586206897,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.31543307086614175,
      "calib/std_conf": 0.04721357281802466,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7388235294117648,
      "calib/step_q_c_n": 493.0,
      "calib/step_q_gap": -0.02081782079920569,
      "calib/step_q_w": 0.7596413502109705,
      "calib/step_q_w_n": 474.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2729.0,
      "completions/max_terminated_length": 2729.0,
      "completions/mean_length": 416.92578125,
      "completions/mean_terminated_length": 416.92578125,
      "completions/min_length": 134.0,
      "completions/min_terminated_length": 134.0,
      "epoch": 0.0448,
      "grad_norm": 0.007271526847034693,
      "kl": 0.020999908447265625,
      "learning_rate": 4.388888888888889e-06,
      "loss": 0.0215,
      "num_tokens": 9650025.0,
      "reward": 0.5809261798858643,
      "reward_std": 0.33942875266075134,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.6443961262702942,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.21120622754096985,
      "step": 42
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5182995495495495,
      "calib/avg_num_step_conf": 3.4375,
      "calib/ece": 0.29317647058823526,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.19607843137254902,
      "calib/gap": 0.00631193693693699,
      "calib/mean_conf": 0.8499607843137256,
      "calib/mu_c": 0.8527083333333333,
      "calib/mu_w": 0.8463963963963963,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2892156862745098,
      "calib/std_conf": 0.08366364984428545,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7397265625,
      "calib/step_q_c_n": 512.0,
      "calib/step_q_gap": -0.005028872282608621,
      "calib/step_q_w": 0.7447554347826086,
      "calib/step_q_w_n": 368.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1084.0,
      "completions/max_terminated_length": 1084.0,
      "completions/mean_length": 472.7890625,
      "completions/mean_terminated_length": 474.6431579589844,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 108.0,
      "epoch": 0.04586666666666667,
      "grad_norm": 0.006691481452435255,
      "kl": 0.022130966186523438,
      "learning_rate": 4.361111111111112e-06,
      "loss": -0.0062,
      "num_tokens": 9876283.0,
      "reward": 0.5755380988121033,
      "reward_std": 0.38558706641197205,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.6614230871200562,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.17949683964252472,
      "step": 43
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5408879969898408,
      "calib/avg_num_step_conf": 4.0234375,
      "calib/ece": 0.3985770750988141,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.22529644268774704,
      "calib/gap": 0.012836448012040536,
      "calib/mean_conf": 0.8638735177865613,
      "calib/mu_c": 0.8706722689075629,
      "calib/mu_w": 0.8578358208955223,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.39604743083003946,
      "calib/std_conf": 0.07086155160819461,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7696687370600414,
      "calib/step_q_c_n": 483.0,
      "calib/step_q_gap": 0.012155025908304529,
      "calib/step_q_w": 0.7575137111517368,
      "calib/step_q_w_n": 547.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1478.0,
      "completions/max_terminated_length": 1478.0,
      "completions/mean_length": 499.76953125,
      "completions/mean_terminated_length": 505.6956787109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 181.0,
      "epoch": 0.046933333333333334,
      "grad_norm": 0.00664907181635499,
      "kl": 0.019491195678710938,
      "learning_rate": 4.333333333333334e-06,
      "loss": -0.0114,
      "num_tokens": 10110544.0,
      "reward": 0.4891369342803955,
      "reward_std": 0.327079176902771,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.5903961062431335,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.09725279361009598,
      "step": 44
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5010208113804004,
      "calib/avg_num_step_conf": 3.9453125,
      "calib/ece": 0.27428,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.196,
      "calib/gap": 0.007297154899894465,
      "calib/mean_conf": 0.8558000000000001,
      "calib/mu_c": 0.858835616438356,
      "calib/mu_w": 0.8515384615384616,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.27304,
      "calib/std_conf": 0.07442822045434111,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7594680851063831,
      "calib/step_q_c_n": 564.0,
      "calib/step_q_gap": -0.0038951435931684486,
      "calib/step_q_w": 0.7633632286995515,
      "calib/step_q_w_n": 446.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2729.0,
      "completions/max_terminated_length": 2729.0,
      "completions/mean_length": 502.17578125,
      "completions/mean_terminated_length": 506.1299133300781,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 153.0,
      "epoch": 0.048,
      "grad_norm": 0.0068217390216887,
      "kl": 0.022533416748046875,
      "learning_rate": 4.305555555555556e-06,
      "loss": 0.0349,
      "num_tokens": 10344149.0,
      "reward": 0.6221271753311157,
      "reward_std": 0.4009205996990204,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.6636918187141418,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.271968811750412,
      "step": 45
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.48097999235376576,
      "calib/avg_num_step_conf": 4.1953125,
      "calib/ece": 0.3437051792828685,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.1952191235059761,
      "calib/gap": -0.01948515356187075,
      "calib/mean_conf": 0.845776892430279,
      "calib/mu_c": 0.8366165413533835,
      "calib/mu_w": 0.8561016949152542,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.329800796812749,
      "calib/std_conf": 0.08657547443019373,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7438319327731092,
      "calib/step_q_c_n": 595.0,
      "calib/step_q_gap": -0.014038630901212268,
      "calib/step_q_w": 0.7578705636743215,
      "calib/step_q_w_n": 479.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2426.0,
      "completions/max_terminated_length": 2426.0,
      "completions/mean_length": 548.8671875,
      "completions/mean_terminated_length": 548.8671875,
      "completions/min_length": 158.0,
      "completions/min_terminated_length": 158.0,
      "epoch": 0.04906666666666667,
      "grad_norm": 0.006014056038111448,
      "kl": 0.023838043212890625,
      "learning_rate": 4.277777777777778e-06,
      "loss": 0.0181,
      "num_tokens": 10589427.0,
      "reward": 0.529619038105011,
      "reward_std": 0.39823025465011597,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.6215183734893799,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.13771970570087433,
      "step": 46
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4879260429003816,
      "calib/avg_num_step_conf": 3.88671875,
      "calib/ece": 0.2801195219123506,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.27091633466135456,
      "calib/gap": -0.003914330833004409,
      "calib/mean_conf": 0.865617529880478,
      "calib/mu_c": 0.8640268456375838,
      "calib/mu_w": 0.8679411764705882,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.27605577689243027,
      "calib/std_conf": 0.0657406252801508,
      "calib/step_conf_rate": 0.98046875,
      "calib/step_q_c": 0.75665568369028,
      "calib/step_q_c_n": 607.0,
      "calib/step_q_gap": -0.009272151361266356,
      "calib/step_q_w": 0.7659278350515464,
      "calib/step_q_w_n": 388.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2340.0,
      "completions/max_terminated_length": 2340.0,
      "completions/mean_length": 553.3203125,
      "completions/mean_terminated_length": 555.490234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 179.0,
      "epoch": 0.050133333333333335,
      "grad_norm": 0.006215677130967379,
      "kl": 0.0251312255859375,
      "learning_rate": 4.25e-06,
      "loss": -0.0096,
      "num_tokens": 10837053.0,
      "reward": 0.6104649305343628,
      "reward_std": 0.33931007981300354,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.6642383337020874,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.24419158697128296,
      "step": 47
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.460375,
      "calib/avg_num_step_conf": 3.89453125,
      "calib/ece": 0.36201581027667984,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.23715415019762845,
      "calib/gap": -0.012888124999999917,
      "calib/mean_conf": 0.8600395256916995,
      "calib/mu_c": 0.8536718750000001,
      "calib/mu_w": 0.86656,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.3580632411067194,
      "calib/std_conf": 0.08404072993211129,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.757972166998012,
      "calib/step_q_c_n": 503.0,
      "calib/step_q_gap": -0.0043962540546196305,
      "calib/step_q_w": 0.7623684210526316,
      "calib/step_q_w_n": 494.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2845.0,
      "completions/max_terminated_length": 2845.0,
      "completions/mean_length": 506.6171875,
      "completions/mean_terminated_length": 506.6171875,
      "completions/min_length": 159.0,
      "completions/min_terminated_length": 159.0,
      "epoch": 0.0512,
      "grad_norm": 0.006817369256168604,
      "kl": 0.028110504150390625,
      "learning_rate": 4.222222222222223e-06,
      "loss": 0.0049,
      "num_tokens": 11070435.0,
      "reward": 0.5294307470321655,
      "reward_std": 0.3830631971359253,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.6033015847206116,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.1586848795413971,
      "step": 48
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.4975291695264242,
      "calib/avg_num_step_conf": 3.94921875,
      "calib/ece": 0.2572289156626505,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.2931726907630522,
      "calib/gap": 0.0009238160603980416,
      "calib/mean_conf": 0.8724899598393574,
      "calib/mu_c": 0.8728387096774193,
      "calib/mu_w": 0.8719148936170212,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.2536144578313252,
      "calib/std_conf": 0.06470371847379203,
      "calib/step_conf_rate": 0.96484375,
      "calib/step_q_c": 0.7685365853658537,
      "calib/step_q_c_n": 656.0,
      "calib/step_q_gap": -0.0008436963242871576,
      "calib/step_q_w": 0.7693802816901408,
      "calib/step_q_w_n": 355.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2730.0,
      "completions/max_terminated_length": 2730.0,
      "completions/mean_length": 562.52734375,
      "completions/mean_terminated_length": 564.7333374023438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 185.0,
      "epoch": 0.05226666666666667,
      "grad_norm": 0.006192709319293499,
      "kl": 0.030986785888671875,
      "learning_rate": 4.194444444444445e-06,
      "loss": 0.0284,
      "num_tokens": 11318978.0,
      "reward": 0.6265689134597778,
      "reward_std": 0.34821343421936035,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.6752644777297974,
      "rewards/format_reward_step": 0.95703125,
      "rewards/step_correlation_reward": 0.2653733491897583,
      "step": 49
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5368070175438596,
      "calib/avg_num_step_conf": 4.0234375,
      "calib/ece": 0.2666122448979591,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.35918367346938773,
      "calib/gap": 0.004743859649122828,
      "calib/mean_conf": 0.8746938775510205,
      "calib/mu_c": 0.8765333333333333,
      "calib/mu_w": 0.8717894736842104,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.2645306122448978,
      "calib/std_conf": 0.0647791236470115,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7780928689883914,
      "calib/step_q_c_n": 603.0,
      "calib/step_q_gap": 0.012519098496588166,
      "calib/step_q_w": 0.7655737704918032,
      "calib/step_q_w_n": 427.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2963.0,
      "completions/max_terminated_length": 2963.0,
      "completions/mean_length": 601.80078125,
      "completions/mean_terminated_length": 604.1608276367188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 152.0,
      "epoch": 0.05333333333333334,
      "grad_norm": 0.22164776921272278,
      "kl": 1.2020721435546875,
      "learning_rate": 4.166666666666667e-06,
      "loss": 0.0174,
      "num_tokens": 11578399.0,
      "reward": 0.5997195243835449,
      "reward_std": 0.3739992380142212,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.6526364088058472,
      "rewards/format_reward_step": 0.9453125,
      "rewards/step_correlation_reward": 0.24055272340774536,
      "step": 50
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.533546752034147,
      "calib/avg_num_step_conf": 4.1015625,
      "calib/ece": 0.27466135458167307,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.398406374501992,
      "calib/gap": 0.01503067893824206,
      "calib/mean_conf": 0.8717131474103587,
      "calib/mu_c": 0.8775816993464051,
      "calib/mu_w": 0.862551020408163,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.26840637450199184,
      "calib/std_conf": 0.09838057955065771,
      "calib/step_conf_rate": 0.97265625,
      "calib/step_q_c": 0.7840649149922719,
      "calib/step_q_c_n": 647.0,
      "calib/step_q_gap": 0.02436268174165157,
      "calib/step_q_w": 0.7597022332506204,
      "calib/step_q_w_n": 403.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2555.0,
      "completions/max_terminated_length": 2555.0,
      "completions/mean_length": 565.29296875,
      "completions/mean_terminated_length": 567.5098266601562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 198.0,
      "epoch": 0.0544,
      "grad_norm": 0.00638164347037673,
      "kl": 0.03388214111328125,
      "learning_rate": 4.138888888888889e-06,
      "loss": 0.0221,
      "num_tokens": 11832410.0,
      "reward": 0.6391332149505615,
      "reward_std": 0.3710130453109741,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.6637433767318726,
      "rewards/format_reward_step": 0.95703125,
      "rewards/step_correlation_reward": 0.3035855293273926,
      "step": 51
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4566133720930233,
      "calib/avg_num_step_conf": 4.16796875,
      "calib/ece": 0.20392857142857143,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.4087301587301587,
      "calib/gap": 0.008645348837209244,
      "calib/mean_conf": 0.8661507936507936,
      "calib/mu_c": 0.8688953488372092,
      "calib/mu_w": 0.86025,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.19376984126984126,
      "calib/std_conf": 0.12019769214832211,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7663157894736842,
      "calib/step_q_c_n": 722.0,
      "calib/step_q_gap": -0.016843630816170907,
      "calib/step_q_w": 0.7831594202898551,
      "calib/step_q_w_n": 345.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2614.0,
      "completions/max_terminated_length": 2614.0,
      "completions/mean_length": 549.65625,
      "completions/mean_terminated_length": 549.65625,
      "completions/min_length": 209.0,
      "completions/min_terminated_length": 209.0,
      "epoch": 0.055466666666666664,
      "grad_norm": 0.034107550978660583,
      "kl": 0.16524124145507812,
      "learning_rate": 4.111111111111111e-06,
      "loss": 0.0429,
      "num_tokens": 12081074.0,
      "reward": 0.7069775462150574,
      "reward_std": 0.37056323885917664,
      "rewards/accuracy_reward_step": 0.67578125,
      "rewards/final_brier_reward_step": 0.7231961488723755,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.3602902293205261,
      "step": 52
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.4809523809523809,
      "calib/avg_num_step_conf": 4.29296875,
      "calib/ece": 0.2893650793650794,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.38095238095238093,
      "calib/gap": -0.0012517006802721387,
      "calib/mean_conf": 0.8726984126984126,
      "calib/mu_c": 0.8721768707482992,
      "calib/mu_w": 0.8734285714285713,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2893650793650794,
      "calib/std_conf": 0.0869812304001554,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7642366612111293,
      "calib/step_q_c_n": 611.0,
      "calib/step_q_gap": -0.0051280928872314835,
      "calib/step_q_w": 0.7693647540983608,
      "calib/step_q_w_n": 488.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2203.0,
      "completions/max_terminated_length": 2203.0,
      "completions/mean_length": 588.453125,
      "completions/mean_terminated_length": 588.453125,
      "completions/min_length": 168.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.05653333333333333,
      "grad_norm": 0.004999403841793537,
      "kl": 0.0465087890625,
      "learning_rate": 4.083333333333334e-06,
      "loss": 0.0229,
      "num_tokens": 12337542.0,
      "reward": 0.612328052520752,
      "reward_std": 0.3517555594444275,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.6539046764373779,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.25981390476226807,
      "step": 53
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.513514500693988,
      "calib/avg_num_step_conf": 4.27734375,
      "calib/ece": 0.20884,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.496,
      "calib/gap": 0.012274819197896303,
      "calib/mean_conf": 0.88052,
      "calib/mu_c": 0.8844970414201184,
      "calib/mu_w": 0.8722222222222221,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.20668,
      "calib/std_conf": 0.09331307303909778,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7815868673050614,
      "calib/step_q_c_n": 731.0,
      "calib/step_q_gap": 0.015542911261105297,
      "calib/step_q_w": 0.7660439560439561,
      "calib/step_q_w_n": 364.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1841.0,
      "completions/max_terminated_length": 1841.0,
      "completions/mean_length": 509.8671875,
      "completions/mean_terminated_length": 513.8818969726562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 172.0,
      "epoch": 0.0576,
      "grad_norm": 0.0054353284649550915,
      "kl": 0.06253814697265625,
      "learning_rate": 4.055555555555556e-06,
      "loss": -0.0021,
      "num_tokens": 12574300.0,
      "reward": 0.6638885736465454,
      "reward_std": 0.40448057651519775,
      "rewards/accuracy_reward_step": 0.6640625,
      "rewards/final_brier_reward_step": 0.7185714840888977,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.28108060359954834,
      "step": 54
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.47657157609397177,
      "calib/avg_num_step_conf": 4.203125,
      "calib/ece": 0.3573493975903614,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.42168674698795183,
      "calib/gap": 0.019470117464825076,
      "calib/mean_conf": 0.8673895582329318,
      "calib/mu_c": 0.8769291338582677,
      "calib/mu_w": 0.8574590163934426,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.3573493975903614,
      "calib/std_conf": 0.11835219397493374,
      "calib/step_conf_rate": 0.984375,
      "calib/step_q_c": 0.7916758747697974,
      "calib/step_q_c_n": 543.0,
      "calib/step_q_gap": 0.0026890079780526044,
      "calib/step_q_w": 0.7889868667917448,
      "calib/step_q_w_n": 533.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2484.0,
      "completions/max_terminated_length": 2484.0,
      "completions/mean_length": 558.59765625,
      "completions/mean_terminated_length": 558.59765625,
      "completions/min_length": 158.0,
      "completions/min_terminated_length": 158.0,
      "epoch": 0.058666666666666666,
      "grad_norm": 0.005009907763451338,
      "kl": 0.06207275390625,
      "learning_rate": 4.027777777777779e-06,
      "loss": 0.0253,
      "num_tokens": 12825125.0,
      "reward": 0.511044979095459,
      "reward_std": 0.39973604679107666,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.597327709197998,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.13179343938827515,
      "step": 55
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.4796370967741936,
      "calib/avg_num_step_conf": 4.37109375,
      "calib/ece": 0.39790983606557373,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.9375,
      "calib/frac_conf_gt_0.9": 0.4918032786885246,
      "calib/gap": -0.00955376344086023,
      "calib/mean_conf": 0.8861885245901638,
      "calib/mu_c": 0.8813333333333333,
      "calib/mu_w": 0.8908870967741935,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.97265625,
      "calib/nonempty_step_conf_rate": 0.95703125,
      "calib/pce": 0.39614754098360655,
      "calib/std_conf": 0.07968573848246181,
      "calib/step_conf_rate": 0.95703125,
      "calib/step_q_c": 0.7779211469534051,
      "calib/step_q_c_n": 558.0,
      "calib/step_q_gap": -0.016517355720391702,
      "calib/step_q_w": 0.7944385026737968,
      "calib/step_q_w_n": 561.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2506.0,
      "completions/max_terminated_length": 2506.0,
      "completions/mean_length": 581.78125,
      "completions/mean_terminated_length": 586.3621826171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.05973333333333333,
      "grad_norm": 0.004738733638077974,
      "kl": 0.06552886962890625,
      "learning_rate": 4.000000000000001e-06,
      "loss": -0.0124,
      "num_tokens": 13080901.0,
      "reward": 0.49623072147369385,
      "reward_std": 0.3822718858718872,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.5537382960319519,
      "rewards/format_reward_step": 0.9375,
      "rewards/step_correlation_reward": 0.1574731469154358,
      "step": 56
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5611590972415715,
      "calib/avg_num_step_conf": 4.515625,
      "calib/ece": 0.279469387755102,
      "calib/final_conf_rate": 0.95703125,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.49795918367346936,
      "calib/gap": 0.040534271384786935,
      "calib/mean_conf": 0.8768571428571428,
      "calib/mu_c": 0.8929054054054054,
      "calib/mu_w": 0.8523711340206185,
      "calib/nonempty_final_conf_rate": 0.95703125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.27612244897959176,
      "calib/std_conf": 0.12209531238390132,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.781980056980057,
      "calib/step_q_c_n": 702.0,
      "calib/step_q_gap": 0.006032920416180421,
      "calib/step_q_w": 0.7759471365638766,
      "calib/step_q_w_n": 454.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2863.0,
      "completions/max_terminated_length": 2863.0,
      "completions/mean_length": 599.22265625,
      "completions/mean_terminated_length": 601.5725708007812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 262.0,
      "epoch": 0.0608,
      "grad_norm": 0.0041455053724348545,
      "kl": 0.066436767578125,
      "learning_rate": 3.972222222222223e-06,
      "loss": 0.0291,
      "num_tokens": 13341094.0,
      "reward": 0.5770140290260315,
      "reward_std": 0.38654983043670654,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.6612206697463989,
      "rewards/format_reward_step": 0.95703125,
      "rewards/step_correlation_reward": 0.18421359360218048,
      "step": 57
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.4436655783880673,
      "calib/avg_num_step_conf": 4.8203125,
      "calib/ece": 0.39148000000000005,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.504,
      "calib/gap": -0.005337046283848679,
      "calib/mean_conf": 0.87076,
      "calib/mu_c": 0.8680487804878048,
      "calib/mu_w": 0.8733858267716534,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.3851200000000001,
      "calib/std_conf": 0.12297569841232861,
      "calib/step_conf_rate": 0.984375,
      "calib/step_q_c": 0.7651885098743268,
      "calib/step_q_c_n": 557.0,
      "calib/step_q_gap": 0.028526766890574984,
      "calib/step_q_w": 0.7366617429837519,
      "calib/step_q_w_n": 677.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2777.0,
      "completions/max_terminated_length": 2777.0,
      "completions/mean_length": 648.44921875,
      "completions/mean_terminated_length": 648.44921875,
      "completions/min_length": 200.0,
      "completions/min_terminated_length": 200.0,
      "epoch": 0.06186666666666667,
      "grad_norm": 0.004322744905948639,
      "kl": 0.07526397705078125,
      "learning_rate": 3.944444444444445e-06,
      "loss": 0.001,
      "num_tokens": 13613417.0,
      "reward": 0.4562169909477234,
      "reward_std": 0.46282821893692017,
      "rewards/accuracy_reward_step": 0.48046875,
      "rewards/final_brier_reward_step": 0.5709011554718018,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.05168907344341278,
      "step": 58
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5150702844066688,
      "calib/avg_num_step_conf": 4.3046875,
      "calib/ece": 0.3415725806451612,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.46774193548387094,
      "calib/gap": 0.01627394573389984,
      "calib/mean_conf": 0.8668145161290324,
      "calib/mu_c": 0.874360902255639,
      "calib/mu_w": 0.8580869565217392,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.33604838709677415,
      "calib/std_conf": 0.12877185795889265,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7930398671096346,
      "calib/step_q_c_n": 602.0,
      "calib/step_q_gap": 0.01533986710963453,
      "calib/step_q_w": 0.7777000000000001,
      "calib/step_q_w_n": 500.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2536.0,
      "completions/max_terminated_length": 2536.0,
      "completions/mean_length": 580.48828125,
      "completions/mean_terminated_length": 582.7647705078125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 172.0,
      "epoch": 0.06293333333333333,
      "grad_norm": 0.004097119905054569,
      "kl": 0.078094482421875,
      "learning_rate": 3.916666666666667e-06,
      "loss": 0.0634,
      "num_tokens": 13868270.0,
      "reward": 0.5046218633651733,
      "reward_std": 0.38234949111938477,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.609442949295044,
      "rewards/format_reward_step": 0.95703125,
      "rewards/step_correlation_reward": 0.10448820888996124,
      "step": 59
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.43834448379902924,
      "calib/avg_num_step_conf": 4.22265625,
      "calib/ece": 0.3957085020242915,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.44534412955465585,
      "calib/gap": -0.021629935720844706,
      "calib/mean_conf": 0.8699595141700405,
      "calib/mu_c": 0.8589256198347107,
      "calib/mu_w": 0.8805555555555554,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.38789473684210524,
      "calib/std_conf": 0.11143705391107282,
      "calib/step_conf_rate": 0.97265625,
      "calib/step_q_c": 0.7760658914728683,
      "calib/step_q_c_n": 516.0,
      "calib/step_q_gap": 0.014685360499416933,
      "calib/step_q_w": 0.7613805309734514,
      "calib/step_q_w_n": 565.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2316.0,
      "completions/max_terminated_length": 2316.0,
      "completions/mean_length": 561.9765625,
      "completions/mean_terminated_length": 566.4015502929688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 235.0,
      "epoch": 0.064,
      "grad_norm": 0.004728895146399736,
      "kl": 0.0794830322265625,
      "learning_rate": 3.88888888888889e-06,
      "loss": 0.0085,
      "num_tokens": 14120992.0,
      "reward": 0.45893681049346924,
      "reward_std": 0.4327120780944824,
      "rewards/accuracy_reward_step": 0.4765625,
      "rewards/final_brier_reward_step": 0.5494043231010437,
      "rewards/format_reward_step": 0.94921875,
      "rewards/step_correlation_reward": 0.08331304788589478,
      "step": 60
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.45793163891323396,
      "calib/avg_num_step_conf": 4.95703125,
      "calib/ece": 0.2676377952755905,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.4881889763779528,
      "calib/gap": -0.01663183442324523,
      "calib/mean_conf": 0.8555905511811024,
      "calib/mu_c": 0.8496319018404909,
      "calib/mu_w": 0.8662637362637361,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.24074803149606291,
      "calib/std_conf": 0.15307428870805093,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7778206896551725,
      "calib/step_q_c_n": 725.0,
      "calib/step_q_gap": 0.089548630831643,
      "calib/step_q_w": 0.6882720588235295,
      "calib/step_q_w_n": 544.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2716.0,
      "completions/max_terminated_length": 2716.0,
      "completions/mean_length": 496.93359375,
      "completions/mean_terminated_length": 496.93359375,
      "completions/min_length": 136.0,
      "completions/min_terminated_length": 136.0,
      "epoch": 0.06506666666666666,
      "grad_norm": 0.0047287000343203545,
      "kl": 0.08477783203125,
      "learning_rate": 3.861111111111112e-06,
      "loss": 0.0132,
      "num_tokens": 14352271.0,
      "reward": 0.6241191625595093,
      "reward_std": 0.40424540638923645,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.6815304756164551,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.24327021837234497,
      "step": 61
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.49229752066115706,
      "calib/avg_num_step_conf": 4.12109375,
      "calib/ece": 0.3784959349593495,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.4349593495934959,
      "calib/gap": 0.0009712396694213643,
      "calib/mean_conf": 0.8573577235772356,
      "calib/mu_c": 0.8578512396694215,
      "calib/mu_w": 0.8568800000000001,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.37199186991869915,
      "calib/std_conf": 0.11819411002075521,
      "calib/step_conf_rate": 0.9765625,
      "calib/step_q_c": 0.7533976833976834,
      "calib/step_q_c_n": 518.0,
      "calib/step_q_gap": 0.005818540008484141,
      "calib/step_q_w": 0.7475791433891993,
      "calib/step_q_w_n": 537.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2563.0,
      "completions/max_terminated_length": 2563.0,
      "completions/mean_length": 572.79296875,
      "completions/mean_terminated_length": 579.5850219726562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 196.0,
      "epoch": 0.06613333333333334,
      "grad_norm": 0.004045703914016485,
      "kl": 0.08335113525390625,
      "learning_rate": 3.833333333333334e-06,
      "loss": 0.0174,
      "num_tokens": 14605986.0,
      "reward": 0.49485641717910767,
      "reward_std": 0.44042736291885376,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.571002721786499,
      "rewards/format_reward_step": 0.9453125,
      "rewards/step_correlation_reward": 0.1351163387298584,
      "step": 62
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5197957839262187,
      "calib/avg_num_step_conf": 4.35546875,
      "calib/ece": 0.3272874493927127,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9453125,
      "calib/frac_conf_gt_0.9": 0.4534412955465587,
      "calib/gap": 0.016361660079051377,
      "calib/mean_conf": 0.8617004048582996,
      "calib/mu_c": 0.8693181818181818,
      "calib/mu_w": 0.8529565217391304,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.96484375,
      "calib/pce": 0.3272874493927127,
      "calib/std_conf": 0.11042585637848189,
      "calib/step_conf_rate": 0.96484375,
      "calib/step_q_c": 0.7688871054039593,
      "calib/step_q_c_n": 623.0,
      "calib/step_q_gap": -0.003064114108235816,
      "calib/step_q_w": 0.7719512195121951,
      "calib/step_q_w_n": 492.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2289.0,
      "completions/max_terminated_length": 2289.0,
      "completions/mean_length": 584.5234375,
      "completions/mean_terminated_length": 586.8157348632812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 144.0,
      "epoch": 0.0672,
      "grad_norm": 0.004193742293864489,
      "kl": 0.0828399658203125,
      "learning_rate": 3.8055555555555556e-06,
      "loss": 0.025,
      "num_tokens": 14864264.0,
      "reward": 0.5371720790863037,
      "reward_std": 0.41586440801620483,
      "rewards/accuracy_reward_step": 0.515625,
      "rewards/final_brier_reward_step": 0.6058300733566284,
      "rewards/format_reward_step": 0.9453125,
      "rewards/step_correlation_reward": 0.17632651329040527,
      "step": 63
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.6001828277356446,
      "calib/avg_num_step_conf": 4.4453125,
      "calib/ece": 0.2853658536585365,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.4105691056910569,
      "calib/gap": 0.04562296858071502,
      "calib/mean_conf": 0.8447967479674796,
      "calib/mu_c": 0.8640845070422535,
      "calib/mu_w": 0.8184615384615385,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.27646341463414625,
      "calib/std_conf": 0.1395629041581844,
      "calib/step_conf_rate": 0.984375,
      "calib/step_q_c": 0.770624,
      "calib/step_q_c_n": 625.0,
      "calib/step_q_gap": 0.027953434697855717,
      "calib/step_q_w": 0.7426705653021443,
      "calib/step_q_w_n": 513.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2935.0,
      "completions/max_terminated_length": 2935.0,
      "completions/mean_length": 550.22265625,
      "completions/mean_terminated_length": 558.9563598632812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.06826666666666667,
      "grad_norm": 0.00438300147652626,
      "kl": 0.090789794921875,
      "learning_rate": 3.777777777777778e-06,
      "loss": 0.0226,
      "num_tokens": 15108897.0,
      "reward": 0.587658703327179,
      "reward_std": 0.45948439836502075,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6535687446594238,
      "rewards/format_reward_step": 0.953125,
      "rewards/step_correlation_reward": 0.2201860547065735,
      "step": 64
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5798196596393193,
      "calib/avg_num_step_conf": 3.9296875,
      "calib/ece": 0.3533864541832669,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.4063745019920319,
      "calib/gap": 0.05891605283210555,
      "calib/mean_conf": 0.8443027888446216,
      "calib/mu_c": 0.8741129032258064,
      "calib/mu_w": 0.8151968503937008,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.3518326693227091,
      "calib/std_conf": 0.1543622400316892,
      "calib/step_conf_rate": 0.984375,
      "calib/step_q_c": 0.7706299212598425,
      "calib/step_q_c_n": 508.0,
      "calib/step_q_gap": 0.032296587926509246,
      "calib/step_q_w": 0.7383333333333333,
      "calib/step_q_w_n": 498.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1564.0,
      "completions/max_terminated_length": 1564.0,
      "completions/mean_length": 457.5234375,
      "completions/mean_terminated_length": 459.31768798828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.06933333333333333,
      "grad_norm": 0.004640705417841673,
      "kl": 0.093994140625,
      "learning_rate": 3.7500000000000005e-06,
      "loss": 0.0027,
      "num_tokens": 15331047.0,
      "reward": 0.5712149143218994,
      "reward_std": 0.37106049060821533,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.6176152229309082,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.23418954014778137,
      "step": 65
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5258597883597884,
      "calib/avg_num_step_conf": 4.37890625,
      "calib/ece": 0.39056680161943313,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.3603238866396761,
      "calib/gap": 0.021560846560846558,
      "calib/mean_conf": 0.8360728744939272,
      "calib/mu_c": 0.847857142857143,
      "calib/mu_w": 0.8262962962962964,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.38659919028340073,
      "calib/std_conf": 0.15139955300073482,
      "calib/step_conf_rate": 0.9765625,
      "calib/step_q_c": 0.7475578947368421,
      "calib/step_q_c_n": 475.0,
      "calib/step_q_gap": 0.006187925696594432,
      "calib/step_q_w": 0.7413699690402477,
      "calib/step_q_w_n": 646.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2799.0,
      "completions/max_terminated_length": 2799.0,
      "completions/mean_length": 595.5625,
      "completions/mean_terminated_length": 597.8980712890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.0704,
      "grad_norm": 0.004484922159463167,
      "kl": 0.0873870849609375,
      "learning_rate": 3.7222222222222225e-06,
      "loss": -0.0164,
      "num_tokens": 15589863.0,
      "reward": 0.45835405588150024,
      "reward_std": 0.4077296555042267,
      "rewards/accuracy_reward_step": 0.4375,
      "rewards/final_brier_reward_step": 0.568659782409668,
      "rewards/format_reward_step": 0.95703125,
      "rewards/step_correlation_reward": 0.06914205849170685,
      "step": 66
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5516579183203175,
      "calib/avg_num_step_conf": 4.375,
      "calib/ece": 0.2704347826086956,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.33201581027667987,
      "calib/gap": 0.040174753552682296,
      "calib/mean_conf": 0.8425296442687746,
      "calib/mu_c": 0.8595205479452054,
      "calib/mu_w": 0.8193457943925231,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.2679446640316205,
      "calib/std_conf": 0.1380751503527436,
      "calib/step_conf_rate": 0.984375,
      "calib/step_q_c": 0.7760596923076923,
      "calib/step_q_c_n": 650.0,
      "calib/step_q_gap": 0.04142139443535198,
      "calib/step_q_w": 0.7346382978723404,
      "calib/step_q_w_n": 470.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2858.0,
      "completions/max_terminated_length": 2858.0,
      "completions/mean_length": 524.83203125,
      "completions/mean_terminated_length": 524.83203125,
      "completions/min_length": 186.0,
      "completions/min_terminated_length": 186.0,
      "epoch": 0.07146666666666666,
      "grad_norm": 0.004559084307402372,
      "kl": 0.0973663330078125,
      "learning_rate": 3.694444444444445e-06,
      "loss": 0.0053,
      "num_tokens": 15829228.0,
      "reward": 0.6144473552703857,
      "reward_std": 0.36117982864379883,
      "rewards/accuracy_reward_step": 0.5703125,
      "rewards/final_brier_reward_step": 0.669489860534668,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.25002986192703247,
      "step": 67
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5365545401516624,
      "calib/avg_num_step_conf": 4.265625,
      "calib/ece": 0.3240799999999999,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.348,
      "calib/gap": -0.0013040378508003547,
      "calib/mean_conf": 0.8343200000000001,
      "calib/mu_c": 0.8337410071942447,
      "calib/mu_w": 0.8350450450450451,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.3011999999999999,
      "calib/std_conf": 0.15953349993026544,
      "calib/step_conf_rate": 0.98046875,
      "calib/step_q_c": 0.7423938879456706,
      "calib/step_q_c_n": 589.0,
      "calib/step_q_gap": 0.0046801702518335375,
      "calib/step_q_w": 0.7377137176938371,
      "calib/step_q_w_n": 503.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2428.0,
      "completions/max_terminated_length": 2428.0,
      "completions/mean_length": 521.02734375,
      "completions/mean_terminated_length": 521.02734375,
      "completions/min_length": 54.0,
      "completions/min_terminated_length": 54.0,
      "epoch": 0.07253333333333334,
      "grad_norm": 0.004508777987211943,
      "kl": 0.10665130615234375,
      "learning_rate": 3.6666666666666666e-06,
      "loss": 0.0179,
      "num_tokens": 16066699.0,
      "reward": 0.5404424071311951,
      "reward_std": 0.3533206582069397,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.626226544380188,
      "rewards/format_reward_step": 0.96484375,
      "rewards/step_correlation_reward": 0.15231452882289886,
      "step": 68
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.4674674674674675,
      "calib/avg_num_step_conf": 3.95703125,
      "calib/ece": 0.3839430894308944,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.27235772357723576,
      "calib/gap": -0.005239239239239191,
      "calib/mean_conf": 0.8054878048780488,
      "calib/mu_c": 0.8026126126126126,
      "calib/mu_w": 0.8078518518518518,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.3691056910569107,
      "calib/std_conf": 0.18090885572276796,
      "calib/step_conf_rate": 0.97265625,
      "calib/step_q_c": 0.7343735224586289,
      "calib/step_q_c_n": 423.0,
      "calib/step_q_gap": -0.011355291100693221,
      "calib/step_q_w": 0.7457288135593221,
      "calib/step_q_w_n": 590.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2557.0,
      "completions/max_terminated_length": 2557.0,
      "completions/mean_length": 588.40234375,
      "completions/mean_terminated_length": 590.7098388671875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 181.0,
      "epoch": 0.0736,
      "grad_norm": 0.004546022973954678,
      "kl": 0.09192657470703125,
      "learning_rate": 3.638888888888889e-06,
      "loss": 0.0411,
      "num_tokens": 16321826.0,
      "reward": 0.4601125717163086,
      "reward_std": 0.37661075592041016,
      "rewards/accuracy_reward_step": 0.43359375,
      "rewards/final_brier_reward_step": 0.5603660345077515,
      "rewards/format_reward_step": 0.94921875,
      "rewards/step_correlation_reward": 0.08329666405916214,
      "step": 69
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5459973753280839,
      "calib/avg_num_step_conf": 4.26953125,
      "calib/ece": 0.34773279352226727,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.32793522267206476,
      "calib/gap": 0.046930446194225595,
      "calib/mean_conf": 0.8230364372469636,
      "calib/mu_c": 0.8471666666666666,
      "calib/mu_w": 0.800236220472441,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.3424696356275304,
      "calib/std_conf": 0.16960089698869982,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7481395348837209,
      "calib/step_q_c_n": 516.0,
      "calib/step_q_gap": 0.06268026278666716,
      "calib/step_q_w": 0.6854592720970537,
      "calib/step_q_w_n": 577.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2470.0,
      "completions/max_terminated_length": 2470.0,
      "completions/mean_length": 540.5078125,
      "completions/mean_terminated_length": 544.7637939453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 146.0,
      "epoch": 0.07466666666666667,
      "grad_norm": 0.00454904418438673,
      "kl": 0.085479736328125,
      "learning_rate": 3.6111111111111115e-06,
      "loss": 0.0238,
      "num_tokens": 16567188.0,
      "reward": 0.5172110795974731,
      "reward_std": 0.39061713218688965,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6089847683906555,
      "rewards/format_reward_step": 0.96484375,
      "rewards/step_correlation_reward": 0.1387186497449875,
      "step": 70
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.4943941778127458,
      "calib/avg_num_step_conf": 4.3515625,
      "calib/ece": 0.3520647773279352,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.27530364372469635,
      "calib/gap": -0.015126540781536746,
      "calib/mean_conf": 0.8280971659919029,
      "calib/mu_c": 0.8205645161290323,
      "calib/mu_w": 0.835691056910569,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.3390688259109311,
      "calib/std_conf": 0.13513834069416286,
      "calib/step_conf_rate": 0.98046875,
      "calib/step_q_c": 0.7431620553359685,
      "calib/step_q_c_n": 506.0,
      "calib/step_q_gap": -0.002956365716663134,
      "calib/step_q_w": 0.7461184210526316,
      "calib/step_q_w_n": 608.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 535.18359375,
      "completions/mean_terminated_length": 537.2824096679688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 175.0,
      "epoch": 0.07573333333333333,
      "grad_norm": 0.0041601029224693775,
      "kl": 0.08847808837890625,
      "learning_rate": 3.5833333333333335e-06,
      "loss": 0.0424,
      "num_tokens": 16808603.0,
      "reward": 0.509465217590332,
      "reward_std": 0.4146815538406372,
      "rewards/accuracy_reward_step": 0.484375,
      "rewards/final_brier_reward_step": 0.5898030996322632,
      "rewards/format_reward_step": 0.953125,
      "rewards/step_correlation_reward": 0.1416272521018982,
      "step": 71
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5870607532647784,
      "calib/avg_num_step_conf": 4.23828125,
      "calib/ece": 0.36829365079365084,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.2857142857142857,
      "calib/gap": 0.032716547851870614,
      "calib/mean_conf": 0.842579365079365,
      "calib/mu_c": 0.8595867768595041,
      "calib/mu_w": 0.8268702290076335,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.3653571428571428,
      "calib/std_conf": 0.11763920178419383,
      "calib/step_conf_rate": 0.9765625,
      "calib/step_q_c": 0.7598109640831759,
      "calib/step_q_c_n": 529.0,
      "calib/step_q_gap": 0.016303769838571625,
      "calib/step_q_w": 0.7435071942446043,
      "calib/step_q_w_n": 556.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2425.0,
      "completions/max_terminated_length": 2425.0,
      "completions/mean_length": 475.34375,
      "completions/mean_terminated_length": 475.34375,
      "completions/min_length": 166.0,
      "completions/min_terminated_length": 166.0,
      "epoch": 0.0768,
      "grad_norm": 0.004811810329556465,
      "kl": 0.10382080078125,
      "learning_rate": 3.555555555555556e-06,
      "loss": 0.0235,
      "num_tokens": 17034699.0,
      "reward": 0.5658941268920898,
      "reward_std": 0.3956160843372345,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.603821873664856,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.23890388011932373,
      "step": 72
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5032398793866684,
      "calib/avg_num_step_conf": 4.1484375,
      "calib/ece": 0.2762698412698412,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.2619047619047619,
      "calib/gap": 0.01088342849810775,
      "calib/mean_conf": 0.8174603174603174,
      "calib/mu_c": 0.8221678321678323,
      "calib/mu_w": 0.8112844036697245,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2631349206349206,
      "calib/std_conf": 0.1501028034378413,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.739360606060606,
      "calib/step_q_c_n": 594.0,
      "calib/step_q_gap": 0.003719580419580426,
      "calib/step_q_w": 0.7356410256410256,
      "calib/step_q_w_n": 468.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2466.0,
      "completions/max_terminated_length": 2466.0,
      "completions/mean_length": 502.109375,
      "completions/mean_terminated_length": 502.109375,
      "completions/min_length": 179.0,
      "completions/min_terminated_length": 179.0,
      "epoch": 0.07786666666666667,
      "grad_norm": 0.004434308968484402,
      "kl": 0.08699798583984375,
      "learning_rate": 3.5277777777777784e-06,
      "loss": -0.0117,
      "num_tokens": 17270271.0,
      "reward": 0.615314245223999,
      "reward_std": 0.418173611164093,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.6604995727539062,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.2623162865638733,
      "step": 73
    },
    {
      "calib/answer_extract_rate": 0.94921875,
      "calib/auroc": 0.5044928522804628,
      "calib/avg_num_step_conf": 4.1640625,
      "calib/ece": 0.3879012345679014,
      "calib/final_conf_rate": 0.94921875,
      "calib/format_rate": 0.94140625,
      "calib/frac_conf_gt_0.9": 0.27983539094650206,
      "calib/gap": 0.00554731109598372,
      "calib/mean_conf": 0.8300411522633746,
      "calib/mu_c": 0.8330088495575221,
      "calib/mu_w": 0.8274615384615384,
      "calib/nonempty_final_conf_rate": 0.94921875,
      "calib/nonempty_reasoning_rate": 0.984375,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.3764609053497944,
      "calib/std_conf": 0.13532919151475983,
      "calib/step_conf_rate": 0.9765625,
      "calib/step_q_c": 0.7249667405764968,
      "calib/step_q_c_n": 451.0,
      "calib/step_q_gap": -0.0083503325942349,
      "calib/step_q_w": 0.7333170731707317,
      "calib/step_q_w_n": 615.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2985.0,
      "completions/max_terminated_length": 2985.0,
      "completions/mean_length": 511.58203125,
      "completions/mean_terminated_length": 515.6102294921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 119.0,
      "epoch": 0.07893333333333333,
      "grad_norm": 0.004572301171720028,
      "kl": 0.1038818359375,
      "learning_rate": 3.5e-06,
      "loss": 0.0399,
      "num_tokens": 17505164.0,
      "reward": 0.4876174330711365,
      "reward_std": 0.41045689582824707,
      "rewards/accuracy_reward_step": 0.44140625,
      "rewards/final_brier_reward_step": 0.5689581632614136,
      "rewards/format_reward_step": 0.94140625,
      "rewards/step_correlation_reward": 0.12971413135528564,
      "step": 74
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5626369112814896,
      "calib/avg_num_step_conf": 4.1875,
      "calib/ece": 0.1933464566929133,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.25984251968503935,
      "calib/gap": 0.029557776560788396,
      "calib/mean_conf": 0.8139763779527559,
      "calib/mu_c": 0.8242168674698794,
      "calib/mu_w": 0.794659090909091,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.1768897637795275,
      "calib/std_conf": 0.15970301988412466,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7443496503496504,
      "calib/step_q_c_n": 715.0,
      "calib/step_q_gap": 0.0305401265401265,
      "calib/step_q_w": 0.7138095238095239,
      "calib/step_q_w_n": 357.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2550.0,
      "completions/max_terminated_length": 2550.0,
      "completions/mean_length": 455.59375,
      "completions/mean_terminated_length": 455.59375,
      "completions/min_length": 161.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.08,
      "grad_norm": 0.004388398490846157,
      "kl": 0.107330322265625,
      "learning_rate": 3.4722222222222224e-06,
      "loss": 0.018,
      "num_tokens": 17726548.0,
      "reward": 0.7035642266273499,
      "reward_std": 0.3689558506011963,
      "rewards/accuracy_reward_step": 0.6484375,
      "rewards/final_brier_reward_step": 0.7299691438674927,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.34903430938720703,
      "step": 75
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5625316937119675,
      "calib/avg_num_step_conf": 4.2421875,
      "calib/ece": 0.29103174603174603,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.2698412698412698,
      "calib/gap": 0.034467545638945385,
      "calib/mean_conf": 0.8072222222222222,
      "calib/mu_c": 0.8230882352941177,
      "calib/mu_w": 0.7886206896551723,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.27928571428571425,
      "calib/std_conf": 0.1687965756483741,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.6995469798657719,
      "calib/step_q_c_n": 596.0,
      "calib/step_q_gap": -0.00010608135871792435,
      "calib/step_q_w": 0.6996530612244898,
      "calib/step_q_w_n": 490.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1422.0,
      "completions/max_terminated_length": 1422.0,
      "completions/mean_length": 459.72265625,
      "completions/mean_terminated_length": 463.342529296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.08106666666666666,
      "grad_norm": 0.0049783228896558285,
      "kl": 0.1104736328125,
      "learning_rate": 3.444444444444445e-06,
      "loss": 0.0207,
      "num_tokens": 17947293.0,
      "reward": 0.5709596276283264,
      "reward_std": 0.4218650758266449,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6571656465530396,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.18240979313850403,
      "step": 76
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.383743842364532,
      "calib/avg_num_step_conf": 4.734375,
      "calib/ece": 0.2728800000000001,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.252,
      "calib/gap": -0.04243021346469622,
      "calib/mean_conf": 0.8051999999999999,
      "calib/mu_c": 0.7873793103448276,
      "calib/mu_w": 0.8298095238095238,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.2490400000000001,
      "calib/std_conf": 0.14694815412246592,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7146539027982327,
      "calib/step_q_c_n": 679.0,
      "calib/step_q_gap": -0.026359230410022527,
      "calib/step_q_w": 0.7410131332082552,
      "calib/step_q_w_n": 533.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2415.0,
      "completions/max_terminated_length": 2415.0,
      "completions/mean_length": 515.83203125,
      "completions/mean_terminated_length": 515.83203125,
      "completions/min_length": 145.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.08213333333333334,
      "grad_norm": 0.004251775331795216,
      "kl": 0.0996856689453125,
      "learning_rate": 3.416666666666667e-06,
      "loss": 0.0346,
      "num_tokens": 18184010.0,
      "reward": 0.5908552408218384,
      "reward_std": 0.39187467098236084,
      "rewards/accuracy_reward_step": 0.56640625,
      "rewards/final_brier_reward_step": 0.6375484466552734,
      "rewards/format_reward_step": 0.9609375,
      "rewards/step_correlation_reward": 0.23869329690933228,
      "step": 77
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.524085637823372,
      "calib/avg_num_step_conf": 4.4765625,
      "calib/ece": 0.30290836653386455,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.21115537848605578,
      "calib/gap": 0.008916146297948457,
      "calib/mean_conf": 0.8063346613545816,
      "calib/mu_c": 0.8105263157894738,
      "calib/mu_w": 0.8016101694915253,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.2896812749003984,
      "calib/std_conf": 0.1452683721736049,
      "calib/step_conf_rate": 0.98046875,
      "calib/step_q_c": 0.71395,
      "calib/step_q_c_n": 600.0,
      "calib/step_q_gap": -0.004474908424908386,
      "calib/step_q_w": 0.7184249084249084,
      "calib/step_q_w_n": 546.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 515.8828125,
      "completions/mean_terminated_length": 515.8828125,
      "completions/min_length": 99.0,
      "completions/min_terminated_length": 99.0,
      "epoch": 0.0832,
      "grad_norm": 0.004391307011246681,
      "kl": 0.1013946533203125,
      "learning_rate": 3.3888888888888893e-06,
      "loss": 0.0049,
      "num_tokens": 18424100.0,
      "reward": 0.5536520481109619,
      "reward_std": 0.3932393789291382,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.6371738314628601,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.17169275879859924,
      "step": 78
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5052199159871511,
      "calib/avg_num_step_conf": 4.6953125,
      "calib/ece": 0.28796875,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 1.0,
      "calib/frac_conf_gt_0.9": 0.2421875,
      "calib/gap": 0.0026809982703236424,
      "calib/mean_conf": 0.807890625,
      "calib/mu_c": 0.8090845070422534,
      "calib/mu_w": 0.8064035087719298,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2705859375,
      "calib/std_conf": 0.15587280643880566,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7383275862068965,
      "calib/step_q_c_n": 696.0,
      "calib/step_q_gap": 0.014039048657489395,
      "calib/step_q_w": 0.7242885375494071,
      "calib/step_q_w_n": 506.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1273.0,
      "completions/max_terminated_length": 1273.0,
      "completions/mean_length": 503.09375,
      "completions/mean_terminated_length": 505.06671142578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.08426666666666667,
      "grad_norm": 0.004289821721613407,
      "kl": 0.0935821533203125,
      "learning_rate": 3.3611111111111117e-06,
      "loss": 0.0034,
      "num_tokens": 18659268.0,
      "reward": 0.5821490287780762,
      "reward_std": 0.3844422698020935,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6659070253372192,
      "rewards/format_reward_step": 1.0,
      "rewards/step_correlation_reward": 0.1874534785747528,
      "step": 79
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5968559837728195,
      "calib/avg_num_step_conf": 4.86328125,
      "calib/ece": 0.2800000000000001,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.23809523809523808,
      "calib/gap": 0.04668356997971601,
      "calib/mean_conf": 0.8167460317460319,
      "calib/mu_c": 0.8382352941176471,
      "calib/mu_w": 0.7915517241379311,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.27853174603174613,
      "calib/std_conf": 0.1320192949166322,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7245037037037036,
      "calib/step_q_c_n": 675.0,
      "calib/step_q_gap": -0.002162962962963033,
      "calib/step_q_w": 0.7266666666666667,
      "calib/step_q_w_n": 570.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1733.0,
      "completions/max_terminated_length": 1733.0,
      "completions/mean_length": 450.1875,
      "completions/mean_terminated_length": 453.7322692871094,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 170.0,
      "epoch": 0.08533333333333333,
      "grad_norm": 0.004870356526225805,
      "kl": 0.108734130859375,
      "learning_rate": 3.3333333333333333e-06,
      "loss": -0.0003,
      "num_tokens": 18876676.0,
      "reward": 0.534776508808136,
      "reward_std": 0.40727341175079346,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6685359477996826,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.09867335855960846,
      "step": 80
    },
    {
      "calib/answer_extract_rate": 0.9609375,
      "calib/auroc": 0.5414757412398922,
      "calib/avg_num_step_conf": 4.6875,
      "calib/ece": 0.2494715447154471,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.1991869918699187,
      "calib/gap": 0.02374258760107828,
      "calib/mean_conf": 0.7789837398373982,
      "calib/mu_c": 0.7892142857142856,
      "calib/mu_w": 0.7654716981132074,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.22967479674796745,
      "calib/std_conf": 0.18046288647293576,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7274414976599064,
      "calib/step_q_c_n": 641.0,
      "calib/step_q_gap": 0.012504109466704327,
      "calib/step_q_w": 0.7149373881932021,
      "calib/step_q_w_n": 559.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2780.0,
      "completions/max_terminated_length": 2780.0,
      "completions/mean_length": 537.35546875,
      "completions/mean_terminated_length": 539.4627685546875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 180.0,
      "epoch": 0.0864,
      "grad_norm": 0.004383731633424759,
      "kl": 0.09157562255859375,
      "learning_rate": 3.3055555555555558e-06,
      "loss": 0.0371,
      "num_tokens": 19120487.0,
      "reward": 0.5714367628097534,
      "reward_std": 0.4036829471588135,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6617749929428101,
      "rewards/format_reward_step": 0.95703125,
      "rewards/step_correlation_reward": 0.18031713366508484,
      "step": 81
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.48127915149191747,
      "calib/avg_num_step_conf": 4.73046875,
      "calib/ece": 0.3093253968253968,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.1865079365079365,
      "calib/gap": -0.030613379336783653,
      "calib/mean_conf": 0.7924206349206349,
      "calib/mu_c": 0.7789361702127658,
      "calib/mu_w": 0.8095495495495495,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.2711111111111111,
      "calib/std_conf": 0.17541360887164556,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7182456140350878,
      "calib/step_q_c_n": 627.0,
      "calib/step_q_gap": -0.020042057197788843,
      "calib/step_q_w": 0.7382876712328766,
      "calib/step_q_w_n": 584.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2046.0,
      "completions/max_terminated_length": 2046.0,
      "completions/mean_length": 481.953125,
      "completions/mean_terminated_length": 481.953125,
      "completions/min_length": 170.0,
      "completions/min_terminated_length": 170.0,
      "epoch": 0.08746666666666666,
      "grad_norm": 0.004382080864161253,
      "kl": 0.1011962890625,
      "learning_rate": 3.277777777777778e-06,
      "loss": 0.0266,
      "num_tokens": 19349419.0,
      "reward": 0.5547659993171692,
      "reward_std": 0.37202906608581543,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.6337062120437622,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.1711382269859314,
      "step": 82
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5297007874015748,
      "calib/avg_num_step_conf": 4.92578125,
      "calib/ece": 0.3159722222222222,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.24206349206349206,
      "calib/gap": 0.032206299212598544,
      "calib/mean_conf": 0.7813690476190476,
      "calib/mu_c": 0.7976,
      "calib/mu_w": 0.7653937007874014,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.30065476190476187,
      "calib/std_conf": 0.18737167755529816,
      "calib/step_conf_rate": 0.98046875,
      "calib/step_q_c": 0.7250498338870432,
      "calib/step_q_c_n": 602.0,
      "calib/step_q_gap": 0.04211356681572309,
      "calib/step_q_w": 0.6829362670713202,
      "calib/step_q_w_n": 659.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2480.0,
      "completions/max_terminated_length": 2480.0,
      "completions/mean_length": 546.04296875,
      "completions/mean_terminated_length": 546.04296875,
      "completions/min_length": 149.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.08853333333333334,
      "grad_norm": 0.0038687747437506914,
      "kl": 0.0943603515625,
      "learning_rate": 3.2500000000000002e-06,
      "loss": 0.0037,
      "num_tokens": 19596470.0,
      "reward": 0.5185824036598206,
      "reward_std": 0.3328552842140198,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.6270061731338501,
      "rewards/format_reward_step": 0.9609375,
      "rewards/step_correlation_reward": 0.11953361332416534,
      "step": 83
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.496384,
      "calib/avg_num_step_conf": 4.4453125,
      "calib/ece": 0.3124,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.16,
      "calib/gap": 0.0033600000000001407,
      "calib/mean_conf": 0.78112,
      "calib/mu_c": 0.7828,
      "calib/mu_w": 0.7794399999999999,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.29676,
      "calib/std_conf": 0.1631451672591009,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7255148342059338,
      "calib/step_q_c_n": 573.0,
      "calib/step_q_gap": 0.014465276683809924,
      "calib/step_q_w": 0.7110495575221238,
      "calib/step_q_w_n": 565.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2708.0,
      "completions/max_terminated_length": 2708.0,
      "completions/mean_length": 495.66015625,
      "completions/mean_terminated_length": 497.60394287109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 125.0,
      "epoch": 0.0896,
      "grad_norm": 0.004789901431649923,
      "kl": 0.10146331787109375,
      "learning_rate": 3.2222222222222227e-06,
      "loss": 0.0466,
      "num_tokens": 19829279.0,
      "reward": 0.536548912525177,
      "reward_std": 0.3873344659805298,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.6269879341125488,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.15392239391803741,
      "step": 84
    },
    {
      "calib/answer_extract_rate": 0.95703125,
      "calib/auroc": 0.5752652519893899,
      "calib/avg_num_step_conf": 4.625,
      "calib/ece": 0.3207317073170731,
      "calib/final_conf_rate": 0.9609375,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.15853658536585366,
      "calib/gap": 0.05572413793103437,
      "calib/mean_conf": 0.7922764227642275,
      "calib/mu_c": 0.8217241379310343,
      "calib/mu_w": 0.7659999999999999,
      "calib/nonempty_final_conf_rate": 0.9609375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.984375,
      "calib/pce": 0.3207317073170731,
      "calib/std_conf": 0.14751223186503895,
      "calib/step_conf_rate": 0.984375,
      "calib/step_q_c": 0.7304512635379061,
      "calib/step_q_c_n": 554.0,
      "calib/step_q_gap": 0.01986396195060458,
      "calib/step_q_w": 0.7105873015873015,
      "calib/step_q_w_n": 630.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2717.0,
      "completions/max_terminated_length": 2717.0,
      "completions/mean_length": 522.79296875,
      "completions/mean_terminated_length": 522.79296875,
      "completions/min_length": 162.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.09066666666666667,
      "grad_norm": 0.004474072251468897,
      "kl": 0.08792877197265625,
      "learning_rate": 3.1944444444444443e-06,
      "loss": 0.0157,
      "num_tokens": 20070938.0,
      "reward": 0.532092273235321,
      "reward_std": 0.39479580521583557,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6251652240753174,
      "rewards/format_reward_step": 0.94921875,
      "rewards/step_correlation_reward": 0.15855059027671814,
      "step": 85
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.556257982120051,
      "calib/avg_num_step_conf": 4.62890625,
      "calib/ece": 0.33151394422310754,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.2151394422310757,
      "calib/gap": 0.04475095785440608,
      "calib/mean_conf": 0.781792828685259,
      "calib/mu_c": 0.8058620689655172,
      "calib/mu_w": 0.7611111111111111,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9765625,
      "calib/pce": 0.3255776892430279,
      "calib/std_conf": 0.17219341051812373,
      "calib/step_conf_rate": 0.9765625,
      "calib/step_q_c": 0.7266970802919709,
      "calib/step_q_c_n": 548.0,
      "calib/step_q_gap": 0.030689230998407258,
      "calib/step_q_w": 0.6960078492935636,
      "calib/step_q_w_n": 637.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1767.0,
      "completions/max_terminated_length": 1767.0,
      "completions/mean_length": 493.74609375,
      "completions/mean_terminated_length": 495.682373046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 135.0,
      "epoch": 0.09173333333333333,
      "grad_norm": 0.004494712222367525,
      "kl": 0.094451904296875,
      "learning_rate": 3.1666666666666667e-06,
      "loss": -0.0006,
      "num_tokens": 20302849.0,
      "reward": 0.5276005268096924,
      "reward_std": 0.3682413399219513,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6204468607902527,
      "rewards/format_reward_step": 0.96484375,
      "rewards/step_correlation_reward": 0.15116043388843536,
      "step": 86
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.502650295222759,
      "calib/avg_num_step_conf": 4.44140625,
      "calib/ece": 0.22822834645669293,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.18110236220472442,
      "calib/gap": -0.01292807300053711,
      "calib/mean_conf": 0.7861023622047245,
      "calib/mu_c": 0.7814197530864198,
      "calib/mu_w": 0.7943478260869569,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.1882677165354331,
      "calib/std_conf": 0.1742005512094152,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7099722607489598,
      "calib/step_q_c_n": 721.0,
      "calib/step_q_gap": 0.020933799210498316,
      "calib/step_q_w": 0.6890384615384615,
      "calib/step_q_w_n": 416.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1076.0,
      "completions/max_terminated_length": 1076.0,
      "completions/mean_length": 436.3203125,
      "completions/mean_terminated_length": 438.0314025878906,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 160.0,
      "epoch": 0.0928,
      "grad_norm": 0.005090489983558655,
      "kl": 0.1055450439453125,
      "learning_rate": 3.138888888888889e-06,
      "loss": 0.0088,
      "num_tokens": 20520043.0,
      "reward": 0.6783336400985718,
      "reward_std": 0.3884207308292389,
      "rewards/accuracy_reward_step": 0.6328125,
      "rewards/final_brier_reward_step": 0.7014593482017517,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.33098912239074707,
      "step": 87
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.530716296527995,
      "calib/avg_num_step_conf": 5.00390625,
      "calib/ece": 0.26669291338582674,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.2440944881889764,
      "calib/gap": 0.02194932749452594,
      "calib/mean_conf": 0.7990551181102361,
      "calib/mu_c": 0.8089928057553956,
      "calib/mu_w": 0.7870434782608696,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.259251968503937,
      "calib/std_conf": 0.1491811405976822,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7272056737588654,
      "calib/step_q_c_n": 705.0,
      "calib/step_q_gap": 0.01809109042553203,
      "calib/step_q_w": 0.7091145833333333,
      "calib/step_q_w_n": 576.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1999.0,
      "completions/max_terminated_length": 1999.0,
      "completions/mean_length": 511.73828125,
      "completions/mean_terminated_length": 513.7451171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 181.0,
      "epoch": 0.09386666666666667,
      "grad_norm": 0.00454970495775342,
      "kl": 0.0915679931640625,
      "learning_rate": 3.1111111111111116e-06,
      "loss": 0.0229,
      "num_tokens": 20760896.0,
      "reward": 0.6263402700424194,
      "reward_std": 0.3962945342063904,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.6637749671936035,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.2849992513656616,
      "step": 88
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5435695538057742,
      "calib/avg_num_step_conf": 4.421875,
      "calib/ece": 0.3020242914979757,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.18218623481781376,
      "calib/gap": 0.027177821522309853,
      "calib/mean_conf": 0.7791093117408907,
      "calib/mu_c": 0.7930833333333334,
      "calib/mu_w": 0.7659055118110235,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.2976518218623482,
      "calib/std_conf": 0.15574301536944893,
      "calib/step_conf_rate": 0.98046875,
      "calib/step_q_c": 0.7219411764705882,
      "calib/step_q_c_n": 510.0,
      "calib/step_q_gap": 0.030799697370909818,
      "calib/step_q_w": 0.6911414790996784,
      "calib/step_q_w_n": 622.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2394.0,
      "completions/max_terminated_length": 2394.0,
      "completions/mean_length": 535.5859375,
      "completions/mean_terminated_length": 537.686279296875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 156.0,
      "epoch": 0.09493333333333333,
      "grad_norm": 0.004490287974476814,
      "kl": 0.08609771728515625,
      "learning_rate": 3.0833333333333336e-06,
      "loss": -0.0259,
      "num_tokens": 21006894.0,
      "reward": 0.5326032638549805,
      "reward_std": 0.30912187695503235,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6257109642028809,
      "rewards/format_reward_step": 0.95703125,
      "rewards/step_correlation_reward": 0.15433922410011292,
      "step": 89
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6094750320102433,
      "calib/avg_num_step_conf": 5.51953125,
      "calib/ece": 0.26063492063492066,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.19047619047619047,
      "calib/gap": 0.046624839948783636,
      "calib/mean_conf": 0.7949999999999999,
      "calib/mu_c": 0.8153521126760563,
      "calib/mu_w": 0.7687272727272727,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.24607142857142858,
      "calib/std_conf": 0.16672974996625384,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7206089309878213,
      "calib/step_q_c_n": 739.0,
      "calib/step_q_gap": 0.02218163128455719,
      "calib/step_q_w": 0.6984272997032641,
      "calib/step_q_w_n": 674.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2317.0,
      "completions/max_terminated_length": 2317.0,
      "completions/mean_length": 509.3828125,
      "completions/mean_terminated_length": 511.38043212890625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.096,
      "grad_norm": 0.004317185375839472,
      "kl": 0.0949859619140625,
      "learning_rate": 3.055555555555556e-06,
      "loss": 0.0119,
      "num_tokens": 21240616.0,
      "reward": 0.6395683884620667,
      "reward_std": 0.3783494830131531,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6792097687721252,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.2936769425868988,
      "step": 90
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5705386237301131,
      "calib/avg_num_step_conf": 5.0546875,
      "calib/ece": 0.2511111111111112,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.20634920634920634,
      "calib/gap": 0.03691010159095254,
      "calib/mean_conf": 0.7865079365079366,
      "calib/mu_c": 0.8027659574468085,
      "calib/mu_w": 0.7658558558558559,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.23904761904761912,
      "calib/std_conf": 0.16857864002749512,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7056178217821782,
      "calib/step_q_c_n": 707.0,
      "calib/step_q_gap": 0.004203852446573397,
      "calib/step_q_w": 0.7014139693356048,
      "calib/step_q_w_n": 587.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1629.0,
      "completions/max_terminated_length": 1629.0,
      "completions/mean_length": 519.99609375,
      "completions/mean_terminated_length": 519.99609375,
      "completions/min_length": 186.0,
      "completions/min_terminated_length": 186.0,
      "epoch": 0.09706666666666666,
      "grad_norm": 0.004235545173287392,
      "kl": 0.0934906005859375,
      "learning_rate": 3.0277777777777776e-06,
      "loss": -0.0043,
      "num_tokens": 21481447.0,
      "reward": 0.5762742757797241,
      "reward_std": 0.4081330895423889,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.6798367500305176,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.16646182537078857,
      "step": 91
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5041772959183672,
      "calib/avg_num_step_conf": 4.41015625,
      "calib/ece": 0.28130952380952373,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.15476190476190477,
      "calib/gap": 0.0013571428571428346,
      "calib/mean_conf": 0.8050396825396826,
      "calib/mu_c": 0.8056428571428571,
      "calib/mu_w": 0.8042857142857143,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.26539682539682535,
      "calib/std_conf": 0.1355075644109864,
      "calib/step_conf_rate": 0.98046875,
      "calib/step_q_c": 0.7301848739495799,
      "calib/step_q_c_n": 595.0,
      "calib/step_q_gap": 0.004106222264186665,
      "calib/step_q_w": 0.7260786516853932,
      "calib/step_q_w_n": 534.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2207.0,
      "completions/max_terminated_length": 2207.0,
      "completions/mean_length": 473.8984375,
      "completions/mean_terminated_length": 475.75689697265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 173.0,
      "epoch": 0.09813333333333334,
      "grad_norm": 0.004517954774200916,
      "kl": 0.09569549560546875,
      "learning_rate": 3e-06,
      "loss": -0.0145,
      "num_tokens": 21709485.0,
      "reward": 0.6152100563049316,
      "reward_std": 0.3936840295791626,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6535598039627075,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.27373531460762024,
      "step": 92
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5125,
      "calib/avg_num_step_conf": 5.43359375,
      "calib/ece": 0.2835458167330677,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.16733067729083664,
      "calib/gap": 0.026701653944020465,
      "calib/mean_conf": 0.7968525896414342,
      "calib/mu_c": 0.8096183206106871,
      "calib/mu_w": 0.7829166666666666,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.27924302788844624,
      "calib/std_conf": 0.15364902243388137,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7297272727272728,
      "calib/step_q_c_n": 660.0,
      "calib/step_q_gap": 0.016033702275836337,
      "calib/step_q_w": 0.7136935704514364,
      "calib/step_q_w_n": 731.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2662.0,
      "completions/max_terminated_length": 2662.0,
      "completions/mean_length": 542.171875,
      "completions/mean_terminated_length": 542.171875,
      "completions/min_length": 173.0,
      "completions/min_terminated_length": 173.0,
      "epoch": 0.0992,
      "grad_norm": 0.00440006610006094,
      "kl": 0.08103179931640625,
      "learning_rate": 2.9722222222222225e-06,
      "loss": 0.0513,
      "num_tokens": 21954057.0,
      "reward": 0.5715123414993286,
      "reward_std": 0.4189292788505554,
      "rewards/accuracy_reward_step": 0.51171875,
      "rewards/final_brier_reward_step": 0.6469859480857849,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.1991637945175171,
      "step": 93
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5303413400758534,
      "calib/avg_num_step_conf": 5.0703125,
      "calib/ece": 0.26458498023715404,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.20553359683794467,
      "calib/gap": 0.031416561314791314,
      "calib/mean_conf": 0.8100395256916996,
      "calib/mu_c": 0.8240714285714285,
      "calib/mu_w": 0.7926548672566371,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2606324110671936,
      "calib/std_conf": 0.12895903698589095,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7232162921348315,
      "calib/step_q_c_n": 712.0,
      "calib/step_q_gap": 0.029717998619473218,
      "calib/step_q_w": 0.6934982935153583,
      "calib/step_q_w_n": 586.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2273.0,
      "completions/max_terminated_length": 2273.0,
      "completions/mean_length": 492.11328125,
      "completions/mean_terminated_length": 495.9881896972656,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 202.0,
      "epoch": 0.10026666666666667,
      "grad_norm": 0.004632134456187487,
      "kl": 0.0995025634765625,
      "learning_rate": 2.944444444444445e-06,
      "loss": -0.0086,
      "num_tokens": 22188718.0,
      "reward": 0.6204875707626343,
      "reward_std": 0.3555172085762024,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6778242588043213,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.25611966848373413,
      "step": 94
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5884313725490197,
      "calib/avg_num_step_conf": 5.56640625,
      "calib/ece": 0.19924901185770755,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.15019762845849802,
      "calib/gap": 0.043509803921568446,
      "calib/mean_conf": 0.7673122529644268,
      "calib/mu_c": 0.7845098039215685,
      "calib/mu_w": 0.7410000000000001,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.18090909090909096,
      "calib/std_conf": 0.18145407777171457,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7088129899216126,
      "calib/step_q_c_n": 893.0,
      "calib/step_q_gap": 0.03947088465845461,
      "calib/step_q_w": 0.669342105263158,
      "calib/step_q_w_n": 532.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2614.0,
      "completions/max_terminated_length": 2614.0,
      "completions/mean_length": 527.21875,
      "completions/mean_terminated_length": 529.2863159179688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 135.0,
      "epoch": 0.10133333333333333,
      "grad_norm": 0.004121353849768639,
      "kl": 0.081787109375,
      "learning_rate": 2.916666666666667e-06,
      "loss": 0.0126,
      "num_tokens": 22429814.0,
      "reward": 0.6321449279785156,
      "reward_std": 0.3281705379486084,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.7107530832290649,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.23791177570819855,
      "step": 95
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5467836257309941,
      "calib/avg_num_step_conf": 5.265625,
      "calib/ece": 0.17492125984251963,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.17716535433070865,
      "calib/gap": 0.006558867047135708,
      "calib/mean_conf": 0.8061023622047244,
      "calib/mu_c": 0.8082456140350875,
      "calib/mu_w": 0.8016867469879518,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.15389763779527557,
      "calib/std_conf": 0.1393212079002976,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7324615384615384,
      "calib/step_q_c_n": 910.0,
      "calib/step_q_gap": 0.03467615033368454,
      "calib/step_q_w": 0.6977853881278538,
      "calib/step_q_w_n": 438.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2744.0,
      "completions/max_terminated_length": 2744.0,
      "completions/mean_length": 490.96484375,
      "completions/mean_terminated_length": 490.96484375,
      "completions/min_length": 152.0,
      "completions/min_terminated_length": 152.0,
      "epoch": 0.1024,
      "grad_norm": 0.004674531519412994,
      "kl": 0.10054779052734375,
      "learning_rate": 2.888888888888889e-06,
      "loss": -0.0372,
      "num_tokens": 22661317.0,
      "reward": 0.6969508528709412,
      "reward_std": 0.30746543407440186,
      "rewards/accuracy_reward_step": 0.66796875,
      "rewards/final_brier_reward_step": 0.7400012016296387,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.32186925411224365,
      "step": 96
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5414687499999999,
      "calib/avg_num_step_conf": 5.140625,
      "calib/ece": 0.30486166007905124,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.16600790513833993,
      "calib/gap": 0.02685499999999985,
      "calib/mean_conf": 0.788893280632411,
      "calib/mu_c": 0.8024799999999999,
      "calib/mu_w": 0.775625,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2998418972332014,
      "calib/std_conf": 0.1536292199667022,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7190620031796502,
      "calib/step_q_c_n": 629.0,
      "calib/step_q_gap": 0.010648611622153803,
      "calib/step_q_w": 0.7084133915574964,
      "calib/step_q_w_n": 687.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2592.0,
      "completions/max_terminated_length": 2592.0,
      "completions/mean_length": 503.1796875,
      "completions/mean_terminated_length": 505.1529846191406,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.10346666666666667,
      "grad_norm": 0.004628482274711132,
      "kl": 0.0954132080078125,
      "learning_rate": 2.861111111111111e-06,
      "loss": -0.0009,
      "num_tokens": 22895203.0,
      "reward": 0.535427987575531,
      "reward_std": 0.4217776358127594,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.6416449546813965,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.13467980921268463,
      "step": 97
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5434121621621621,
      "calib/avg_num_step_conf": 5.2265625,
      "calib/ece": 0.23955645161290323,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.18951612903225806,
      "calib/gap": 0.013878378378378575,
      "calib/mean_conf": 0.7977822580645162,
      "calib/mu_c": 0.8033783783783784,
      "calib/mu_w": 0.7894999999999999,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.22028225806451612,
      "calib/std_conf": 0.1505350704226733,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7124238544474394,
      "calib/step_q_c_n": 742.0,
      "calib/step_q_gap": 0.004319827601801718,
      "calib/step_q_w": 0.7081040268456377,
      "calib/step_q_w_n": 596.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2832.0,
      "completions/max_terminated_length": 2832.0,
      "completions/mean_length": 580.94140625,
      "completions/mean_terminated_length": 583.2196655273438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 163.0,
      "epoch": 0.10453333333333334,
      "grad_norm": 0.004246114287525415,
      "kl": 0.07635498046875,
      "learning_rate": 2.8333333333333335e-06,
      "loss": 0.0686,
      "num_tokens": 23150108.0,
      "reward": 0.6070481538772583,
      "reward_std": 0.3663785457611084,
      "rewards/accuracy_reward_step": 0.578125,
      "rewards/final_brier_reward_step": 0.6810113191604614,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.22370991110801697,
      "step": 98
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5169593117551654,
      "calib/avg_num_step_conf": 5.63671875,
      "calib/ece": 0.45336,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.196,
      "calib/gap": 0.018194767646851573,
      "calib/mean_conf": 0.79216,
      "calib/mu_c": 0.8040229885057472,
      "calib/mu_w": 0.7858282208588956,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.44876,
      "calib/std_conf": 0.1691039159806774,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7280092843326886,
      "calib/step_q_c_n": 517.0,
      "calib/step_q_gap": 0.03957515906271025,
      "calib/step_q_w": 0.6884341252699784,
      "calib/step_q_w_n": 926.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2670.0,
      "completions/max_terminated_length": 2670.0,
      "completions/mean_length": 654.35546875,
      "completions/mean_terminated_length": 654.35546875,
      "completions/min_length": 144.0,
      "completions/min_terminated_length": 144.0,
      "epoch": 0.1056,
      "grad_norm": 0.004029831383377314,
      "kl": 0.07680511474609375,
      "learning_rate": 2.805555555555556e-06,
      "loss": 0.0207,
      "num_tokens": 23423423.0,
      "reward": 0.3443741202354431,
      "reward_std": 0.39202311635017395,
      "rewards/accuracy_reward_step": 0.33984375,
      "rewards/final_brier_reward_step": 0.5394691824913025,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": -0.11243967711925507,
      "step": 99
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.6053447420634921,
      "calib/avg_num_step_conf": 5.41015625,
      "calib/ece": 0.2951968503937008,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.1377952755905512,
      "calib/gap": 0.057713293650793585,
      "calib/mean_conf": 0.7754330708661418,
      "calib/mu_c": 0.8040625,
      "calib/mu_w": 0.7463492063492064,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2833464566929134,
      "calib/std_conf": 0.16153425762553517,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7255524079320114,
      "calib/step_q_c_n": 706.0,
      "calib/step_q_gap": 0.03624460233554605,
      "calib/step_q_w": 0.6893078055964653,
      "calib/step_q_w_n": 679.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1979.0,
      "completions/max_terminated_length": 1979.0,
      "completions/mean_length": 556.93359375,
      "completions/mean_terminated_length": 559.11767578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 138.0,
      "epoch": 0.10666666666666667,
      "grad_norm": 0.004563473630696535,
      "kl": 0.08527374267578125,
      "learning_rate": 2.7777777777777783e-06,
      "loss": 0.0299,
      "num_tokens": 23673406.0,
      "reward": 0.5480512976646423,
      "reward_std": 0.3751283586025238,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.6646746397018433,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.13533419370651245,
      "step": 100
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5611288392500997,
      "calib/avg_num_step_conf": 5.953125,
      "calib/ece": 0.355587044534413,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.9609375,
      "calib/frac_conf_gt_0.9": 0.18218623481781376,
      "calib/gap": 0.060533173780082494,
      "calib/mean_conf": 0.79251012145749,
      "calib/mu_c": 0.8263302752293579,
      "calib/mu_w": 0.7657971014492754,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3534008097165992,
      "calib/std_conf": 0.18371386531917389,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.732985315712188,
      "calib/step_q_c_n": 681.0,
      "calib/step_q_gap": 0.0508738091878701,
      "calib/step_q_w": 0.6821115065243178,
      "calib/step_q_w_n": 843.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2217.0,
      "completions/max_terminated_length": 2217.0,
      "completions/mean_length": 619.71484375,
      "completions/mean_terminated_length": 622.1451416015625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 200.0,
      "epoch": 0.10773333333333333,
      "grad_norm": 0.004082713276147842,
      "kl": 0.08042144775390625,
      "learning_rate": 2.7500000000000004e-06,
      "loss": 0.0349,
      "num_tokens": 23939045.0,
      "reward": 0.5178720951080322,
      "reward_std": 0.37443262338638306,
      "rewards/accuracy_reward_step": 0.4296875,
      "rewards/final_brier_reward_step": 0.6013695001602173,
      "rewards/format_reward_step": 0.9609375,
      "rewards/step_correlation_reward": 0.15624955296516418,
      "step": 101
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.517766329832904,
      "calib/avg_num_step_conf": 5.7109375,
      "calib/ece": 0.23312,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.212,
      "calib/gap": 0.01893402021002588,
      "calib/mean_conf": 0.80424,
      "calib/mu_c": 0.8120408163265307,
      "calib/mu_w": 0.7931067961165048,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.22468,
      "calib/std_conf": 0.14438982789656618,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7196905940594059,
      "calib/step_q_c_n": 808.0,
      "calib/step_q_gap": -0.016349161292276038,
      "calib/step_q_w": 0.736039755351682,
      "calib/step_q_w_n": 654.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2713.0,
      "completions/max_terminated_length": 2713.0,
      "completions/mean_length": 534.40234375,
      "completions/mean_terminated_length": 534.40234375,
      "completions/min_length": 167.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.1088,
      "grad_norm": 0.004449573345482349,
      "kl": 0.0848236083984375,
      "learning_rate": 2.7222222222222224e-06,
      "loss": 0.012,
      "num_tokens": 24182548.0,
      "reward": 0.5914692878723145,
      "reward_std": 0.316368043422699,
      "rewards/accuracy_reward_step": 0.57421875,
      "rewards/final_brier_reward_step": 0.6829195022583008,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.18986272811889648,
      "step": 102
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5246334698190463,
      "calib/avg_num_step_conf": 5.19921875,
      "calib/ece": 0.26611336032388666,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.95703125,
      "calib/frac_conf_gt_0.9": 0.17813765182186234,
      "calib/gap": 0.030098401796328167,
      "calib/mean_conf": 0.7933198380566802,
      "calib/mu_c": 0.807089552238806,
      "calib/mu_w": 0.7769911504424778,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 0.98046875,
      "calib/nonempty_step_conf_rate": 0.97265625,
      "calib/pce": 0.25846153846153846,
      "calib/std_conf": 0.1614067426411123,
      "calib/step_conf_rate": 0.97265625,
      "calib/step_q_c": 0.7253805970149254,
      "calib/step_q_c_n": 670.0,
      "calib/step_q_gap": -0.006268419626526911,
      "calib/step_q_w": 0.7316490166414523,
      "calib/step_q_w_n": 661.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2265.0,
      "completions/max_terminated_length": 2265.0,
      "completions/mean_length": 630.05078125,
      "completions/mean_terminated_length": 632.5216064453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 164.0,
      "epoch": 0.10986666666666667,
      "grad_norm": 0.004228085279464722,
      "kl": 0.0755157470703125,
      "learning_rate": 2.6944444444444444e-06,
      "loss": 0.0224,
      "num_tokens": 24448393.0,
      "reward": 0.5414041876792908,
      "reward_std": 0.33534783124923706,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.6473226547241211,
      "rewards/format_reward_step": 0.95703125,
      "rewards/step_correlation_reward": 0.13939201831817627,
      "step": 103
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5704055059523809,
      "calib/avg_num_step_conf": 5.53515625,
      "calib/ece": 0.30763779527559054,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.12992125984251968,
      "calib/gap": 0.037669890873015976,
      "calib/mean_conf": 0.7881102362204725,
      "calib/mu_c": 0.8067968750000001,
      "calib/mu_w": 0.7691269841269841,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.29590551181102365,
      "calib/std_conf": 0.15763624170338938,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.737034990791897,
      "calib/step_q_c_n": 724.0,
      "calib/step_q_gap": 0.0328502866071928,
      "calib/step_q_w": 0.7041847041847042,
      "calib/step_q_w_n": 693.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2902.0,
      "completions/max_terminated_length": 2902.0,
      "completions/mean_length": 571.26171875,
      "completions/mean_terminated_length": 571.26171875,
      "completions/min_length": 153.0,
      "completions/min_terminated_length": 153.0,
      "epoch": 0.11093333333333333,
      "grad_norm": 0.004389981739223003,
      "kl": 0.0796051025390625,
      "learning_rate": 2.666666666666667e-06,
      "loss": 0.0152,
      "num_tokens": 24701316.0,
      "reward": 0.5887982249259949,
      "reward_std": 0.2890593409538269,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.658064067363739,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.22109487652778625,
      "step": 104
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5200482172313158,
      "calib/avg_num_step_conf": 5.84375,
      "calib/ece": 0.25849802371541497,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.14624505928853754,
      "calib/gap": 0.024042634183479183,
      "calib/mean_conf": 0.7691699604743083,
      "calib/mu_c": 0.7797183098591549,
      "calib/mu_w": 0.7556756756756757,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.23320158102766794,
      "calib/std_conf": 0.17764517741301394,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7048106060606061,
      "calib/step_q_c_n": 792.0,
      "calib/step_q_gap": 0.00456912878787874,
      "calib/step_q_w": 0.7002414772727273,
      "calib/step_q_w_n": 704.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2781.0,
      "completions/max_terminated_length": 2781.0,
      "completions/mean_length": 584.81640625,
      "completions/mean_terminated_length": 584.81640625,
      "completions/min_length": 206.0,
      "completions/min_terminated_length": 206.0,
      "epoch": 0.112,
      "grad_norm": 0.00401820195838809,
      "kl": 0.08333587646484375,
      "learning_rate": 2.6388888888888893e-06,
      "loss": 0.0686,
      "num_tokens": 24956789.0,
      "reward": 0.6147169470787048,
      "reward_std": 0.4103122353553772,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.678810179233551,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.2428111732006073,
      "step": 105
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5796992481203007,
      "calib/avg_num_step_conf": 5.5234375,
      "calib/ece": 0.28367588932806315,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.1383399209486166,
      "calib/gap": 0.03851002506265677,
      "calib/mean_conf": 0.8024110671936759,
      "calib/mu_c": 0.8206766917293233,
      "calib/mu_w": 0.7821666666666666,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.28019762845849794,
      "calib/std_conf": 0.13706161519451995,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7453986013986015,
      "calib/step_q_c_n": 715.0,
      "calib/step_q_gap": 0.02011963144151996,
      "calib/step_q_w": 0.7252789699570815,
      "calib/step_q_w_n": 699.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1616.0,
      "completions/max_terminated_length": 1616.0,
      "completions/mean_length": 527.32421875,
      "completions/mean_terminated_length": 529.3922119140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 186.0,
      "epoch": 0.11306666666666666,
      "grad_norm": 0.004746657330542803,
      "kl": 0.08556365966796875,
      "learning_rate": 2.6111111111111113e-06,
      "loss": 0.0056,
      "num_tokens": 25196368.0,
      "reward": 0.5568984746932983,
      "reward_std": 0.35886260867118835,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.6666004061698914,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.1456340253353119,
      "step": 106
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5674223602484472,
      "calib/avg_num_step_conf": 5.6484375,
      "calib/ece": 0.28168627450980394,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.1803921568627451,
      "calib/gap": 0.025102484472049658,
      "calib/mean_conf": 0.785607843137255,
      "calib/mu_c": 0.7969285714285714,
      "calib/mu_w": 0.7718260869565218,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2591372549019608,
      "calib/std_conf": 0.15821654042474634,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7204850746268656,
      "calib/step_q_c_n": 804.0,
      "calib/step_q_gap": 0.008226507648672499,
      "calib/step_q_w": 0.7122585669781931,
      "calib/step_q_w_n": 642.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1325.0,
      "completions/max_terminated_length": 1325.0,
      "completions/mean_length": 511.19921875,
      "completions/mean_terminated_length": 513.2039794921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 160.0,
      "epoch": 0.11413333333333334,
      "grad_norm": 0.004514755215495825,
      "kl": 0.0935516357421875,
      "learning_rate": 2.5833333333333337e-06,
      "loss": 0.0336,
      "num_tokens": 25431851.0,
      "reward": 0.6089458465576172,
      "reward_std": 0.37728819251060486,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6811558604240417,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.22814197838306427,
      "step": 107
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5240497076023392,
      "calib/avg_num_step_conf": 6.40625,
      "calib/ece": 0.1715537848605578,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.20318725099601595,
      "calib/gap": 0.00889400584795308,
      "calib/mean_conf": 0.8214342629482072,
      "calib/mu_c": 0.8242690058479532,
      "calib/mu_w": 0.8153750000000001,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.1558565737051793,
      "calib/std_conf": 0.12750570122928112,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7346252788104088,
      "calib/step_q_c_n": 1076.0,
      "calib/step_q_gap": -0.01913358643781815,
      "calib/step_q_w": 0.753758865248227,
      "calib/step_q_w_n": 564.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2303.0,
      "completions/max_terminated_length": 2303.0,
      "completions/mean_length": 572.57421875,
      "completions/mean_terminated_length": 574.8196411132812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 31.0,
      "epoch": 0.1152,
      "grad_norm": 0.004031268414109945,
      "kl": 0.07091522216796875,
      "learning_rate": 2.5555555555555557e-06,
      "loss": -0.021,
      "num_tokens": 25681662.0,
      "reward": 0.696212887763977,
      "reward_std": 0.34606415033340454,
      "rewards/accuracy_reward_step": 0.66796875,
      "rewards/final_brier_reward_step": 0.7361562252044678,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.3265819549560547,
      "step": 108
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.6031997414350355,
      "calib/avg_num_step_conf": 6.375,
      "calib/ece": 0.3474698795180722,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.17670682730923695,
      "calib/gap": 0.051499676793794524,
      "calib/mean_conf": 0.8253815261044176,
      "calib/mu_c": 0.8522689075630252,
      "calib/mu_w": 0.8007692307692307,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3474698795180722,
      "calib/std_conf": 0.12503796620246393,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7522580645161291,
      "calib/step_q_c_n": 744.0,
      "calib/step_q_gap": 0.04583914559721014,
      "calib/step_q_w": 0.706418918918919,
      "calib/step_q_w_n": 888.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2096.0,
      "completions/max_terminated_length": 2096.0,
      "completions/mean_length": 584.484375,
      "completions/mean_terminated_length": 586.7764892578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 211.0,
      "epoch": 0.11626666666666667,
      "grad_norm": 0.004027488175779581,
      "kl": 0.0707855224609375,
      "learning_rate": 2.5277777777777778e-06,
      "loss": -0.035,
      "num_tokens": 25935890.0,
      "reward": 0.5401857495307922,
      "reward_std": 0.2600173056125641,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.622322678565979,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.17054879665374756,
      "step": 109
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.45666666666666667,
      "calib/avg_num_step_conf": 5.484375,
      "calib/ece": 0.31175999999999987,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.128,
      "calib/gap": -0.01839743589743592,
      "calib/mean_conf": 0.7956,
      "calib/mu_c": 0.7867692307692308,
      "calib/mu_w": 0.8051666666666667,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2936799999999999,
      "calib/std_conf": 0.14804810029176327,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7333526850507982,
      "calib/step_q_c_n": 689.0,
      "calib/step_q_gap": 0.011856181554294687,
      "calib/step_q_w": 0.7214965034965035,
      "calib/step_q_w_n": 715.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2667.0,
      "completions/max_terminated_length": 2667.0,
      "completions/mean_length": 552.046875,
      "completions/mean_terminated_length": 554.2117919921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.11733333333333333,
      "grad_norm": 0.004308481700718403,
      "kl": 0.0877532958984375,
      "learning_rate": 2.5e-06,
      "loss": 0.062,
      "num_tokens": 26182134.0,
      "reward": 0.549569845199585,
      "reward_std": 0.41790106892585754,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.6282640695571899,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.1740006059408188,
      "step": 110
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5433070866141732,
      "calib/avg_num_step_conf": 6.07421875,
      "calib/ece": 0.3258565737051793,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.14342629482071714,
      "calib/gap": 0.012054864109728114,
      "calib/mean_conf": 0.8142231075697212,
      "calib/mu_c": 0.8203225806451613,
      "calib/mu_w": 0.8082677165354332,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.32302788844621516,
      "calib/std_conf": 0.11807087320548772,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7442020497803807,
      "calib/step_q_c_n": 683.0,
      "calib/step_q_gap": 0.013708930514325646,
      "calib/step_q_w": 0.7304931192660551,
      "calib/step_q_w_n": 872.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2788.0,
      "completions/max_terminated_length": 2788.0,
      "completions/mean_length": 589.40234375,
      "completions/mean_terminated_length": 589.40234375,
      "completions/min_length": 186.0,
      "completions/min_terminated_length": 186.0,
      "epoch": 0.1184,
      "grad_norm": 0.003989416174590588,
      "kl": 0.0763092041015625,
      "learning_rate": 2.4722222222222226e-06,
      "loss": 0.0195,
      "num_tokens": 26440429.0,
      "reward": 0.5446423292160034,
      "reward_std": 0.36422672867774963,
      "rewards/accuracy_reward_step": 0.48828125,
      "rewards/final_brier_reward_step": 0.6271018981933594,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.1684325933456421,
      "step": 111
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5472689075630253,
      "calib/avg_num_step_conf": 5.4375,
      "calib/ece": 0.280161943319838,
      "calib/final_conf_rate": 0.96484375,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.145748987854251,
      "calib/gap": 0.024548975840336218,
      "calib/mean_conf": 0.7724696356275305,
      "calib/mu_c": 0.784296875,
      "calib/mu_w": 0.7597478991596638,
      "calib/nonempty_final_conf_rate": 0.96484375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.2672064777327935,
      "calib/std_conf": 0.1633975495515841,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7214992927864214,
      "calib/step_q_c_n": 707.0,
      "calib/step_q_gap": 0.034798562859414184,
      "calib/step_q_w": 0.6867007299270073,
      "calib/step_q_w_n": 685.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2989.0,
      "completions/max_terminated_length": 2989.0,
      "completions/mean_length": 619.1953125,
      "completions/mean_terminated_length": 621.6235961914062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.11946666666666667,
      "grad_norm": 0.004297351930290461,
      "kl": 0.07457733154296875,
      "learning_rate": 2.4444444444444447e-06,
      "loss": 0.0347,
      "num_tokens": 26706863.0,
      "reward": 0.5232102870941162,
      "reward_std": 0.36772620677948,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.6476492285728455,
      "rewards/format_reward_step": 0.96484375,
      "rewards/step_correlation_reward": 0.10502135753631592,
      "step": 112
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5253927997030805,
      "calib/avg_num_step_conf": 5.76171875,
      "calib/ece": 0.2656078431372549,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.11764705882352941,
      "calib/gap": 0.015623530867252322,
      "calib/mean_conf": 0.8015294117647058,
      "calib/mu_c": 0.8087591240875913,
      "calib/mu_w": 0.793135593220339,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.26494117647058824,
      "calib/std_conf": 0.12424049305818236,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7267310167310168,
      "calib/step_q_c_n": 777.0,
      "calib/step_q_gap": 0.03375107403760702,
      "calib/step_q_w": 0.6929799426934098,
      "calib/step_q_w_n": 698.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2699.0,
      "completions/max_terminated_length": 2699.0,
      "completions/mean_length": 535.06640625,
      "completions/mean_terminated_length": 535.06640625,
      "completions/min_length": 205.0,
      "completions/min_terminated_length": 205.0,
      "epoch": 0.12053333333333334,
      "grad_norm": 0.004804201424121857,
      "kl": 0.0929718017578125,
      "learning_rate": 2.4166666666666667e-06,
      "loss": 0.0349,
      "num_tokens": 26949040.0,
      "reward": 0.5936443209648132,
      "reward_std": 0.371960312128067,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.6712472438812256,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.2097914069890976,
      "step": 113
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.562398334309324,
      "calib/avg_num_step_conf": 5.953125,
      "calib/ece": 0.25304,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.136,
      "calib/gap": 0.02785477259418323,
      "calib/mean_conf": 0.8035999999999999,
      "calib/mu_c": 0.8157446808510638,
      "calib/mu_w": 0.7878899082568805,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.24631999999999998,
      "calib/std_conf": 0.118222840432803,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7349100257069409,
      "calib/step_q_c_n": 778.0,
      "calib/step_q_gap": 0.026223698629192826,
      "calib/step_q_w": 0.708686327077748,
      "calib/step_q_w_n": 746.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2387.0,
      "completions/max_terminated_length": 2387.0,
      "completions/mean_length": 575.57421875,
      "completions/mean_terminated_length": 575.57421875,
      "completions/min_length": 179.0,
      "completions/min_terminated_length": 179.0,
      "epoch": 0.1216,
      "grad_norm": 0.004351339768618345,
      "kl": 0.084808349609375,
      "learning_rate": 2.388888888888889e-06,
      "loss": 0.057,
      "num_tokens": 27201411.0,
      "reward": 0.6057231426239014,
      "reward_std": 0.3419015407562256,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.6800882816314697,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.22588929533958435,
      "step": 114
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5503722084367245,
      "calib/avg_num_step_conf": 5.84765625,
      "calib/ece": 0.3108661417322835,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.11811023622047244,
      "calib/gap": 0.018978908188585608,
      "calib/mean_conf": 0.798503937007874,
      "calib/mu_c": 0.8077692307692307,
      "calib/mu_w": 0.7887903225806451,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2987795275590552,
      "calib/std_conf": 0.1413298968664433,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7183467741935484,
      "calib/step_q_c_n": 744.0,
      "calib/step_q_gap": -0.005624009338988212,
      "calib/step_q_w": 0.7239707835325366,
      "calib/step_q_w_n": 753.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2196.0,
      "completions/max_terminated_length": 2196.0,
      "completions/mean_length": 540.47265625,
      "completions/mean_terminated_length": 540.47265625,
      "completions/min_length": 218.0,
      "completions/min_terminated_length": 218.0,
      "epoch": 0.12266666666666666,
      "grad_norm": 0.004473676439374685,
      "kl": 0.0851287841796875,
      "learning_rate": 2.361111111111111e-06,
      "loss": 0.0075,
      "num_tokens": 27445036.0,
      "reward": 0.5372321605682373,
      "reward_std": 0.40835297107696533,
      "rewards/accuracy_reward_step": 0.5078125,
      "rewards/final_brier_reward_step": 0.6523203253746033,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.12214392423629761,
      "step": 115
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.4656552077331797,
      "calib/avg_num_step_conf": 6.015625,
      "calib/ece": 0.3060400000000001,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.124,
      "calib/gap": 0.008253632930030008,
      "calib/mean_conf": 0.79964,
      "calib/mu_c": 0.8037007874015748,
      "calib/mu_w": 0.7954471544715448,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.29884000000000005,
      "calib/std_conf": 0.13730357023763076,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7131073446327684,
      "calib/step_q_c_n": 708.0,
      "calib/step_q_gap": -0.00922438613646237,
      "calib/step_q_w": 0.7223317307692307,
      "calib/step_q_w_n": 832.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2417.0,
      "completions/max_terminated_length": 2417.0,
      "completions/mean_length": 610.34765625,
      "completions/mean_terminated_length": 612.7412109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 141.0,
      "epoch": 0.12373333333333333,
      "grad_norm": 0.004027099348604679,
      "kl": 0.07471466064453125,
      "learning_rate": 2.3333333333333336e-06,
      "loss": 0.0287,
      "num_tokens": 27705805.0,
      "reward": 0.5554635524749756,
      "reward_std": 0.3605683445930481,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.6302086114883423,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.1877496838569641,
      "step": 116
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5348603070727409,
      "calib/avg_num_step_conf": 5.96484375,
      "calib/ece": 0.35620158102766797,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.16996047430830039,
      "calib/gap": 0.022886735464384578,
      "calib/mean_conf": 0.808296442687747,
      "calib/mu_c": 0.8206896551724138,
      "calib/mu_w": 0.7978029197080292,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.353,
      "calib/std_conf": 0.1259064645418673,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7259905362776026,
      "calib/step_q_c_n": 634.0,
      "calib/step_q_gap": -0.004468590262150962,
      "calib/step_q_w": 0.7304591265397535,
      "calib/step_q_w_n": 893.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2424.0,
      "completions/max_terminated_length": 2424.0,
      "completions/mean_length": 561.42578125,
      "completions/mean_terminated_length": 561.42578125,
      "completions/min_length": 191.0,
      "completions/min_terminated_length": 191.0,
      "epoch": 0.1248,
      "grad_norm": 0.004113308619707823,
      "kl": 0.077239990234375,
      "learning_rate": 2.305555555555556e-06,
      "loss": 0.0105,
      "num_tokens": 27956130.0,
      "reward": 0.4996691346168518,
      "reward_std": 0.37606483697891235,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6175527572631836,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.09350422024726868,
      "step": 117
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5593154915079773,
      "calib/avg_num_step_conf": 6.54296875,
      "calib/ece": 0.27880000000000005,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.14,
      "calib/gap": 0.02534096757591353,
      "calib/mean_conf": 0.7916,
      "calib/mu_c": 0.8033582089552239,
      "calib/mu_w": 0.7780172413793104,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.26720000000000005,
      "calib/std_conf": 0.14713748672585109,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7081600955794505,
      "calib/step_q_c_n": 837.0,
      "calib/step_q_gap": 0.03423408125964145,
      "calib/step_q_w": 0.673926014319809,
      "calib/step_q_w_n": 838.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 581.3671875,
      "completions/mean_terminated_length": 585.9448852539062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.12586666666666665,
      "grad_norm": 0.004203206859529018,
      "kl": 0.07547760009765625,
      "learning_rate": 2.277777777777778e-06,
      "loss": 0.0018,
      "num_tokens": 28208968.0,
      "reward": 0.5645748376846313,
      "reward_std": 0.35757777094841003,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.6591734290122986,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.17075751721858978,
      "step": 118
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5735979836168872,
      "calib/avg_num_step_conf": 5.546875,
      "calib/ece": 0.2559288537549407,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.1225296442687747,
      "calib/gap": 0.05443478260869561,
      "calib/mean_conf": 0.7843873517786562,
      "calib/mu_c": 0.8091304347826087,
      "calib/mu_w": 0.7546956521739131,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.24743083003952565,
      "calib/std_conf": 0.15600534651228387,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7256300268096514,
      "calib/step_q_c_n": 746.0,
      "calib/step_q_gap": 0.03079323155742586,
      "calib/step_q_w": 0.6948367952522255,
      "calib/step_q_w_n": 674.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 606.953125,
      "completions/mean_terminated_length": 606.953125,
      "completions/min_length": 145.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.12693333333333334,
      "grad_norm": 0.004040359519422054,
      "kl": 0.07463836669921875,
      "learning_rate": 2.25e-06,
      "loss": -0.0244,
      "num_tokens": 28469412.0,
      "reward": 0.5990918278694153,
      "reward_std": 0.38760697841644287,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.6894567012786865,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.20325830578804016,
      "step": 119
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.49453114579165,
      "calib/avg_num_step_conf": 5.5546875,
      "calib/ece": 0.20059760956175296,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.07569721115537849,
      "calib/gap": 0.00023609443777505845,
      "calib/mean_conf": 0.7691235059760957,
      "calib/mu_c": 0.7692156862745098,
      "calib/mu_w": 0.7689795918367347,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.18007968127490037,
      "calib/std_conf": 0.155783678437086,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.6988690476190476,
      "calib/step_q_c_n": 840.0,
      "calib/step_q_gap": -0.036732326951399163,
      "calib/step_q_w": 0.7356013745704467,
      "calib/step_q_w_n": 582.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2504.0,
      "completions/max_terminated_length": 2504.0,
      "completions/mean_length": 566.359375,
      "completions/mean_terminated_length": 566.359375,
      "completions/min_length": 210.0,
      "completions/min_terminated_length": 210.0,
      "epoch": 0.128,
      "grad_norm": 0.00447692908346653,
      "kl": 0.07639312744140625,
      "learning_rate": 2.222222222222222e-06,
      "loss": 0.0449,
      "num_tokens": 28721088.0,
      "reward": 0.6634672284126282,
      "reward_std": 0.33212563395500183,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.6975241899490356,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.31456655263900757,
      "step": 120
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.4832033788174139,
      "calib/avg_num_step_conf": 6.30859375,
      "calib/ece": 0.2799196787148594,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.0963855421686747,
      "calib/gap": 0.0013664717348929045,
      "calib/mean_conf": 0.7843373493975903,
      "calib/mu_c": 0.784962962962963,
      "calib/mu_w": 0.7835964912280701,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2610441767068273,
      "calib/std_conf": 0.14496247489302483,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7130761664564943,
      "calib/step_q_c_n": 793.0,
      "calib/step_q_gap": -0.0034116680933840016,
      "calib/step_q_w": 0.7164878345498783,
      "calib/step_q_w_n": 822.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2724.0,
      "completions/max_terminated_length": 2724.0,
      "completions/mean_length": 630.87109375,
      "completions/mean_terminated_length": 638.351806640625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 188.0,
      "epoch": 0.12906666666666666,
      "grad_norm": 0.00422104774042964,
      "kl": 0.0763702392578125,
      "learning_rate": 2.1944444444444445e-06,
      "loss": 0.0163,
      "num_tokens": 28987647.0,
      "reward": 0.6108550429344177,
      "reward_std": 0.3837067186832428,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.6519953012466431,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.2704959511756897,
      "step": 121
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5428459119496855,
      "calib/avg_num_step_conf": 5.67578125,
      "calib/ece": 0.20047058823529418,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.12156862745098039,
      "calib/gap": 0.0361871069182389,
      "calib/mean_conf": 0.7963137254901961,
      "calib/mu_c": 0.809937106918239,
      "calib/mu_w": 0.77375,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1866274509803922,
      "calib/std_conf": 0.13981689569367087,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7348118586088939,
      "calib/step_q_c_n": 877.0,
      "calib/step_q_gap": 0.02901324749778278,
      "calib/step_q_w": 0.7057986111111111,
      "calib/step_q_w_n": 576.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2533.0,
      "completions/max_terminated_length": 2533.0,
      "completions/mean_length": 553.546875,
      "completions/mean_terminated_length": 553.546875,
      "completions/min_length": 207.0,
      "completions/min_terminated_length": 207.0,
      "epoch": 0.13013333333333332,
      "grad_norm": 0.004603228531777859,
      "kl": 0.07869720458984375,
      "learning_rate": 2.166666666666667e-06,
      "loss": -0.0038,
      "num_tokens": 29236699.0,
      "reward": 0.6796014308929443,
      "reward_std": 0.33652737736701965,
      "rewards/accuracy_reward_step": 0.62109375,
      "rewards/final_brier_reward_step": 0.729982852935791,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.3057824969291687,
      "step": 122
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.49417317708333336,
      "calib/avg_num_step_conf": 6.140625,
      "calib/ece": 0.2578225806451613,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.0846774193548387,
      "calib/gap": 0.01717187500000006,
      "calib/mean_conf": 0.7616129032258064,
      "calib/mu_c": 0.769921875,
      "calib/mu_w": 0.7527499999999999,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2516532258064516,
      "calib/std_conf": 0.16044488098255663,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7184172661870504,
      "calib/step_q_c_n": 695.0,
      "calib/step_q_gap": 0.010948623085567988,
      "calib/step_q_w": 0.7074686431014824,
      "calib/step_q_w_n": 877.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2781.0,
      "completions/max_terminated_length": 2781.0,
      "completions/mean_length": 638.91796875,
      "completions/mean_terminated_length": 641.423583984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 210.0,
      "epoch": 0.1312,
      "grad_norm": 0.00413132319226861,
      "kl": 0.07390594482421875,
      "learning_rate": 2.138888888888889e-06,
      "loss": 0.0571,
      "num_tokens": 29505550.0,
      "reward": 0.5470514297485352,
      "reward_std": 0.35542139410972595,
      "rewards/accuracy_reward_step": 0.5,
      "rewards/final_brier_reward_step": 0.6500972509384155,
      "rewards/format_reward_step": 0.96484375,
      "rewards/step_correlation_reward": 0.15103670954704285,
      "step": 123
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5037012987012988,
      "calib/avg_num_step_conf": 5.84765625,
      "calib/ece": 0.21070866141732283,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.07874015748031496,
      "calib/gap": -0.0017311688311688611,
      "calib/mean_conf": 0.7968503937007875,
      "calib/mu_c": 0.7961688311688312,
      "calib/mu_w": 0.7979,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2006299212598425,
      "calib/std_conf": 0.1221571260497188,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7401129943502824,
      "calib/step_q_c_n": 885.0,
      "calib/step_q_gap": -0.0028281821203057778,
      "calib/step_q_w": 0.7429411764705882,
      "calib/step_q_w_n": 612.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1372.0,
      "completions/max_terminated_length": 1372.0,
      "completions/mean_length": 561.4140625,
      "completions/mean_terminated_length": 563.61572265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 88.0,
      "epoch": 0.13226666666666667,
      "grad_norm": 0.004335741512477398,
      "kl": 0.07759857177734375,
      "learning_rate": 2.1111111111111114e-06,
      "loss": 0.0041,
      "num_tokens": 29756088.0,
      "reward": 0.6853257417678833,
      "reward_std": 0.3310573697090149,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.7027504444122314,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.34836989641189575,
      "step": 124
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5575951825951827,
      "calib/avg_num_step_conf": 5.6015625,
      "calib/ece": 0.33441767068273087,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.10843373493975904,
      "calib/gap": 0.02896270396270406,
      "calib/mean_conf": 0.7895180722891566,
      "calib/mu_c": 0.804871794871795,
      "calib/mu_w": 0.775909090909091,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.3270281124497991,
      "calib/std_conf": 0.14176090183068885,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7411801242236025,
      "calib/step_q_c_n": 644.0,
      "calib/step_q_gap": 0.026273288780564563,
      "calib/step_q_w": 0.7149068354430379,
      "calib/step_q_w_n": 790.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2795.0,
      "completions/max_terminated_length": 2795.0,
      "completions/mean_length": 611.23828125,
      "completions/mean_terminated_length": 611.23828125,
      "completions/min_length": 188.0,
      "completions/min_terminated_length": 188.0,
      "epoch": 0.13333333333333333,
      "grad_norm": 0.004059802275151014,
      "kl": 0.07111358642578125,
      "learning_rate": 2.0833333333333334e-06,
      "loss": 0.0432,
      "num_tokens": 30017373.0,
      "reward": 0.5046956539154053,
      "reward_std": 0.35496553778648376,
      "rewards/accuracy_reward_step": 0.45703125,
      "rewards/final_brier_reward_step": 0.6204128861427307,
      "rewards/format_reward_step": 0.96484375,
      "rewards/step_correlation_reward": 0.10460340976715088,
      "step": 125
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5254450464396285,
      "calib/avg_num_step_conf": 5.88671875,
      "calib/ece": 0.26948,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.144,
      "calib/gap": 0.01557404540763685,
      "calib/mean_conf": 0.7735599999999999,
      "calib/mu_c": 0.7806617647058823,
      "calib/mu_w": 0.7650877192982455,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.24951999999999996,
      "calib/std_conf": 0.17800372580370333,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7304551365409622,
      "calib/step_q_c_n": 769.0,
      "calib/step_q_gap": 0.040184133830935,
      "calib/step_q_w": 0.6902710027100272,
      "calib/step_q_w_n": 738.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2935.0,
      "completions/max_terminated_length": 2935.0,
      "completions/mean_length": 597.98046875,
      "completions/mean_terminated_length": 600.3255004882812,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 184.0,
      "epoch": 0.1344,
      "grad_norm": 0.004169043619185686,
      "kl": 0.06622314453125,
      "learning_rate": 2.0555555555555555e-06,
      "loss": 0.0089,
      "num_tokens": 30275920.0,
      "reward": 0.5446716547012329,
      "reward_std": 0.34021419286727905,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6581730246543884,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.13038893043994904,
      "step": 126
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5178362573099415,
      "calib/avg_num_step_conf": 5.97265625,
      "calib/ece": 0.3217269076305221,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.09236947791164658,
      "calib/gap": 0.025413255360623888,
      "calib/mean_conf": 0.7757831325301205,
      "calib/mu_c": 0.789561403508772,
      "calib/mu_w": 0.7641481481481481,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.31983935742971886,
      "calib/std_conf": 0.14432048630633593,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7253539823008849,
      "calib/step_q_c_n": 678.0,
      "calib/step_q_gap": 0.02907901167808824,
      "calib/step_q_w": 0.6962749706227966,
      "calib/step_q_w_n": 851.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2382.0,
      "completions/max_terminated_length": 2382.0,
      "completions/mean_length": 576.84375,
      "completions/mean_terminated_length": 579.1058959960938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 185.0,
      "epoch": 0.13546666666666668,
      "grad_norm": 0.004571571014821529,
      "kl": 0.07588958740234375,
      "learning_rate": 2.027777777777778e-06,
      "loss": 0.0189,
      "num_tokens": 30527264.0,
      "reward": 0.49010002613067627,
      "reward_std": 0.3242504596710205,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.6249051094055176,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.07170122116804123,
      "step": 127
    },
    {
      "calib/answer_extract_rate": 0.96484375,
      "calib/auroc": 0.5613778298204528,
      "calib/avg_num_step_conf": 5.0859375,
      "calib/ece": 0.27471774193548393,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96484375,
      "calib/frac_conf_gt_0.9": 0.07258064516129033,
      "calib/gap": 0.04295472287275559,
      "calib/mean_conf": 0.76125,
      "calib/mu_c": 0.7823809523809523,
      "calib/mu_w": 0.7394262295081967,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2639516129032259,
      "calib/std_conf": 0.15535745073861118,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7149917491749175,
      "calib/step_q_c_n": 606.0,
      "calib/step_q_gap": 0.04333945032434283,
      "calib/step_q_w": 0.6716522988505746,
      "calib/step_q_w_n": 696.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2431.0,
      "completions/max_terminated_length": 2431.0,
      "completions/mean_length": 569.40234375,
      "completions/mean_terminated_length": 576.1541748046875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 150.0,
      "epoch": 0.13653333333333334,
      "grad_norm": 0.004363252781331539,
      "kl": 0.074676513671875,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0028,
      "num_tokens": 30779695.0,
      "reward": 0.5591524839401245,
      "reward_std": 0.387980580329895,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.6609953045845032,
      "rewards/format_reward_step": 0.96484375,
      "rewards/step_correlation_reward": 0.1659032702445984,
      "step": 128
    },
    {
      "calib/answer_extract_rate": 1.0,
      "calib/auroc": 0.5172991071428571,
      "calib/avg_num_step_conf": 5.6328125,
      "calib/ece": 0.231484375,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.08984375,
      "calib/gap": 0.004444444444444362,
      "calib/mean_conf": 0.780625,
      "calib/mu_c": 0.7825694444444444,
      "calib/mu_w": 0.7781250000000001,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.22480468750000002,
      "calib/std_conf": 0.12434358899838785,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7359466019417477,
      "calib/step_q_c_n": 824.0,
      "calib/step_q_gap": 0.0282119741100324,
      "calib/step_q_w": 0.7077346278317153,
      "calib/step_q_w_n": 618.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1344.0,
      "completions/max_terminated_length": 1344.0,
      "completions/mean_length": 508.23046875,
      "completions/mean_terminated_length": 510.22357177734375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 188.0,
      "epoch": 0.1376,
      "grad_norm": 0.004571618977934122,
      "kl": 0.08074951171875,
      "learning_rate": 1.9722222222222224e-06,
      "loss": 0.0073,
      "num_tokens": 31012186.0,
      "reward": 0.6112266182899475,
      "reward_std": 0.3370286226272583,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.6892741918563843,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.2214602380990982,
      "step": 129
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5843045843045843,
      "calib/avg_num_step_conf": 5.53515625,
      "calib/ece": 0.1876470588235294,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.07058823529411765,
      "calib/gap": 0.033640248640248616,
      "calib/mean_conf": 0.7788627450980393,
      "calib/mu_c": 0.791923076923077,
      "calib/mu_w": 0.7582828282828283,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.17737254901960783,
      "calib/std_conf": 0.13884072121310173,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7269721577726218,
      "calib/step_q_c_n": 862.0,
      "calib/step_q_gap": 0.005963148763612813,
      "calib/step_q_w": 0.721009009009009,
      "calib/step_q_w_n": 555.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1380.0,
      "completions/max_terminated_length": 1380.0,
      "completions/mean_length": 497.7421875,
      "completions/mean_terminated_length": 499.69415283203125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 182.0,
      "epoch": 0.13866666666666666,
      "grad_norm": 0.004695298615843058,
      "kl": 0.08007049560546875,
      "learning_rate": 1.944444444444445e-06,
      "loss": -0.0013,
      "num_tokens": 31244896.0,
      "reward": 0.6643092632293701,
      "reward_std": 0.2820330858230591,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.7284159660339355,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.27910876274108887,
      "step": 130
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5473628968774601,
      "calib/avg_num_step_conf": 5.40234375,
      "calib/ece": 0.3800398406374502,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.06772908366533864,
      "calib/gap": 0.010362109682498066,
      "calib/mean_conf": 0.766414342629482,
      "calib/mu_c": 0.7725242718446602,
      "calib/mu_w": 0.7621621621621621,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.36804780876494025,
      "calib/std_conf": 0.14329882560505158,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7322448979591837,
      "calib/step_q_c_n": 588.0,
      "calib/step_q_gap": 0.029490180978051672,
      "calib/step_q_w": 0.702754716981132,
      "calib/step_q_w_n": 795.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2650.0,
      "completions/max_terminated_length": 2650.0,
      "completions/mean_length": 548.11328125,
      "completions/mean_terminated_length": 550.2627563476562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 195.0,
      "epoch": 0.13973333333333332,
      "grad_norm": 0.004741629585623741,
      "kl": 0.07027435302734375,
      "learning_rate": 1.916666666666667e-06,
      "loss": -0.0036,
      "num_tokens": 31491421.0,
      "reward": 0.47297048568725586,
      "reward_std": 0.30465346574783325,
      "rewards/accuracy_reward_step": 0.40234375,
      "rewards/final_brier_reward_step": 0.6037136912345886,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.06566473841667175,
      "step": 131
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5562135922330097,
      "calib/avg_num_step_conf": 6.23046875,
      "calib/ece": 0.18948616600790508,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.06324110671936758,
      "calib/gap": 0.0357197411003235,
      "calib/mean_conf": 0.7773913043478261,
      "calib/mu_c": 0.7919333333333333,
      "calib/mu_w": 0.7562135922330098,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.18699604743082998,
      "calib/std_conf": 0.1312523782712374,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.733171974522293,
      "calib/step_q_c_n": 785.0,
      "calib/step_q_gap": 0.07576456711488555,
      "calib/step_q_w": 0.6574074074074074,
      "calib/step_q_w_n": 810.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1863.0,
      "completions/max_terminated_length": 1863.0,
      "completions/mean_length": 547.80859375,
      "completions/mean_terminated_length": 549.9569091796875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 153.0,
      "epoch": 0.1408,
      "grad_norm": 0.004259465262293816,
      "kl": 0.0728912353515625,
      "learning_rate": 1.888888888888889e-06,
      "loss": 0.0205,
      "num_tokens": 31737252.0,
      "reward": 0.6615229845046997,
      "reward_std": 0.3438699543476105,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.7125101685523987,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.2964732348918915,
      "step": 132
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5110264900662251,
      "calib/avg_num_step_conf": 5.83203125,
      "calib/ece": 0.38274900398406375,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.09561752988047809,
      "calib/gap": 0.019666225165562823,
      "calib/mean_conf": 0.7677689243027888,
      "calib/mu_c": 0.7796,
      "calib/mu_w": 0.7599337748344371,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.37605577689243025,
      "calib/std_conf": 0.1573808358108324,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7125268817204302,
      "calib/step_q_c_n": 558.0,
      "calib/step_q_gap": 0.02055896728192752,
      "calib/step_q_w": 0.6919679144385027,
      "calib/step_q_w_n": 935.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2411.0,
      "completions/max_terminated_length": 2411.0,
      "completions/mean_length": 629.44921875,
      "completions/mean_terminated_length": 634.405517578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 217.0,
      "epoch": 0.14186666666666667,
      "grad_norm": 0.004305138718336821,
      "kl": 0.06676483154296875,
      "learning_rate": 1.8611111111111113e-06,
      "loss": 0.008,
      "num_tokens": 32004735.0,
      "reward": 0.44489115476608276,
      "reward_std": 0.4102109670639038,
      "rewards/accuracy_reward_step": 0.390625,
      "rewards/final_brier_reward_step": 0.59170001745224,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.025426030158996582,
      "step": 133
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6138782051282051,
      "calib/avg_num_step_conf": 5.51171875,
      "calib/ece": 0.30723999999999996,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.092,
      "calib/gap": 0.049352564102564056,
      "calib/mean_conf": 0.78492,
      "calib/mu_c": 0.8105833333333333,
      "calib/mu_w": 0.7612307692307693,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.30607999999999996,
      "calib/std_conf": 0.12528445075108083,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7180841121495327,
      "calib/step_q_c_n": 642.0,
      "calib/step_q_gap": 0.03820114726006585,
      "calib/step_q_w": 0.6798829648894669,
      "calib/step_q_w_n": 769.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1600.0,
      "completions/max_terminated_length": 1600.0,
      "completions/mean_length": 631.734375,
      "completions/mean_terminated_length": 634.2117919921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 194.0,
      "epoch": 0.14293333333333333,
      "grad_norm": 0.004371109418570995,
      "kl": 0.0604248046875,
      "learning_rate": 1.8333333333333333e-06,
      "loss": -0.0015,
      "num_tokens": 32275411.0,
      "reward": 0.5322444438934326,
      "reward_std": 0.4109966456890106,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6493402719497681,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.1268673986196518,
      "step": 134
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5147754137115839,
      "calib/avg_num_step_conf": 5.90625,
      "calib/ece": 0.2445882352941177,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.10588235294117647,
      "calib/gap": 0.009561403508771882,
      "calib/mean_conf": 0.7823921568627451,
      "calib/mu_c": 0.7866666666666666,
      "calib/mu_w": 0.7771052631578947,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.23701960784313728,
      "calib/std_conf": 0.12859178128090068,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7156049382716049,
      "calib/step_q_c_n": 810.0,
      "calib/step_q_gap": 0.01512060778727442,
      "calib/step_q_w": 0.7004843304843305,
      "calib/step_q_w_n": 702.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1789.0,
      "completions/max_terminated_length": 1789.0,
      "completions/mean_length": 570.97265625,
      "completions/mean_terminated_length": 573.2117919921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 203.0,
      "epoch": 0.144,
      "grad_norm": 0.004456392489373684,
      "kl": 0.06621551513671875,
      "learning_rate": 1.8055555555555557e-06,
      "loss": 0.0222,
      "num_tokens": 32527460.0,
      "reward": 0.6077929735183716,
      "reward_std": 0.3696410655975342,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.6856574416160583,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.22055339813232422,
      "step": 135
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6299169859514687,
      "calib/avg_num_step_conf": 6.046875,
      "calib/ece": 0.3193227091633467,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.06374501992031872,
      "calib/gap": 0.06505810983397176,
      "calib/mean_conf": 0.7814741035856574,
      "calib/mu_c": 0.8164655172413792,
      "calib/mu_w": 0.7514074074074074,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.3193227091633467,
      "calib/std_conf": 0.13342193476332617,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7216311874105866,
      "calib/step_q_c_n": 699.0,
      "calib/step_q_gap": 0.03562412027277739,
      "calib/step_q_w": 0.6860070671378092,
      "calib/step_q_w_n": 849.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2287.0,
      "completions/max_terminated_length": 2287.0,
      "completions/mean_length": 572.515625,
      "completions/mean_terminated_length": 574.7608032226562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.14506666666666668,
      "grad_norm": 0.0045397537760436535,
      "kl": 0.06781768798828125,
      "learning_rate": 1.777777777777778e-06,
      "loss": 0.0246,
      "num_tokens": 32782512.0,
      "reward": 0.5561786890029907,
      "reward_std": 0.34354400634765625,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.6489378809928894,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.17748203873634338,
      "step": 136
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5825439361834726,
      "calib/avg_num_step_conf": 6.1640625,
      "calib/ece": 0.2304313725490197,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.07058823529411765,
      "calib/gap": 0.032907889816776725,
      "calib/mean_conf": 0.7744313725490196,
      "calib/mu_c": 0.7890140845070422,
      "calib/mu_w": 0.7561061946902655,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2240000000000001,
      "calib/std_conf": 0.13749142645669404,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7308179723502305,
      "calib/step_q_c_n": 868.0,
      "calib/step_q_gap": 0.04591656389952625,
      "calib/step_q_w": 0.6849014084507042,
      "calib/step_q_w_n": 710.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1576.0,
      "completions/max_terminated_length": 1576.0,
      "completions/mean_length": 565.7890625,
      "completions/mean_terminated_length": 568.0078735351562,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 207.0,
      "epoch": 0.14613333333333334,
      "grad_norm": 0.004167406819760799,
      "kl": 0.06789398193359375,
      "learning_rate": 1.75e-06,
      "loss": 0.035,
      "num_tokens": 33034338.0,
      "reward": 0.6721287369728088,
      "reward_std": 0.3198203444480896,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6990191340446472,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.3358633518218994,
      "step": 137
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5444679544058426,
      "calib/avg_num_step_conf": 5.47265625,
      "calib/ece": 0.15186507936507934,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.06746031746031746,
      "calib/gap": 0.02965599617773551,
      "calib/mean_conf": 0.7634523809523809,
      "calib/mu_c": 0.7741614906832298,
      "calib/mu_w": 0.7445054945054943,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.1382142857142857,
      "calib/std_conf": 0.13909994734533124,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7092874396135265,
      "calib/step_q_c_n": 828.0,
      "calib/step_q_gap": 0.04131187242329959,
      "calib/step_q_w": 0.6679755671902269,
      "calib/step_q_w_n": 573.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 547.69140625,
      "completions/mean_terminated_length": 549.8392333984375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 152.0,
      "epoch": 0.1472,
      "grad_norm": 0.004642318934202194,
      "kl": 0.06908416748046875,
      "learning_rate": 1.7222222222222224e-06,
      "loss": 0.0184,
      "num_tokens": 33278883.0,
      "reward": 0.6913992166519165,
      "reward_std": 0.3592164218425751,
      "rewards/accuracy_reward_step": 0.62890625,
      "rewards/final_brier_reward_step": 0.7364199161529541,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.3237221837043762,
      "step": 138
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5907372400756143,
      "calib/avg_num_step_conf": 5.41796875,
      "calib/ece": 0.16612648221343873,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.03557312252964427,
      "calib/gap": 0.03127329192546602,
      "calib/mean_conf": 0.7714229249011858,
      "calib/mu_c": 0.7827950310559006,
      "calib/mu_w": 0.7515217391304346,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.1505928853754941,
      "calib/std_conf": 0.12221745259785197,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7193537015276147,
      "calib/step_q_c_n": 851.0,
      "calib/step_q_gap": 0.021909671676868303,
      "calib/step_q_w": 0.6974440298507464,
      "calib/step_q_w_n": 536.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2808.0,
      "completions/max_terminated_length": 2808.0,
      "completions/mean_length": 549.0390625,
      "completions/mean_terminated_length": 549.0390625,
      "completions/min_length": 145.0,
      "completions/min_terminated_length": 145.0,
      "epoch": 0.14826666666666666,
      "grad_norm": 0.004201711155474186,
      "kl": 0.06937408447265625,
      "learning_rate": 1.6944444444444446e-06,
      "loss": 0.013,
      "num_tokens": 33522533.0,
      "reward": 0.7442376613616943,
      "reward_std": 0.28208547830581665,
      "rewards/accuracy_reward_step": 0.62890625,
      "rewards/final_brier_reward_step": 0.7411026954650879,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.42393502593040466,
      "step": 139
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6013111268603828,
      "calib/avg_num_step_conf": 5.30078125,
      "calib/ece": 0.12808764940239045,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.07569721115537849,
      "calib/gap": 0.04478454996456416,
      "calib/mean_conf": 0.7806772908366534,
      "calib/mu_c": 0.7958433734939759,
      "calib/mu_w": 0.7510588235294118,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.12370517928286853,
      "calib/std_conf": 0.12407080712845006,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7337,
      "calib/step_q_c_n": 900.0,
      "calib/step_q_gap": 0.0634374179431072,
      "calib/step_q_w": 0.6702625820568928,
      "calib/step_q_w_n": 457.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0078125,
      "completions/max_length": 1905.0,
      "completions/max_terminated_length": 1905.0,
      "completions/mean_length": 534.484375,
      "completions/mean_terminated_length": 538.6929321289062,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.14933333333333335,
      "grad_norm": 0.004485068377107382,
      "kl": 0.07003021240234375,
      "learning_rate": 1.6666666666666667e-06,
      "loss": -0.0128,
      "num_tokens": 33764377.0,
      "reward": 0.7021718621253967,
      "reward_std": 0.32926690578460693,
      "rewards/accuracy_reward_step": 0.6484375,
      "rewards/final_brier_reward_step": 0.7483788728713989,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.3309648036956787,
      "step": 140
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.62610668032487,
      "calib/avg_num_step_conf": 5.51953125,
      "calib/ece": 0.08313492063492071,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.06349206349206349,
      "calib/gap": 0.06611472890905079,
      "calib/mean_conf": 0.7601984126984127,
      "calib/mu_c": 0.7809248554913294,
      "calib/mu_w": 0.7148101265822786,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.0784126984126985,
      "calib/std_conf": 0.13927092016580372,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.69957805907173,
      "calib/step_q_c_n": 948.0,
      "calib/step_q_gap": 0.01944902681366545,
      "calib/step_q_w": 0.6801290322580645,
      "calib/step_q_w_n": 465.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2821.0,
      "completions/max_terminated_length": 2821.0,
      "completions/mean_length": 582.05078125,
      "completions/mean_terminated_length": 584.3333740234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 231.0,
      "epoch": 0.1504,
      "grad_norm": 0.004221404902637005,
      "kl": 0.06427001953125,
      "learning_rate": 1.638888888888889e-06,
      "loss": 0.0391,
      "num_tokens": 34020478.0,
      "reward": 0.7531248331069946,
      "reward_std": 0.33923524618148804,
      "rewards/accuracy_reward_step": 0.67578125,
      "rewards/final_brier_reward_step": 0.7742726802825928,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.40072697401046753,
      "step": 141
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5905033238366572,
      "calib/avg_num_step_conf": 5.61328125,
      "calib/ece": 0.25210317460317466,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.10714285714285714,
      "calib/gap": 0.044119658119658234,
      "calib/mean_conf": 0.7699603174603175,
      "calib/mu_c": 0.7904444444444445,
      "calib/mu_w": 0.7463247863247863,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.24317460317460327,
      "calib/std_conf": 0.1531015277205796,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7170850767085079,
      "calib/step_q_c_n": 717.0,
      "calib/step_q_gap": 0.03491841004184115,
      "calib/step_q_w": 0.6821666666666667,
      "calib/step_q_w_n": 720.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2188.0,
      "completions/max_terminated_length": 2188.0,
      "completions/mean_length": 600.36328125,
      "completions/mean_terminated_length": 600.36328125,
      "completions/min_length": 172.0,
      "completions/min_terminated_length": 172.0,
      "epoch": 0.15146666666666667,
      "grad_norm": 0.004155976232141256,
      "kl": 0.06311798095703125,
      "learning_rate": 1.6111111111111113e-06,
      "loss": 0.0204,
      "num_tokens": 34279331.0,
      "reward": 0.6085727214813232,
      "reward_std": 0.2957516610622406,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.6840535402297974,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.23074819147586823,
      "step": 142
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5991659821823466,
      "calib/avg_num_step_conf": 5.61328125,
      "calib/ece": 0.23373015873015876,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.015873015873015872,
      "calib/gap": 0.042750995134896,
      "calib/mean_conf": 0.7564285714285715,
      "calib/mu_c": 0.7766165413533834,
      "calib/mu_w": 0.7338655462184874,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.23119047619047625,
      "calib/std_conf": 0.1319148612258865,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7176137931034483,
      "calib/step_q_c_n": 725.0,
      "calib/step_q_gap": 0.056118357710189914,
      "calib/step_q_w": 0.6614954353932584,
      "calib/step_q_w_n": 712.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2900.0,
      "completions/max_terminated_length": 2900.0,
      "completions/mean_length": 572.58203125,
      "completions/mean_terminated_length": 574.8275146484375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 235.0,
      "epoch": 0.15253333333333333,
      "grad_norm": 0.004387116525322199,
      "kl": 0.06169891357421875,
      "learning_rate": 1.5833333333333333e-06,
      "loss": -0.0056,
      "num_tokens": 34533248.0,
      "reward": 0.63039231300354,
      "reward_std": 0.31463029980659485,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.6914234161376953,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.2685798406600952,
      "step": 143
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5462456655108029,
      "calib/avg_num_step_conf": 5.55078125,
      "calib/ece": 0.1612549019607843,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.07450980392156863,
      "calib/gap": 0.023340890904241296,
      "calib/mean_conf": 0.7551372549019608,
      "calib/mu_c": 0.763558282208589,
      "calib/mu_w": 0.7402173913043477,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.13858823529411762,
      "calib/std_conf": 0.15238200444520655,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7097377049180328,
      "calib/step_q_c_n": 915.0,
      "calib/step_q_gap": 0.03125944404846748,
      "calib/step_q_w": 0.6784782608695653,
      "calib/step_q_w_n": 506.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1782.0,
      "completions/max_terminated_length": 1782.0,
      "completions/mean_length": 536.90625,
      "completions/mean_terminated_length": 536.90625,
      "completions/min_length": 165.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.1536,
      "grad_norm": 0.005076371133327484,
      "kl": 0.0737457275390625,
      "learning_rate": 1.5555555555555558e-06,
      "loss": 0.002,
      "num_tokens": 34774824.0,
      "reward": 0.6651709079742432,
      "reward_std": 0.3426276743412018,
      "rewards/accuracy_reward_step": 0.63671875,
      "rewards/final_brier_reward_step": 0.7367910146713257,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.26776954531669617,
      "step": 144
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5868158711295967,
      "calib/avg_num_step_conf": 5.82421875,
      "calib/ece": 0.19749999999999998,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0873015873015873,
      "calib/gap": 0.030166369578134455,
      "calib/mean_conf": 0.7871031746031746,
      "calib/mu_c": 0.7989542483660131,
      "calib/mu_w": 0.7687878787878787,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.1887301587301587,
      "calib/std_conf": 0.11359887165247524,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7066727272727272,
      "calib/step_q_c_n": 847.0,
      "calib/step_q_gap": 0.0201199322416713,
      "calib/step_q_w": 0.6865527950310559,
      "calib/step_q_w_n": 644.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2102.0,
      "completions/max_terminated_length": 2102.0,
      "completions/mean_length": 545.01953125,
      "completions/mean_terminated_length": 547.1569213867188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.15466666666666667,
      "grad_norm": 0.004416238516569138,
      "kl": 0.06328582763671875,
      "learning_rate": 1.527777777777778e-06,
      "loss": 0.0301,
      "num_tokens": 35017053.0,
      "reward": 0.6805466413497925,
      "reward_std": 0.37044334411621094,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.7191644906997681,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.3255225419998169,
      "step": 145
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5401602564102564,
      "calib/avg_num_step_conf": 5.76171875,
      "calib/ece": 0.3114399999999999,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.072,
      "calib/gap": 0.01771153846153839,
      "calib/mean_conf": 0.78904,
      "calib/mu_c": 0.7982499999999999,
      "calib/mu_w": 0.7805384615384615,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3102399999999999,
      "calib/std_conf": 0.11241298145676949,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7241643059490085,
      "calib/step_q_c_n": 706.0,
      "calib/step_q_gap": 0.023071978250698844,
      "calib/step_q_w": 0.7010923276983096,
      "calib/step_q_w_n": 769.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2467.0,
      "completions/max_terminated_length": 2467.0,
      "completions/mean_length": 556.53515625,
      "completions/mean_terminated_length": 563.1343994140625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 217.0,
      "epoch": 0.15573333333333333,
      "grad_norm": 0.004498027730733156,
      "kl": 0.06241607666015625,
      "learning_rate": 1.5e-06,
      "loss": 0.0025,
      "num_tokens": 35266742.0,
      "reward": 0.5354191064834595,
      "reward_std": 0.38119083642959595,
      "rewards/accuracy_reward_step": 0.46875,
      "rewards/final_brier_reward_step": 0.6358391046524048,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.14593657851219177,
      "step": 146
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5771493212669684,
      "calib/avg_num_step_conf": 6.05859375,
      "calib/ece": 0.3080321285140561,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.05622489959839357,
      "calib/gap": 0.027821590174531008,
      "calib/mean_conf": 0.7779116465863454,
      "calib/mu_c": 0.7924369747899157,
      "calib/mu_w": 0.7646153846153847,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.30401606425702804,
      "calib/std_conf": 0.12988586364651236,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7223343373493976,
      "calib/step_q_c_n": 664.0,
      "calib/step_q_gap": 0.036347866097988235,
      "calib/step_q_w": 0.6859864712514093,
      "calib/step_q_w_n": 887.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 3035.0,
      "completions/max_terminated_length": 3035.0,
      "completions/mean_length": 591.16015625,
      "completions/mean_terminated_length": 593.4784545898438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 196.0,
      "epoch": 0.1568,
      "grad_norm": 0.004279103595763445,
      "kl": 0.059326171875,
      "learning_rate": 1.4722222222222225e-06,
      "loss": 0.0406,
      "num_tokens": 35521759.0,
      "reward": 0.4943665862083435,
      "reward_std": 0.3192962110042572,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.6381163597106934,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.0638980120420456,
      "step": 147
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5623113854595336,
      "calib/avg_num_step_conf": 5.30859375,
      "calib/ece": 0.16972222222222225,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.11507936507936507,
      "calib/gap": 0.01374074074074072,
      "calib/mean_conf": 0.7775000000000001,
      "calib/mu_c": 0.7824074074074074,
      "calib/mu_w": 0.7686666666666667,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1521825396825397,
      "calib/std_conf": 0.13714962574236156,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7106971153846153,
      "calib/step_q_c_n": 832.0,
      "calib/step_q_gap": -0.0025097157349293076,
      "calib/step_q_w": 0.7132068311195446,
      "calib/step_q_w_n": 527.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3060.0,
      "completions/max_terminated_length": 3060.0,
      "completions/mean_length": 529.66796875,
      "completions/mean_terminated_length": 529.66796875,
      "completions/min_length": 143.0,
      "completions/min_terminated_length": 143.0,
      "epoch": 0.15786666666666666,
      "grad_norm": 0.004526691976934671,
      "kl": 0.06888580322265625,
      "learning_rate": 1.4444444444444445e-06,
      "loss": 0.0661,
      "num_tokens": 35762466.0,
      "reward": 0.7067558169364929,
      "reward_std": 0.2901526391506195,
      "rewards/accuracy_reward_step": 0.6328125,
      "rewards/final_brier_reward_step": 0.7282199263572693,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.3618541955947876,
      "step": 148
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5815457725562669,
      "calib/avg_num_step_conf": 5.3828125,
      "calib/ece": 0.21729838709677418,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.953125,
      "calib/frac_conf_gt_0.9": 0.07258064516129033,
      "calib/gap": 0.05388621213121281,
      "calib/mean_conf": 0.7728629032258065,
      "calib/mu_c": 0.7965467625899281,
      "calib/mu_w": 0.7426605504587153,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.98046875,
      "calib/pce": 0.21483870967741936,
      "calib/std_conf": 0.14994982142840313,
      "calib/step_conf_rate": 0.98046875,
      "calib/step_q_c": 0.7293567954220316,
      "calib/step_q_c_n": 699.0,
      "calib/step_q_gap": 0.059842804258555815,
      "calib/step_q_w": 0.6695139911634758,
      "calib/step_q_w_n": 679.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 627.4296875,
      "completions/mean_terminated_length": 627.4296875,
      "completions/min_length": 159.0,
      "completions/min_terminated_length": 159.0,
      "epoch": 0.15893333333333334,
      "grad_norm": 0.004020726308226585,
      "kl": 0.06133270263671875,
      "learning_rate": 1.4166666666666667e-06,
      "loss": 0.0344,
      "num_tokens": 36027544.0,
      "reward": 0.6291335821151733,
      "reward_std": 0.30805811285972595,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.6803281307220459,
      "rewards/format_reward_step": 0.953125,
      "rewards/step_correlation_reward": 0.2787202000617981,
      "step": 149
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6293198113816937,
      "calib/avg_num_step_conf": 5.30859375,
      "calib/ece": 0.26472,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.12,
      "calib/gap": 0.037041534784574615,
      "calib/mean_conf": 0.80552,
      "calib/mu_c": 0.8222627737226278,
      "calib/mu_w": 0.7852212389380532,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.26112,
      "calib/std_conf": 0.11956558702235356,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.71859375,
      "calib/step_q_c_n": 704.0,
      "calib/step_q_gap": 0.02201665076335879,
      "calib/step_q_w": 0.6965770992366412,
      "calib/step_q_w_n": 655.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1897.0,
      "completions/max_terminated_length": 1897.0,
      "completions/mean_length": 496.05078125,
      "completions/mean_terminated_length": 497.99609375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 149.0,
      "epoch": 0.16,
      "grad_norm": 0.0046783071011304855,
      "kl": 0.0672607421875,
      "learning_rate": 1.3888888888888892e-06,
      "loss": 0.0248,
      "num_tokens": 36259493.0,
      "reward": 0.6306412220001221,
      "reward_std": 0.32415205240249634,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.6738687753677368,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.28507000207901,
      "step": 150
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5477752057613168,
      "calib/avg_num_step_conf": 5.30859375,
      "calib/ece": 0.36337301587301585,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.0873015873015873,
      "calib/gap": 0.02909722222222233,
      "calib/mean_conf": 0.7883730158730159,
      "calib/mu_c": 0.8050000000000002,
      "calib/mu_w": 0.7759027777777778,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.36158730158730157,
      "calib/std_conf": 0.1212244562605517,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7051403508771931,
      "calib/step_q_c_n": 570.0,
      "calib/step_q_gap": -0.02286978853979038,
      "calib/step_q_w": 0.7280101394169834,
      "calib/step_q_w_n": 789.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2477.0,
      "completions/max_terminated_length": 2477.0,
      "completions/mean_length": 567.50390625,
      "completions/mean_terminated_length": 571.972412109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 191.0,
      "epoch": 0.16106666666666666,
      "grad_norm": 0.0045959255658090115,
      "kl": 0.05704498291015625,
      "learning_rate": 1.3611111111111112e-06,
      "loss": 0.0222,
      "num_tokens": 36511798.0,
      "reward": 0.4811530113220215,
      "reward_std": 0.3167354166507721,
      "rewards/accuracy_reward_step": 0.421875,
      "rewards/final_brier_reward_step": 0.6115511655807495,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.0702861025929451,
      "step": 151
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5257395313100269,
      "calib/avg_num_step_conf": 5.3046875,
      "calib/ece": 0.33665338645418336,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.099601593625498,
      "calib/gap": 0.014992956844666394,
      "calib/mean_conf": 0.78199203187251,
      "calib/mu_c": 0.7901754385964912,
      "calib/mu_w": 0.7751824817518248,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.3322310756972113,
      "calib/std_conf": 0.1295103248366623,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7425299145299146,
      "calib/step_q_c_n": 585.0,
      "calib/step_q_gap": 0.045311285810638924,
      "calib/step_q_w": 0.6972186287192756,
      "calib/step_q_w_n": 773.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2924.0,
      "completions/max_terminated_length": 2924.0,
      "completions/mean_length": 557.63671875,
      "completions/mean_terminated_length": 557.63671875,
      "completions/min_length": 201.0,
      "completions/min_terminated_length": 201.0,
      "epoch": 0.16213333333333332,
      "grad_norm": 0.004607094451785088,
      "kl": 0.061553955078125,
      "learning_rate": 1.3333333333333334e-06,
      "loss": 0.0075,
      "num_tokens": 36759945.0,
      "reward": 0.5449575781822205,
      "reward_std": 0.35911184549331665,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.6194121241569519,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.18612799048423767,
      "step": 152
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.49685656879901485,
      "calib/avg_num_step_conf": 5.2421875,
      "calib/ece": 0.24752000000000005,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.076,
      "calib/gap": 0.002478449672694283,
      "calib/mean_conf": 0.7833600000000002,
      "calib/mu_c": 0.7844604316546763,
      "calib/mu_w": 0.781981981981982,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.98828125,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.23744000000000004,
      "calib/std_conf": 0.1306901312265008,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7036757532281205,
      "calib/step_q_c_n": 697.0,
      "calib/step_q_gap": 0.013009086561453831,
      "calib/step_q_w": 0.6906666666666667,
      "calib/step_q_w_n": 645.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2376.0,
      "completions/max_terminated_length": 2376.0,
      "completions/mean_length": 568.30859375,
      "completions/mean_terminated_length": 568.30859375,
      "completions/min_length": 184.0,
      "completions/min_terminated_length": 184.0,
      "epoch": 0.1632,
      "grad_norm": 0.00460352748632431,
      "kl": 0.06215667724609375,
      "learning_rate": 1.3055555555555556e-06,
      "loss": 0.0523,
      "num_tokens": 37012752.0,
      "reward": 0.6116618514060974,
      "reward_std": 0.3292079567909241,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.669518768787384,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.2498987466096878,
      "step": 153
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5392928535732133,
      "calib/avg_num_step_conf": 5.1953125,
      "calib/ece": 0.3261811023622046,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.06692913385826772,
      "calib/gap": 0.01358945527236377,
      "calib/mean_conf": 0.7764960629921259,
      "calib/mu_c": 0.7838793103448276,
      "calib/mu_w": 0.7702898550724638,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3229921259842518,
      "calib/std_conf": 0.13031430049550902,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7374398625429552,
      "calib/step_q_c_n": 582.0,
      "calib/step_q_gap": 0.03261365933439908,
      "calib/step_q_w": 0.7048262032085562,
      "calib/step_q_w_n": 748.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2537.0,
      "completions/max_terminated_length": 2537.0,
      "completions/mean_length": 543.01953125,
      "completions/mean_terminated_length": 543.01953125,
      "completions/min_length": 182.0,
      "completions/min_terminated_length": 182.0,
      "epoch": 0.16426666666666667,
      "grad_norm": 0.004732536617666483,
      "kl": 0.0688323974609375,
      "learning_rate": 1.2777777777777779e-06,
      "loss": 0.0078,
      "num_tokens": 37256205.0,
      "reward": 0.5141885280609131,
      "reward_std": 0.3324345648288727,
      "rewards/accuracy_reward_step": 0.453125,
      "rewards/final_brier_reward_step": 0.634368360042572,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.10494619607925415,
      "step": 154
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.47309670136711396,
      "calib/avg_num_step_conf": 5.16015625,
      "calib/ece": 0.3157312252964428,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.05928853754940711,
      "calib/gap": -0.007140348676784036,
      "calib/mean_conf": 0.7693280632411067,
      "calib/mu_c": 0.765546218487395,
      "calib/mu_w": 0.772686567164179,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.3073517786561266,
      "calib/std_conf": 0.1342345562436322,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.6943515358361775,
      "calib/step_q_c_n": 586.0,
      "calib/step_q_gap": -8.383831348246495e-05,
      "calib/step_q_w": 0.69443537414966,
      "calib/step_q_w_n": 735.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1818.0,
      "completions/max_terminated_length": 1818.0,
      "completions/mean_length": 498.0859375,
      "completions/mean_terminated_length": 502.00787353515625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.16533333333333333,
      "grad_norm": 0.004977459087967873,
      "kl": 0.069793701171875,
      "learning_rate": 1.25e-06,
      "loss": -0.0006,
      "num_tokens": 37490931.0,
      "reward": 0.5263241529464722,
      "reward_std": 0.36180227994918823,
      "rewards/accuracy_reward_step": 0.46484375,
      "rewards/final_brier_reward_step": 0.6324187517166138,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.1296045333147049,
      "step": 155
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5958965209634255,
      "calib/avg_num_step_conf": 5.5234375,
      "calib/ece": 0.23553784860557767,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.07569721115537849,
      "calib/gap": 0.06108449088823753,
      "calib/mean_conf": 0.7654183266932271,
      "calib/mu_c": 0.7941353383458647,
      "calib/mu_w": 0.7330508474576272,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.23553784860557767,
      "calib/std_conf": 0.1484478938685652,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7180300957592337,
      "calib/step_q_c_n": 731.0,
      "calib/step_q_gap": 0.01617065212819413,
      "calib/step_q_w": 0.7018594436310396,
      "calib/step_q_w_n": 683.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/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": 546.1015625,
      "completions/mean_terminated_length": 546.1015625,
      "completions/min_length": 159.0,
      "completions/min_terminated_length": 159.0,
      "epoch": 0.1664,
      "grad_norm": 0.004691540729254484,
      "kl": 0.06121063232421875,
      "learning_rate": 1.2222222222222223e-06,
      "loss": 0.0518,
      "num_tokens": 37735493.0,
      "reward": 0.6621110439300537,
      "reward_std": 0.3094662129878998,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.6900647878646851,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.3333760201931,
      "step": 156
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5312153796024763,
      "calib/avg_num_step_conf": 5.8515625,
      "calib/ece": 0.17783464566929125,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.06692913385826772,
      "calib/gap": 0.022703812316715588,
      "calib/mean_conf": 0.7747637795275589,
      "calib/mu_c": 0.7836129032258065,
      "calib/mu_w": 0.7609090909090909,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.17118110236220463,
      "calib/std_conf": 0.1261847137935653,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7137927461139896,
      "calib/step_q_c_n": 965.0,
      "calib/step_q_gap": 0.0181079431121135,
      "calib/step_q_w": 0.6956848030018761,
      "calib/step_q_w_n": 533.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2275.0,
      "completions/max_terminated_length": 2275.0,
      "completions/mean_length": 529.3515625,
      "completions/mean_terminated_length": 529.3515625,
      "completions/min_length": 155.0,
      "completions/min_terminated_length": 155.0,
      "epoch": 0.16746666666666668,
      "grad_norm": 0.004464243073016405,
      "kl": 0.0630035400390625,
      "learning_rate": 1.1944444444444446e-06,
      "loss": 0.0011,
      "num_tokens": 37974735.0,
      "reward": 0.6527853012084961,
      "reward_std": 0.35439929366111755,
      "rewards/accuracy_reward_step": 0.60546875,
      "rewards/final_brier_reward_step": 0.7229777574539185,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.26384279131889343,
      "step": 157
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.48933588761174973,
      "calib/avg_num_step_conf": 5.10546875,
      "calib/ece": 0.27366533864541837,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.04780876494023904,
      "calib/gap": -0.017674329501915942,
      "calib/mean_conf": 0.7723904382470119,
      "calib/mu_c": 0.7642222222222221,
      "calib/mu_w": 0.7818965517241381,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2541035856573705,
      "calib/std_conf": 0.13399717153837998,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7058136094674556,
      "calib/step_q_c_n": 676.0,
      "calib/step_q_gap": 0.004598078564127639,
      "calib/step_q_w": 0.701215530903328,
      "calib/step_q_w_n": 631.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2767.0,
      "completions/max_terminated_length": 2767.0,
      "completions/mean_length": 501.10546875,
      "completions/mean_terminated_length": 503.07061767578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 151.0,
      "epoch": 0.16853333333333334,
      "grad_norm": 0.004322138614952564,
      "kl": 0.0720062255859375,
      "learning_rate": 1.1666666666666668e-06,
      "loss": 0.0199,
      "num_tokens": 38208258.0,
      "reward": 0.5553004741668701,
      "reward_std": 0.3481801748275757,
      "rewards/accuracy_reward_step": 0.52734375,
      "rewards/final_brier_reward_step": 0.6541483998298645,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.15567129850387573,
      "step": 158
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6022792385560578,
      "calib/avg_num_step_conf": 4.89453125,
      "calib/ece": 0.24646825396825386,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.10317460317460317,
      "calib/gap": 0.0364238874387216,
      "calib/mean_conf": 0.7980555555555556,
      "calib/mu_c": 0.8143884892086332,
      "calib/mu_w": 0.7779646017699116,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.24646825396825386,
      "calib/std_conf": 0.10246074010462054,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7284534534534535,
      "calib/step_q_c_n": 666.0,
      "calib/step_q_gap": 0.011639143402346175,
      "calib/step_q_w": 0.7168143100511073,
      "calib/step_q_w_n": 587.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1939.0,
      "completions/max_terminated_length": 1939.0,
      "completions/mean_length": 492.5,
      "completions/mean_terminated_length": 498.3399353027344,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 201.0,
      "epoch": 0.1696,
      "grad_norm": 0.005003445781767368,
      "kl": 0.0658721923828125,
      "learning_rate": 1.138888888888889e-06,
      "loss": -0.0221,
      "num_tokens": 38439122.0,
      "reward": 0.657343864440918,
      "reward_std": 0.33716249465942383,
      "rewards/accuracy_reward_step": 0.54296875,
      "rewards/final_brier_reward_step": 0.6885058283805847,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.3207131028175354,
      "step": 159
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.4872908622908623,
      "calib/avg_num_step_conf": 5.3515625,
      "calib/ece": 0.24545816733067732,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.055776892430278883,
      "calib/gap": -0.015550193050193006,
      "calib/mean_conf": 0.7729482071713148,
      "calib/mu_c": 0.7660714285714286,
      "calib/mu_w": 0.7816216216216216,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.23031872509960163,
      "calib/std_conf": 0.12740465268063883,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.6951212553495006,
      "calib/step_q_c_n": 701.0,
      "calib/step_q_gap": -0.010289805936000151,
      "calib/step_q_w": 0.7054110612855008,
      "calib/step_q_w_n": 669.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2762.0,
      "completions/max_terminated_length": 2762.0,
      "completions/mean_length": 552.41796875,
      "completions/mean_terminated_length": 552.41796875,
      "completions/min_length": 160.0,
      "completions/min_terminated_length": 160.0,
      "epoch": 0.17066666666666666,
      "grad_norm": 0.004410970024764538,
      "kl": 0.0662689208984375,
      "learning_rate": 1.111111111111111e-06,
      "loss": 0.0435,
      "num_tokens": 38685381.0,
      "reward": 0.6169449687004089,
      "reward_std": 0.363558292388916,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6697894930839539,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.2586316764354706,
      "step": 160
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6014192785334121,
      "calib/avg_num_step_conf": 5.08203125,
      "calib/ece": 0.11787401574803148,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.07480314960629922,
      "calib/gap": 0.027143701951507948,
      "calib/mean_conf": 0.7716535433070866,
      "calib/mu_c": 0.7797752808988763,
      "calib/mu_w": 0.7526315789473683,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.09437007874015745,
      "calib/std_conf": 0.14143674348875832,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7118466522678186,
      "calib/step_q_c_n": 926.0,
      "calib/step_q_gap": 0.0029933189344852895,
      "calib/step_q_w": 0.7088533333333333,
      "calib/step_q_w_n": 375.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2132.0,
      "completions/max_terminated_length": 2132.0,
      "completions/mean_length": 498.08984375,
      "completions/mean_terminated_length": 498.08984375,
      "completions/min_length": 143.0,
      "completions/min_terminated_length": 143.0,
      "epoch": 0.17173333333333332,
      "grad_norm": 0.005129239987581968,
      "kl": 0.06465911865234375,
      "learning_rate": 1.0833333333333335e-06,
      "loss": 0.0135,
      "num_tokens": 38916812.0,
      "reward": 0.7501875758171082,
      "reward_std": 0.2995142638683319,
      "rewards/accuracy_reward_step": 0.6953125,
      "rewards/final_brier_reward_step": 0.7687234282493591,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.3949328362941742,
      "step": 161
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5519480519480519,
      "calib/avg_num_step_conf": 5.11328125,
      "calib/ece": 0.14769531249999998,
      "calib/final_conf_rate": 1.0,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.07421875,
      "calib/gap": 0.02688444888444874,
      "calib/mean_conf": 0.7823828125,
      "calib/mu_c": 0.7919393939393938,
      "calib/mu_w": 0.7650549450549451,
      "calib/nonempty_final_conf_rate": 1.0,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.1427734375,
      "calib/std_conf": 0.11652934321272836,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7153942428035044,
      "calib/step_q_c_n": 799.0,
      "calib/step_q_gap": 0.008982478097622093,
      "calib/step_q_w": 0.7064117647058823,
      "calib/step_q_w_n": 510.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1145.0,
      "completions/max_terminated_length": 1145.0,
      "completions/mean_length": 480.0625,
      "completions/mean_terminated_length": 481.94512939453125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 188.0,
      "epoch": 0.1728,
      "grad_norm": 0.004994014278054237,
      "kl": 0.06990814208984375,
      "learning_rate": 1.0555555555555557e-06,
      "loss": 0.0274,
      "num_tokens": 39143852.0,
      "reward": 0.7359399795532227,
      "reward_std": 0.3165464401245117,
      "rewards/accuracy_reward_step": 0.64453125,
      "rewards/final_brier_reward_step": 0.7494761943817139,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.39427870512008667,
      "step": 162
    },
    {
      "calib/answer_extract_rate": 0.97265625,
      "calib/auroc": 0.5603504218040234,
      "calib/avg_num_step_conf": 5.05859375,
      "calib/ece": 0.31429718875502016,
      "calib/final_conf_rate": 0.97265625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.07630522088353414,
      "calib/gap": 0.028836469824788935,
      "calib/mean_conf": 0.7699598393574297,
      "calib/mu_c": 0.7854782608695651,
      "calib/mu_w": 0.7566417910447761,
      "calib/nonempty_final_conf_rate": 0.97265625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.31120481927710847,
      "calib/std_conf": 0.13173692292223305,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7043589743589743,
      "calib/step_q_c_n": 546.0,
      "calib/step_q_gap": 0.01997980212933481,
      "calib/step_q_w": 0.6843791722296395,
      "calib/step_q_w_n": 749.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2435.0,
      "completions/max_terminated_length": 2435.0,
      "completions/mean_length": 542.75,
      "completions/mean_terminated_length": 547.0236206054688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 166.0,
      "epoch": 0.17386666666666667,
      "grad_norm": 0.004544574301689863,
      "kl": 0.06464385986328125,
      "learning_rate": 1.0277777777777777e-06,
      "loss": -0.017,
      "num_tokens": 39387628.0,
      "reward": 0.506102442741394,
      "reward_std": 0.33533090353012085,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.6356328725814819,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.09219704568386078,
      "step": 163
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.537001287001287,
      "calib/avg_num_step_conf": 5.16796875,
      "calib/ece": 0.22402390438247025,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.05179282868525897,
      "calib/gap": 0.030009009009009047,
      "calib/mean_conf": 0.7777290836653387,
      "calib/mu_c": 0.791,
      "calib/mu_w": 0.760990990990991,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.22199203187251007,
      "calib/std_conf": 0.11736152436447117,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.6988037634408601,
      "calib/step_q_c_n": 744.0,
      "calib/step_q_gap": 0.018561967240514843,
      "calib/step_q_w": 0.6802417962003453,
      "calib/step_q_w_n": 579.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2157.0,
      "completions/max_terminated_length": 2157.0,
      "completions/mean_length": 567.21875,
      "completions/mean_terminated_length": 571.68505859375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 185.0,
      "epoch": 0.17493333333333333,
      "grad_norm": 0.004481378011405468,
      "kl": 0.0655364990234375,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0257,
      "num_tokens": 39638972.0,
      "reward": 0.6318373084068298,
      "reward_std": 0.324283629655838,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6921964883804321,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.2660093605518341,
      "step": 164
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5638165662089107,
      "calib/avg_num_step_conf": 5.4453125,
      "calib/ece": 0.3218503937007874,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.07480314960629922,
      "calib/gap": 0.022084757347915263,
      "calib/mean_conf": 0.7857086614173229,
      "calib/mu_c": 0.7972727272727272,
      "calib/mu_w": 0.775187969924812,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3155905511811023,
      "calib/std_conf": 0.12280870536715315,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7209561128526646,
      "calib/step_q_c_n": 638.0,
      "calib/step_q_gap": 0.048046059942611685,
      "calib/step_q_w": 0.6729100529100529,
      "calib/step_q_w_n": 756.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2947.0,
      "completions/max_terminated_length": 2947.0,
      "completions/mean_length": 565.6875,
      "completions/mean_terminated_length": 565.6875,
      "completions/min_length": 168.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.176,
      "grad_norm": 0.0046351272612810135,
      "kl": 0.0653533935546875,
      "learning_rate": 9.722222222222224e-07,
      "loss": 0.0127,
      "num_tokens": 39889364.0,
      "reward": 0.5509390830993652,
      "reward_std": 0.3410285711288452,
      "rewards/accuracy_reward_step": 0.47265625,
      "rewards/final_brier_reward_step": 0.6437867283821106,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.16590389609336853,
      "step": 165
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.5079512324410284,
      "calib/avg_num_step_conf": 5.33984375,
      "calib/ece": 0.20920634920634926,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.06746031746031746,
      "calib/gap": 0.008144712430426582,
      "calib/mean_conf": 0.7943650793650794,
      "calib/mu_c": 0.7975324675324675,
      "calib/mu_w": 0.789387755102041,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.19623015873015878,
      "calib/std_conf": 0.12337512825773576,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7252061855670103,
      "calib/step_q_c_n": 776.0,
      "calib/step_q_gap": 0.007625813316587293,
      "calib/step_q_w": 0.717580372250423,
      "calib/step_q_w_n": 591.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2191.0,
      "completions/max_terminated_length": 2191.0,
      "completions/mean_length": 553.09765625,
      "completions/mean_terminated_length": 553.09765625,
      "completions/min_length": 93.0,
      "completions/min_terminated_length": 93.0,
      "epoch": 0.17706666666666668,
      "grad_norm": 0.004448858089745045,
      "kl": 0.05725860595703125,
      "learning_rate": 9.444444444444445e-07,
      "loss": 0.0238,
      "num_tokens": 40137141.0,
      "reward": 0.663632333278656,
      "reward_std": 0.33324968814849854,
      "rewards/accuracy_reward_step": 0.6015625,
      "rewards/final_brier_reward_step": 0.7015999555587769,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.3108208179473877,
      "step": 166
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5180018318889593,
      "calib/avg_num_step_conf": 5.32421875,
      "calib/ece": 0.15547244094488194,
      "calib/final_conf_rate": 0.9921875,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.06692913385826772,
      "calib/gap": 0.005885295568237914,
      "calib/mean_conf": 0.7946850393700787,
      "calib/mu_c": 0.7966081871345029,
      "calib/mu_w": 0.790722891566265,
      "calib/nonempty_final_conf_rate": 0.9921875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.1384645669291339,
      "calib/std_conf": 0.11801716148976306,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7187671794871795,
      "calib/step_q_c_n": 975.0,
      "calib/step_q_gap": -0.005493129791171114,
      "calib/step_q_w": 0.7242603092783506,
      "calib/step_q_w_n": 388.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2118.0,
      "completions/max_terminated_length": 2118.0,
      "completions/mean_length": 536.91796875,
      "completions/mean_terminated_length": 539.0235595703125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 203.0,
      "epoch": 0.17813333333333334,
      "grad_norm": 0.004431822337210178,
      "kl": 0.06308746337890625,
      "learning_rate": 9.166666666666666e-07,
      "loss": -0.0022,
      "num_tokens": 40380200.0,
      "reward": 0.6855368614196777,
      "reward_std": 0.31911522150039673,
      "rewards/accuracy_reward_step": 0.66796875,
      "rewards/final_brier_reward_step": 0.744168758392334,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.29565489292144775,
      "step": 167
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5900554981930819,
      "calib/avg_num_step_conf": 5.140625,
      "calib/ece": 0.17351778656126482,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.07114624505928854,
      "calib/gap": 0.058692565823438514,
      "calib/mean_conf": 0.7624505928853755,
      "calib/mu_c": 0.7865771812080538,
      "calib/mu_w": 0.7278846153846152,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.17351778656126482,
      "calib/std_conf": 0.13785884733373494,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7120529801324503,
      "calib/step_q_c_n": 755.0,
      "calib/step_q_gap": 0.056634085301790815,
      "calib/step_q_w": 0.6554188948306595,
      "calib/step_q_w_n": 561.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2766.0,
      "completions/max_terminated_length": 2766.0,
      "completions/mean_length": 560.04296875,
      "completions/mean_terminated_length": 562.2392578125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.1792,
      "grad_norm": 0.0045144446194171906,
      "kl": 0.06415557861328125,
      "learning_rate": 8.88888888888889e-07,
      "loss": 0.0411,
      "num_tokens": 40628243.0,
      "reward": 0.6357038021087646,
      "reward_std": 0.3899627923965454,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7285742163658142,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.228770911693573,
      "step": 168
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.635352366255144,
      "calib/avg_num_step_conf": 5.19140625,
      "calib/ece": 0.2372619047619047,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.05952380952380952,
      "calib/gap": 0.038148148148148264,
      "calib/mean_conf": 0.7883730158730159,
      "calib/mu_c": 0.8047222222222222,
      "calib/mu_w": 0.766574074074074,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.22710317460317456,
      "calib/std_conf": 0.13560611929156152,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7267403314917127,
      "calib/step_q_c_n": 724.0,
      "calib/step_q_gap": 0.10915355463220855,
      "calib/step_q_w": 0.6175867768595041,
      "calib/step_q_w_n": 605.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2591.0,
      "completions/max_terminated_length": 2591.0,
      "completions/mean_length": 547.890625,
      "completions/mean_terminated_length": 550.0392456054688,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 171.0,
      "epoch": 0.18026666666666666,
      "grad_norm": 0.004493064247071743,
      "kl": 0.06183624267578125,
      "learning_rate": 8.611111111111112e-07,
      "loss": 0.0265,
      "num_tokens": 40872687.0,
      "reward": 0.6389569640159607,
      "reward_std": 0.3134586811065674,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.6957355737686157,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.2735845744609833,
      "step": 169
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6196836788942052,
      "calib/avg_num_step_conf": 5.30859375,
      "calib/ece": 0.18948207171314746,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.07569721115537849,
      "calib/gap": 0.04203216374268981,
      "calib/mean_conf": 0.7943426294820717,
      "calib/mu_c": 0.8109210526315789,
      "calib/mu_w": 0.7688888888888891,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.18912350597609567,
      "calib/std_conf": 0.09476678301391761,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.702200956937799,
      "calib/step_q_c_n": 836.0,
      "calib/step_q_gap": 0.0036541118135160477,
      "calib/step_q_w": 0.698546845124283,
      "calib/step_q_w_n": 523.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1956.0,
      "completions/max_terminated_length": 1956.0,
      "completions/mean_length": 553.328125,
      "completions/mean_terminated_length": 553.328125,
      "completions/min_length": 179.0,
      "completions/min_terminated_length": 179.0,
      "epoch": 0.18133333333333335,
      "grad_norm": 0.004648624919354916,
      "kl": 0.05800628662109375,
      "learning_rate": 8.333333333333333e-07,
      "loss": 0.0087,
      "num_tokens": 41118491.0,
      "reward": 0.6535503268241882,
      "reward_std": 0.3317633271217346,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.7185636758804321,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.27447444200515747,
      "step": 170
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5978896516653954,
      "calib/avg_num_step_conf": 5.1640625,
      "calib/ece": 0.3296031746031746,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.07539682539682539,
      "calib/gap": 0.048039664378336955,
      "calib/mean_conf": 0.7732539682539683,
      "calib/mu_c": 0.7995614035087718,
      "calib/mu_w": 0.7515217391304349,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.3252380952380952,
      "calib/std_conf": 0.14069740899160643,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7059758203799654,
      "calib/step_q_c_n": 579.0,
      "calib/step_q_gap": 0.048815658872562895,
      "calib/step_q_w": 0.6571601615074025,
      "calib/step_q_w_n": 743.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2710.0,
      "completions/max_terminated_length": 2710.0,
      "completions/mean_length": 537.76953125,
      "completions/mean_terminated_length": 537.76953125,
      "completions/min_length": 170.0,
      "completions/min_terminated_length": 170.0,
      "epoch": 0.1824,
      "grad_norm": 0.004815216176211834,
      "kl": 0.0618438720703125,
      "learning_rate": 8.055555555555557e-07,
      "loss": 0.0305,
      "num_tokens": 41363056.0,
      "reward": 0.5328998565673828,
      "reward_std": 0.35678592324256897,
      "rewards/accuracy_reward_step": 0.4453125,
      "rewards/final_brier_reward_step": 0.6431062817573547,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.13675591349601746,
      "step": 171
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6530531324345757,
      "calib/avg_num_step_conf": 4.74609375,
      "calib/ece": 0.1715810276679842,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.05928853754940711,
      "calib/gap": 0.062312318265926314,
      "calib/mean_conf": 0.7850197628458498,
      "calib/mu_c": 0.8089102564102564,
      "calib/mu_w": 0.7465979381443301,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.17,
      "calib/std_conf": 0.11636328729110729,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7164840764331211,
      "calib/step_q_c_n": 785.0,
      "calib/step_q_gap": 0.03211198340986532,
      "calib/step_q_w": 0.6843720930232557,
      "calib/step_q_w_n": 430.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2483.0,
      "completions/max_terminated_length": 2483.0,
      "completions/mean_length": 509.65234375,
      "completions/mean_terminated_length": 511.6510009765625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.18346666666666667,
      "grad_norm": 0.004593435674905777,
      "kl": 0.06084442138671875,
      "learning_rate": 7.777777777777779e-07,
      "loss": 0.0067,
      "num_tokens": 41596879.0,
      "reward": 0.69252610206604,
      "reward_std": 0.3409295082092285,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.7423496246337891,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.3231714069843292,
      "step": 172
    },
    {
      "calib/answer_extract_rate": 0.9921875,
      "calib/auroc": 0.5530224075039083,
      "calib/avg_num_step_conf": 5.21875,
      "calib/ece": 0.1972727272727273,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.08300395256916997,
      "calib/gap": 0.023839239187076422,
      "calib/mean_conf": 0.7771936758893281,
      "calib/mu_c": 0.7867105263157894,
      "calib/mu_w": 0.762871287128713,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.18683794466403167,
      "calib/std_conf": 0.143737622046652,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7323192019950124,
      "calib/step_q_c_n": 802.0,
      "calib/step_q_gap": 0.02555890236954428,
      "calib/step_q_w": 0.7067602996254682,
      "calib/step_q_w_n": 534.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2323.0,
      "completions/max_terminated_length": 2323.0,
      "completions/mean_length": 558.58984375,
      "completions/mean_terminated_length": 558.58984375,
      "completions/min_length": 137.0,
      "completions/min_terminated_length": 137.0,
      "epoch": 0.18453333333333333,
      "grad_norm": 0.0045933169312775135,
      "kl": 0.0599365234375,
      "learning_rate": 7.5e-07,
      "loss": 0.0606,
      "num_tokens": 41843038.0,
      "reward": 0.6417874097824097,
      "reward_std": 0.3373648524284363,
      "rewards/accuracy_reward_step": 0.59375,
      "rewards/final_brier_reward_step": 0.7074738144874573,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.26047593355178833,
      "step": 173
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5402089101434684,
      "calib/avg_num_step_conf": 5.4140625,
      "calib/ece": 0.2594861660079052,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.08695652173913043,
      "calib/gap": 0.03405424112761135,
      "calib/mean_conf": 0.7881818181818182,
      "calib/mu_c": 0.8037956204379563,
      "calib/mu_w": 0.7697413793103449,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2530830039525692,
      "calib/std_conf": 0.1330512046868835,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.6936849132176236,
      "calib/step_q_c_n": 749.0,
      "calib/step_q_gap": 0.014925101600669066,
      "calib/step_q_w": 0.6787598116169545,
      "calib/step_q_w_n": 637.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2928.0,
      "completions/max_terminated_length": 2928.0,
      "completions/mean_length": 590.55859375,
      "completions/mean_terminated_length": 590.55859375,
      "completions/min_length": 191.0,
      "completions/min_terminated_length": 191.0,
      "epoch": 0.1856,
      "grad_norm": 0.005183553323149681,
      "kl": 0.058837890625,
      "learning_rate": 7.222222222222222e-07,
      "loss": 0.0599,
      "num_tokens": 42098453.0,
      "reward": 0.5652014017105103,
      "reward_std": 0.39359307289123535,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.6796855926513672,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.14681102335453033,
      "step": 174
    },
    {
      "calib/answer_extract_rate": 0.96875,
      "calib/auroc": 0.5470371610311349,
      "calib/avg_num_step_conf": 5.203125,
      "calib/ece": 0.34661290322580646,
      "calib/final_conf_rate": 0.96875,
      "calib/format_rate": 0.96875,
      "calib/frac_conf_gt_0.9": 0.07258064516129033,
      "calib/gap": 0.033510545698024585,
      "calib/mean_conf": 0.7541935483870967,
      "calib/mu_c": 0.7737864077669901,
      "calib/mu_w": 0.7402758620689656,
      "calib/nonempty_final_conf_rate": 0.96875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.34274193548387094,
      "calib/std_conf": 0.15626485361240294,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.6955212355212355,
      "calib/step_q_c_n": 518.0,
      "calib/step_q_gap": 0.023113373113373137,
      "calib/step_q_w": 0.6724078624078623,
      "calib/step_q_w_n": 814.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 3054.0,
      "completions/max_terminated_length": 3054.0,
      "completions/mean_length": 605.14453125,
      "completions/mean_terminated_length": 607.5177001953125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 95.0,
      "epoch": 0.18666666666666668,
      "grad_norm": 0.004421654157340527,
      "kl": 0.05449676513671875,
      "learning_rate": 6.944444444444446e-07,
      "loss": -0.0002,
      "num_tokens": 42359194.0,
      "reward": 0.48904335498809814,
      "reward_std": 0.3712538182735443,
      "rewards/accuracy_reward_step": 0.40234375,
      "rewards/final_brier_reward_step": 0.6143742203712463,
      "rewards/format_reward_step": 0.96875,
      "rewards/step_correlation_reward": 0.0894937515258789,
      "step": 175
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5567669172932331,
      "calib/avg_num_step_conf": 5.31640625,
      "calib/ece": 0.25316205533596836,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.08695652173913043,
      "calib/gap": 0.02729385964912301,
      "calib/mean_conf": 0.7712648221343873,
      "calib/mu_c": 0.7842105263157896,
      "calib/mu_w": 0.7569166666666666,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2493675889328063,
      "calib/std_conf": 0.130116932161979,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7219310344827584,
      "calib/step_q_c_n": 725.0,
      "calib/step_q_gap": 0.020563109954456538,
      "calib/step_q_w": 0.7013679245283019,
      "calib/step_q_w_n": 636.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2310.0,
      "completions/max_terminated_length": 2310.0,
      "completions/mean_length": 541.265625,
      "completions/mean_terminated_length": 541.265625,
      "completions/min_length": 171.0,
      "completions/min_terminated_length": 171.0,
      "epoch": 0.18773333333333334,
      "grad_norm": 0.0042267730459570885,
      "kl": 0.05854034423828125,
      "learning_rate": 6.666666666666667e-07,
      "loss": -0.0064,
      "num_tokens": 42601822.0,
      "reward": 0.6104744672775269,
      "reward_std": 0.31826311349868774,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.6766769886016846,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.2434908151626587,
      "step": 176
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6396663928729387,
      "calib/avg_num_step_conf": 5.16796875,
      "calib/ece": 0.26234126984126993,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.06746031746031746,
      "calib/gap": 0.04010172490048647,
      "calib/mean_conf": 0.7823412698412698,
      "calib/mu_c": 0.8012781954887217,
      "calib/mu_w": 0.7611764705882352,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.25845238095238104,
      "calib/std_conf": 0.12524450136511792,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7173524962178517,
      "calib/step_q_c_n": 661.0,
      "calib/step_q_gap": 0.04609871978280633,
      "calib/step_q_w": 0.6712537764350454,
      "calib/step_q_w_n": 662.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2584.0,
      "completions/max_terminated_length": 2584.0,
      "completions/mean_length": 549.38671875,
      "completions/mean_terminated_length": 549.38671875,
      "completions/min_length": 206.0,
      "completions/min_terminated_length": 206.0,
      "epoch": 0.1888,
      "grad_norm": 0.004867099225521088,
      "kl": 0.05817413330078125,
      "learning_rate": 6.388888888888889e-07,
      "loss": 0.0114,
      "num_tokens": 42846297.0,
      "reward": 0.6099656224250793,
      "reward_std": 0.3393414318561554,
      "rewards/accuracy_reward_step": 0.51953125,
      "rewards/final_brier_reward_step": 0.6782711148262024,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.2416602075099945,
      "step": 177
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6612476870208829,
      "calib/avg_num_step_conf": 5.16015625,
      "calib/ece": 0.17920948616600793,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.06719367588932806,
      "calib/gap": 0.06232421358710016,
      "calib/mean_conf": 0.7777075098814229,
      "calib/mu_c": 0.8016025641025641,
      "calib/mu_w": 0.739278350515464,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.17015810276679846,
      "calib/std_conf": 0.13258855428056515,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7318061088977423,
      "calib/step_q_c_n": 753.0,
      "calib/step_q_gap": 0.036225122982249336,
      "calib/step_q_w": 0.695580985915493,
      "calib/step_q_w_n": 568.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2538.0,
      "completions/max_terminated_length": 2538.0,
      "completions/mean_length": 519.85546875,
      "completions/mean_terminated_length": 521.8941650390625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 189.0,
      "epoch": 0.18986666666666666,
      "grad_norm": 0.004670095629990101,
      "kl": 0.059825897216796875,
      "learning_rate": 6.111111111111112e-07,
      "loss": 0.0078,
      "num_tokens": 43085452.0,
      "reward": 0.6979852914810181,
      "reward_std": 0.3313940167427063,
      "rewards/accuracy_reward_step": 0.609375,
      "rewards/final_brier_reward_step": 0.7407445311546326,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.33569473028182983,
      "step": 178
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6188888888888888,
      "calib/avg_num_step_conf": 5.171875,
      "calib/ece": 0.30617529880478084,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.04780876494023904,
      "calib/gap": 0.0348679365079364,
      "calib/mean_conf": 0.7941434262948208,
      "calib/mu_c": 0.8115079365079364,
      "calib/mu_w": 0.77664,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.29916334661354577,
      "calib/std_conf": 0.10948067989163843,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7024519940915805,
      "calib/step_q_c_n": 677.0,
      "calib/step_q_gap": 0.008402535049849336,
      "calib/step_q_w": 0.6940494590417312,
      "calib/step_q_w_n": 647.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2673.0,
      "completions/max_terminated_length": 2673.0,
      "completions/mean_length": 535.08203125,
      "completions/mean_terminated_length": 537.180419921875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 117.0,
      "epoch": 0.19093333333333334,
      "grad_norm": 0.004549046047031879,
      "kl": 0.06024932861328125,
      "learning_rate": 5.833333333333334e-07,
      "loss": 0.0058,
      "num_tokens": 43328697.0,
      "reward": 0.5557443499565125,
      "reward_std": 0.3748915195465088,
      "rewards/accuracy_reward_step": 0.4921875,
      "rewards/final_brier_reward_step": 0.6570112705230713,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.15994608402252197,
      "step": 179
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6261399210562134,
      "calib/avg_num_step_conf": 5.0546875,
      "calib/ece": 0.152390438247012,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.055776892430278883,
      "calib/gap": 0.049441949094868654,
      "calib/mean_conf": 0.779402390438247,
      "calib/mu_c": 0.7977215189873418,
      "calib/mu_w": 0.7482795698924731,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.15115537848605584,
      "calib/std_conf": 0.11593394118801192,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7070868014268727,
      "calib/step_q_c_n": 841.0,
      "calib/step_q_gap": 0.015033821294422434,
      "calib/step_q_w": 0.6920529801324503,
      "calib/step_q_w_n": 453.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 1670.0,
      "completions/max_terminated_length": 1670.0,
      "completions/mean_length": 563.22265625,
      "completions/mean_terminated_length": 572.1627197265625,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 192.0,
      "epoch": 0.192,
      "grad_norm": 0.004343854729086161,
      "kl": 0.0619049072265625,
      "learning_rate": 5.555555555555555e-07,
      "loss": -0.0097,
      "num_tokens": 43576738.0,
      "reward": 0.6902997493743896,
      "reward_std": 0.3136835992336273,
      "rewards/accuracy_reward_step": 0.6171875,
      "rewards/final_brier_reward_step": 0.7354074716567993,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.32644209265708923,
      "step": 180
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6090808416389811,
      "calib/avg_num_step_conf": 4.84375,
      "calib/ece": 0.28729411764705876,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.054901960784313725,
      "calib/gap": 0.033252122554448205,
      "calib/mean_conf": 0.7869803921568628,
      "calib/mu_c": 0.8034108527131784,
      "calib/mu_w": 0.7701587301587302,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.2841960784313725,
      "calib/std_conf": 0.11060572800046516,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7113057324840765,
      "calib/step_q_c_n": 628.0,
      "calib/step_q_gap": 0.02302141875858621,
      "calib/step_q_w": 0.6882843137254903,
      "calib/step_q_w_n": 612.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 1545.0,
      "completions/max_terminated_length": 1545.0,
      "completions/mean_length": 484.0,
      "completions/mean_terminated_length": 484.0,
      "completions/min_length": 198.0,
      "completions/min_terminated_length": 198.0,
      "epoch": 0.19306666666666666,
      "grad_norm": 0.005096070468425751,
      "kl": 0.0666961669921875,
      "learning_rate": 5.277777777777779e-07,
      "loss": -0.0033,
      "num_tokens": 43806906.0,
      "reward": 0.578808069229126,
      "reward_std": 0.3632180094718933,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.6692156195640564,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.18918174505233765,
      "step": 181
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5387970021966662,
      "calib/avg_num_step_conf": 5.39453125,
      "calib/ece": 0.22143426294820712,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.98046875,
      "calib/frac_conf_gt_0.9": 0.043824701195219126,
      "calib/gap": 0.024952836283757773,
      "calib/mean_conf": 0.7831075697211155,
      "calib/mu_c": 0.7939436619718311,
      "calib/mu_w": 0.7689908256880733,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.21940239043824694,
      "calib/std_conf": 0.12184138496467316,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7168958868894602,
      "calib/step_q_c_n": 778.0,
      "calib/step_q_gap": 0.009897545264252905,
      "calib/step_q_w": 0.7069983416252072,
      "calib/step_q_w_n": 603.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2210.0,
      "completions/max_terminated_length": 2210.0,
      "completions/mean_length": 522.02734375,
      "completions/mean_terminated_length": 526.1378173828125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 233.0,
      "epoch": 0.19413333333333332,
      "grad_norm": 0.0045974161475896835,
      "kl": 0.05859375,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0018,
      "num_tokens": 44046705.0,
      "reward": 0.5793590545654297,
      "reward_std": 0.3560647964477539,
      "rewards/accuracy_reward_step": 0.5546875,
      "rewards/final_brier_reward_step": 0.6907273530960083,
      "rewards/format_reward_step": 0.98046875,
      "rewards/step_correlation_reward": 0.160959392786026,
      "step": 182
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5923833703586163,
      "calib/avg_num_step_conf": 4.80859375,
      "calib/ece": 0.32178571428571423,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.0873015873015873,
      "calib/gap": 0.04948206918438591,
      "calib/mean_conf": 0.7780555555555555,
      "calib/mu_c": 0.8049565217391305,
      "calib/mu_w": 0.7554744525547445,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.32174603174603167,
      "calib/std_conf": 0.14135887042175166,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7299630996309963,
      "calib/step_q_c_n": 542.0,
      "calib/step_q_gap": 0.037394159137527416,
      "calib/step_q_w": 0.6925689404934688,
      "calib/step_q_w_n": 689.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2259.0,
      "completions/max_terminated_length": 2259.0,
      "completions/mean_length": 548.87890625,
      "completions/mean_terminated_length": 548.87890625,
      "completions/min_length": 184.0,
      "completions/min_terminated_length": 184.0,
      "epoch": 0.1952,
      "grad_norm": 0.004615063313394785,
      "kl": 0.05975341796875,
      "learning_rate": 4.7222222222222226e-07,
      "loss": 0.0156,
      "num_tokens": 44293898.0,
      "reward": 0.5459011793136597,
      "reward_std": 0.37487709522247314,
      "rewards/accuracy_reward_step": 0.44921875,
      "rewards/final_brier_reward_step": 0.642777681350708,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.16230589151382446,
      "step": 183
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5727848101265822,
      "calib/avg_num_step_conf": 5.0390625,
      "calib/ece": 0.17000000000000007,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.08300395256916997,
      "calib/gap": 0.0353637574950032,
      "calib/mean_conf": 0.7945059288537549,
      "calib/mu_c": 0.8077848101265822,
      "calib/mu_w": 0.772421052631579,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.17000000000000007,
      "calib/std_conf": 0.10058845125616318,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7023876765083441,
      "calib/step_q_c_n": 779.0,
      "calib/step_q_gap": 0.004775152046504583,
      "calib/step_q_w": 0.6976125244618395,
      "calib/step_q_w_n": 511.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2373.0,
      "completions/max_terminated_length": 2373.0,
      "completions/mean_length": 539.79296875,
      "completions/mean_terminated_length": 541.9098510742188,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.19626666666666667,
      "grad_norm": 0.0043892087414860725,
      "kl": 0.05843353271484375,
      "learning_rate": 4.444444444444445e-07,
      "loss": 0.0487,
      "num_tokens": 44537365.0,
      "reward": 0.6674075722694397,
      "reward_std": 0.39005720615386963,
      "rewards/accuracy_reward_step": 0.6171875,
      "rewards/final_brier_reward_step": 0.7324241995811462,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.2820783853530884,
      "step": 184
    },
    {
      "calib/answer_extract_rate": 0.953125,
      "calib/auroc": 0.541623631481099,
      "calib/avg_num_step_conf": 5.24609375,
      "calib/ece": 0.22389344262295074,
      "calib/final_conf_rate": 0.953125,
      "calib/format_rate": 0.94921875,
      "calib/frac_conf_gt_0.9": 0.09836065573770492,
      "calib/gap": 0.02085863802244703,
      "calib/mean_conf": 0.8017622950819672,
      "calib/mu_c": 0.8105673758865248,
      "calib/mu_w": 0.7897087378640778,
      "calib/nonempty_final_conf_rate": 0.953125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.22389344262295074,
      "calib/std_conf": 0.10399287123098268,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7203409090909091,
      "calib/step_q_c_n": 704.0,
      "calib/step_q_gap": 0.060700846493100036,
      "calib/step_q_w": 0.659640062597809,
      "calib/step_q_w_n": 639.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2798.0,
      "completions/max_terminated_length": 2798.0,
      "completions/mean_length": 593.890625,
      "completions/mean_terminated_length": 596.2196655273438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 167.0,
      "epoch": 0.19733333333333333,
      "grad_norm": 0.004370216280221939,
      "kl": 0.04863739013671875,
      "learning_rate": 4.1666666666666667e-07,
      "loss": -0.0136,
      "num_tokens": 44796321.0,
      "reward": 0.581261396408081,
      "reward_std": 0.3171704411506653,
      "rewards/accuracy_reward_step": 0.55078125,
      "rewards/final_brier_reward_step": 0.6714941263198853,
      "rewards/format_reward_step": 0.94921875,
      "rewards/step_correlation_reward": 0.19102858006954193,
      "step": 185
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5160457516339869,
      "calib/avg_num_step_conf": 5.12890625,
      "calib/ece": 0.2179446640316206,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.05533596837944664,
      "calib/gap": 0.0024692810457518366,
      "calib/mean_conf": 0.778893280632411,
      "calib/mu_c": 0.7798692810457516,
      "calib/mu_w": 0.7773999999999998,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.1960474308300396,
      "calib/std_conf": 0.15070703090545,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7165880893300248,
      "calib/step_q_c_n": 806.0,
      "calib/step_q_gap": 0.01714035757460075,
      "calib/step_q_w": 0.6994477317554241,
      "calib/step_q_w_n": 507.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2380.0,
      "completions/max_terminated_length": 2380.0,
      "completions/mean_length": 549.40234375,
      "completions/mean_terminated_length": 549.40234375,
      "completions/min_length": 175.0,
      "completions/min_terminated_length": 175.0,
      "epoch": 0.1984,
      "grad_norm": 0.004544900264590979,
      "kl": 0.0554046630859375,
      "learning_rate": 3.8888888888888895e-07,
      "loss": 0.0202,
      "num_tokens": 45042008.0,
      "reward": 0.6229519844055176,
      "reward_std": 0.3632119297981262,
      "rewards/accuracy_reward_step": 0.59765625,
      "rewards/final_brier_reward_step": 0.6970667839050293,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.23243087530136108,
      "step": 186
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5849218452154021,
      "calib/avg_num_step_conf": 5.4375,
      "calib/ece": 0.2889641434262948,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.05976095617529881,
      "calib/gap": 0.025349472614054958,
      "calib/mean_conf": 0.7901593625498008,
      "calib/mu_c": 0.8024806201550387,
      "calib/mu_w": 0.7771311475409838,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.98828125,
      "calib/pce": 0.28258964143426296,
      "calib/std_conf": 0.11937742927167588,
      "calib/step_conf_rate": 0.98828125,
      "calib/step_q_c": 0.7098208955223881,
      "calib/step_q_c_n": 670.0,
      "calib/step_q_gap": 0.0552087071567372,
      "calib/step_q_w": 0.6546121883656509,
      "calib/step_q_w_n": 722.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2619.0,
      "completions/max_terminated_length": 2619.0,
      "completions/mean_length": 599.6796875,
      "completions/mean_terminated_length": 599.6796875,
      "completions/min_length": 190.0,
      "completions/min_terminated_length": 190.0,
      "epoch": 0.19946666666666665,
      "grad_norm": 0.025320569053292274,
      "kl": 0.0690155029296875,
      "learning_rate": 3.611111111111111e-07,
      "loss": -0.0079,
      "num_tokens": 45297070.0,
      "reward": 0.5355815887451172,
      "reward_std": 0.3840063214302063,
      "rewards/accuracy_reward_step": 0.50390625,
      "rewards/final_brier_reward_step": 0.6569265127182007,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.11814286559820175,
      "step": 187
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.577579905641571,
      "calib/avg_num_step_conf": 5.28515625,
      "calib/ece": 0.19904000000000002,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.068,
      "calib/gap": 0.03922652667951365,
      "calib/mean_conf": 0.78328,
      "calib/mu_c": 0.7991275167785235,
      "calib/mu_w": 0.7599009900990098,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.19316000000000003,
      "calib/std_conf": 0.13080535768843718,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7264237516869095,
      "calib/step_q_c_n": 741.0,
      "calib/step_q_gap": 0.029773424889523836,
      "calib/step_q_w": 0.6966503267973857,
      "calib/step_q_w_n": 612.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2802.0,
      "completions/max_terminated_length": 2802.0,
      "completions/mean_length": 579.109375,
      "completions/mean_terminated_length": 583.6693115234375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 169.0,
      "epoch": 0.20053333333333334,
      "grad_norm": 0.004090098664164543,
      "kl": 0.050445556640625,
      "learning_rate": 3.3333333333333335e-07,
      "loss": -0.0056,
      "num_tokens": 45549394.0,
      "reward": 0.640600860118866,
      "reward_std": 0.3480183482170105,
      "rewards/accuracy_reward_step": 0.58203125,
      "rewards/final_brier_reward_step": 0.7073789238929749,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.26288533210754395,
      "step": 188
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.5781945130993573,
      "calib/avg_num_step_conf": 5.00390625,
      "calib/ece": 0.24784313725490187,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.9921875,
      "calib/frac_conf_gt_0.9": 0.058823529411764705,
      "calib/gap": 0.039180672268907646,
      "calib/mean_conf": 0.772156862745098,
      "calib/mu_c": 0.7904411764705882,
      "calib/mu_w": 0.7512605042016806,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.24333333333333323,
      "calib/std_conf": 0.14273125336868883,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7104017216642755,
      "calib/step_q_c_n": 697.0,
      "calib/step_q_gap": 0.03707980385605625,
      "calib/step_q_w": 0.6733219178082193,
      "calib/step_q_w_n": 584.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2940.0,
      "completions/max_terminated_length": 2940.0,
      "completions/mean_length": 546.1328125,
      "completions/mean_terminated_length": 546.1328125,
      "completions/min_length": 176.0,
      "completions/min_terminated_length": 176.0,
      "epoch": 0.2016,
      "grad_norm": 0.004661476239562035,
      "kl": 0.06034088134765625,
      "learning_rate": 3.055555555555556e-07,
      "loss": 0.0078,
      "num_tokens": 45796972.0,
      "reward": 0.5992229580879211,
      "reward_std": 0.3520023822784424,
      "rewards/accuracy_reward_step": 0.53125,
      "rewards/final_brier_reward_step": 0.6865929961204529,
      "rewards/format_reward_step": 0.9921875,
      "rewards/step_correlation_reward": 0.20716550946235657,
      "step": 189
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.6374180378572313,
      "calib/avg_num_step_conf": 5.203125,
      "calib/ece": 0.24294117647058813,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.0784313725490196,
      "calib/gap": 0.0629871334900407,
      "calib/mean_conf": 0.7801960784313725,
      "calib/mu_c": 0.8093430656934306,
      "calib/mu_w": 0.7463559322033899,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.24294117647058813,
      "calib/std_conf": 0.1258896216825136,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7085185185185185,
      "calib/step_q_c_n": 729.0,
      "calib/step_q_gap": 0.01792150359314537,
      "calib/step_q_w": 0.6905970149253732,
      "calib/step_q_w_n": 603.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 1710.0,
      "completions/max_terminated_length": 1710.0,
      "completions/mean_length": 593.7578125,
      "completions/mean_terminated_length": 596.0863037109375,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 216.0,
      "epoch": 0.20266666666666666,
      "grad_norm": 0.0045477100647985935,
      "kl": 0.05255889892578125,
      "learning_rate": 2.7777777777777776e-07,
      "loss": -0.0021,
      "num_tokens": 46054582.0,
      "reward": 0.6030627489089966,
      "reward_std": 0.3603215515613556,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.7050730586051941,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.19480247795581818,
      "step": 190
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.6133603238866397,
      "calib/avg_num_step_conf": 5.4375,
      "calib/ece": 0.33456,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.096,
      "calib/gap": 0.04238673607094667,
      "calib/mean_conf": 0.80104,
      "calib/mu_c": 0.8235897435897437,
      "calib/mu_w": 0.781203007518797,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.33380000000000004,
      "calib/std_conf": 0.11470186746518123,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7104006410256409,
      "calib/step_q_c_n": 624.0,
      "calib/step_q_gap": 0.0047495993589742325,
      "calib/step_q_w": 0.7056510416666667,
      "calib/step_q_w_n": 768.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2482.0,
      "completions/max_terminated_length": 2482.0,
      "completions/mean_length": 540.63671875,
      "completions/mean_terminated_length": 540.63671875,
      "completions/min_length": 153.0,
      "completions/min_terminated_length": 153.0,
      "epoch": 0.20373333333333332,
      "grad_norm": 0.004615450277924538,
      "kl": 0.059356689453125,
      "learning_rate": 2.5000000000000004e-07,
      "loss": 0.0366,
      "num_tokens": 46297153.0,
      "reward": 0.524100661277771,
      "reward_std": 0.2861478328704834,
      "rewards/accuracy_reward_step": 0.4609375,
      "rewards/final_brier_reward_step": 0.6304647922515869,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.13101768493652344,
      "step": 191
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.6018273471959673,
      "calib/avg_num_step_conf": 5.09375,
      "calib/ece": 0.2530434782608696,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.06324110671936758,
      "calib/gap": 0.03463768115942012,
      "calib/mean_conf": 0.7780237154150197,
      "calib/mu_c": 0.7937681159420289,
      "calib/mu_w": 0.7591304347826088,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.24280632411067196,
      "calib/std_conf": 0.12691866396896104,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7158591549295774,
      "calib/step_q_c_n": 710.0,
      "calib/step_q_gap": 0.06244164651206896,
      "calib/step_q_w": 0.6534175084175085,
      "calib/step_q_w_n": 594.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2399.0,
      "completions/max_terminated_length": 2399.0,
      "completions/mean_length": 559.578125,
      "completions/mean_terminated_length": 559.578125,
      "completions/min_length": 155.0,
      "completions/min_terminated_length": 155.0,
      "epoch": 0.2048,
      "grad_norm": 0.004527461715042591,
      "kl": 0.06067657470703125,
      "learning_rate": 2.2222222222222224e-07,
      "loss": 0.0512,
      "num_tokens": 46545381.0,
      "reward": 0.6329571604728699,
      "reward_std": 0.3509492874145508,
      "rewards/accuracy_reward_step": 0.5390625,
      "rewards/final_brier_reward_step": 0.6870129108428955,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.27421388030052185,
      "step": 192
    },
    {
      "calib/answer_extract_rate": 0.99609375,
      "calib/auroc": 0.62677842385253,
      "calib/avg_num_step_conf": 5.0078125,
      "calib/ece": 0.26494117647058824,
      "calib/final_conf_rate": 0.99609375,
      "calib/format_rate": 0.99609375,
      "calib/frac_conf_gt_0.9": 0.06666666666666667,
      "calib/gap": 0.04801002103179497,
      "calib/mean_conf": 0.7943529411764706,
      "calib/mu_c": 0.8165693430656935,
      "calib/mu_w": 0.7685593220338985,
      "calib/nonempty_final_conf_rate": 0.99609375,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 1.0,
      "calib/pce": 0.26101960784313727,
      "calib/std_conf": 0.1261817586533001,
      "calib/step_conf_rate": 1.0,
      "calib/step_q_c": 0.7286369958275383,
      "calib/step_q_c_n": 719.0,
      "calib/step_q_gap": 0.030484242719190036,
      "calib/step_q_w": 0.6981527531083482,
      "calib/step_q_w_n": 563.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 3026.0,
      "completions/max_terminated_length": 3026.0,
      "completions/mean_length": 548.2421875,
      "completions/mean_terminated_length": 548.2421875,
      "completions/min_length": 170.0,
      "completions/min_terminated_length": 170.0,
      "epoch": 0.20586666666666667,
      "grad_norm": 0.004557074513286352,
      "kl": 0.05731964111328125,
      "learning_rate": 1.9444444444444447e-07,
      "loss": 0.0423,
      "num_tokens": 46791443.0,
      "reward": 0.5993220806121826,
      "reward_std": 0.38422849774360657,
      "rewards/accuracy_reward_step": 0.53515625,
      "rewards/final_brier_reward_step": 0.6905304789543152,
      "rewards/format_reward_step": 0.99609375,
      "rewards/step_correlation_reward": 0.20186357200145721,
      "step": 193
    },
    {
      "calib/answer_extract_rate": 0.98046875,
      "calib/auroc": 0.5816294132588266,
      "calib/avg_num_step_conf": 4.99609375,
      "calib/ece": 0.29075697211155377,
      "calib/final_conf_rate": 0.98046875,
      "calib/format_rate": 0.97265625,
      "calib/frac_conf_gt_0.9": 0.07171314741035857,
      "calib/gap": 0.03610744221488449,
      "calib/mean_conf": 0.7697211155378486,
      "calib/mu_c": 0.7875590551181102,
      "calib/mu_w": 0.7514516129032257,
      "calib/nonempty_final_conf_rate": 0.98046875,
      "calib/nonempty_reasoning_rate": 1.0,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2772509960159362,
      "calib/std_conf": 0.1634693640424067,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.708179012345679,
      "calib/step_q_c_n": 648.0,
      "calib/step_q_gap": 0.0453580931697678,
      "calib/step_q_w": 0.6628209191759112,
      "calib/step_q_w_n": 631.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2415.0,
      "completions/max_terminated_length": 2415.0,
      "completions/mean_length": 519.83984375,
      "completions/mean_terminated_length": 523.9330444335938,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 174.0,
      "epoch": 0.20693333333333333,
      "grad_norm": 0.004988906439393759,
      "kl": 0.055816650390625,
      "learning_rate": 1.6666666666666668e-07,
      "loss": 0.0221,
      "num_tokens": 47030466.0,
      "reward": 0.554186999797821,
      "reward_std": 0.4383990168571472,
      "rewards/accuracy_reward_step": 0.49609375,
      "rewards/final_brier_reward_step": 0.6530015468597412,
      "rewards/format_reward_step": 0.97265625,
      "rewards/step_correlation_reward": 0.16162237524986267,
      "step": 194
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5845663265306122,
      "calib/avg_num_step_conf": 5.17578125,
      "calib/ece": 0.22968253968253974,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.07936507936507936,
      "calib/gap": 0.036374999999999824,
      "calib/mean_conf": 0.779047619047619,
      "calib/mu_c": 0.7952142857142857,
      "calib/mu_w": 0.7588392857142858,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.22658730158730164,
      "calib/std_conf": 0.12304295656801217,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7152106741573033,
      "calib/step_q_c_n": 712.0,
      "calib/step_q_gap": 0.051426008578184135,
      "calib/step_q_w": 0.6637846655791192,
      "calib/step_q_w_n": 613.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2504.0,
      "completions/max_terminated_length": 2504.0,
      "completions/mean_length": 551.60546875,
      "completions/mean_terminated_length": 551.60546875,
      "completions/min_length": 165.0,
      "completions/min_terminated_length": 165.0,
      "epoch": 0.208,
      "grad_norm": 0.004569363314658403,
      "kl": 0.0572357177734375,
      "learning_rate": 1.3888888888888888e-07,
      "loss": 0.0088,
      "num_tokens": 47277661.0,
      "reward": 0.5922529697418213,
      "reward_std": 0.3348928391933441,
      "rewards/accuracy_reward_step": 0.546875,
      "rewards/final_brier_reward_step": 0.6949304938316345,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.1833253800868988,
      "step": 195
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.5556579783852511,
      "calib/avg_num_step_conf": 4.85546875,
      "calib/ece": 0.23015810276679843,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.98828125,
      "calib/frac_conf_gt_0.9": 0.07905138339920949,
      "calib/gap": 0.03444055944055946,
      "calib/mean_conf": 0.7953754940711463,
      "calib/mu_c": 0.8103496503496503,
      "calib/mu_w": 0.7759090909090909,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.23015810276679843,
      "calib/std_conf": 0.11368928824264636,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7309170305676855,
      "calib/step_q_c_n": 687.0,
      "calib/step_q_gap": 0.039449404668404964,
      "calib/step_q_w": 0.6914676258992806,
      "calib/step_q_w_n": 556.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2226.0,
      "completions/max_terminated_length": 2226.0,
      "completions/mean_length": 487.3359375,
      "completions/mean_terminated_length": 487.3359375,
      "completions/min_length": 168.0,
      "completions/min_terminated_length": 168.0,
      "epoch": 0.20906666666666668,
      "grad_norm": 0.005070297047495842,
      "kl": 0.062015533447265625,
      "learning_rate": 1.1111111111111112e-07,
      "loss": 0.0075,
      "num_tokens": 47504963.0,
      "reward": 0.6088164448738098,
      "reward_std": 0.3506855368614197,
      "rewards/accuracy_reward_step": 0.55859375,
      "rewards/final_brier_reward_step": 0.6970175504684448,
      "rewards/format_reward_step": 0.98828125,
      "rewards/step_correlation_reward": 0.21124020218849182,
      "step": 196
    },
    {
      "calib/answer_extract_rate": 0.9765625,
      "calib/auroc": 0.6139024704065877,
      "calib/avg_num_step_conf": 5.296875,
      "calib/ece": 0.2788400000000001,
      "calib/final_conf_rate": 0.9765625,
      "calib/format_rate": 0.9765625,
      "calib/frac_conf_gt_0.9": 0.1,
      "calib/gap": 0.03703937210499231,
      "calib/mean_conf": 0.8035599999999999,
      "calib/mu_c": 0.8207462686567165,
      "calib/mu_w": 0.7837068965517242,
      "calib/nonempty_final_conf_rate": 0.9765625,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.27320000000000005,
      "calib/std_conf": 0.11233755560808682,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7165542521994136,
      "calib/step_q_c_n": 682.0,
      "calib/step_q_gap": 0.0035127091727072957,
      "calib/step_q_w": 0.7130415430267063,
      "calib/step_q_w_n": 674.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 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": 2342.0,
      "completions/max_terminated_length": 2342.0,
      "completions/mean_length": 540.3359375,
      "completions/mean_terminated_length": 546.7431030273438,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 173.0,
      "epoch": 0.21013333333333334,
      "grad_norm": 0.004491974599659443,
      "kl": 0.057811737060546875,
      "learning_rate": 8.333333333333334e-08,
      "loss": 0.0028,
      "num_tokens": 47748345.0,
      "reward": 0.5784967541694641,
      "reward_std": 0.3455749452114105,
      "rewards/accuracy_reward_step": 0.5234375,
      "rewards/final_brier_reward_step": 0.6694449186325073,
      "rewards/format_reward_step": 0.9765625,
      "rewards/step_correlation_reward": 0.18754860758781433,
      "step": 197
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.6222222222222222,
      "calib/avg_num_step_conf": 4.88671875,
      "calib/ece": 0.2284523809523809,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.06746031746031746,
      "calib/gap": 0.019258823529411884,
      "calib/mean_conf": 0.794404761904762,
      "calib/mu_c": 0.8022,
      "calib/mu_w": 0.7829411764705881,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.21380952380952375,
      "calib/std_conf": 0.12721701734097443,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7308748317631224,
      "calib/step_q_c_n": 743.0,
      "calib/step_q_gap": 0.025776406566271892,
      "calib/step_q_w": 0.7050984251968505,
      "calib/step_q_w_n": 508.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2599.0,
      "completions/max_terminated_length": 2599.0,
      "completions/mean_length": 496.71484375,
      "completions/mean_terminated_length": 498.66278076171875,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 161.0,
      "epoch": 0.2112,
      "grad_norm": 0.004678445868194103,
      "kl": 0.06182861328125,
      "learning_rate": 5.555555555555556e-08,
      "loss": 0.0301,
      "num_tokens": 47980888.0,
      "reward": 0.6651490926742554,
      "reward_std": 0.322966992855072,
      "rewards/accuracy_reward_step": 0.5859375,
      "rewards/final_brier_reward_step": 0.7013660073280334,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.3148697316646576,
      "step": 198
    },
    {
      "calib/answer_extract_rate": 0.98828125,
      "calib/auroc": 0.59375,
      "calib/avg_num_step_conf": 4.97265625,
      "calib/ece": 0.22723320158102772,
      "calib/final_conf_rate": 0.98828125,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.04743083003952569,
      "calib/gap": 0.02434378185524977,
      "calib/mean_conf": 0.7779841897233201,
      "calib/mu_c": 0.7884722222222221,
      "calib/mu_w": 0.7641284403669724,
      "calib/nonempty_final_conf_rate": 0.98828125,
      "calib/nonempty_reasoning_rate": 0.9921875,
      "calib/nonempty_step_conf_rate": 0.9921875,
      "calib/pce": 0.2180237154150198,
      "calib/std_conf": 0.13423876929138315,
      "calib/step_conf_rate": 0.9921875,
      "calib/step_q_c": 0.7115448275862067,
      "calib/step_q_c_n": 725.0,
      "calib/step_q_gap": 0.022639718097155703,
      "calib/step_q_w": 0.688905109489051,
      "calib/step_q_w_n": 548.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.0,
      "completions/max_length": 2676.0,
      "completions/max_terminated_length": 2676.0,
      "completions/mean_length": 556.4921875,
      "completions/mean_terminated_length": 556.4921875,
      "completions/min_length": 162.0,
      "completions/min_terminated_length": 162.0,
      "epoch": 0.21226666666666666,
      "grad_norm": 0.004453481640666723,
      "kl": 0.059539794921875,
      "learning_rate": 2.777777777777778e-08,
      "loss": 0.0177,
      "num_tokens": 48227550.0,
      "reward": 0.62669438123703,
      "reward_std": 0.36970627307891846,
      "rewards/accuracy_reward_step": 0.5625,
      "rewards/final_brier_reward_step": 0.6939073801040649,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.2501063346862793,
      "step": 199
    },
    {
      "calib/answer_extract_rate": 0.984375,
      "calib/auroc": 0.5132642068408295,
      "calib/avg_num_step_conf": 5.01171875,
      "calib/ece": 0.16654761904761894,
      "calib/final_conf_rate": 0.984375,
      "calib/format_rate": 0.984375,
      "calib/frac_conf_gt_0.9": 0.051587301587301584,
      "calib/gap": 0.01234715863183411,
      "calib/mean_conf": 0.7821031746031746,
      "calib/mu_c": 0.7867088607594938,
      "calib/mu_w": 0.7743617021276596,
      "calib/nonempty_final_conf_rate": 0.984375,
      "calib/nonempty_reasoning_rate": 0.99609375,
      "calib/nonempty_step_conf_rate": 0.99609375,
      "calib/pce": 0.16083333333333322,
      "calib/std_conf": 0.1195292571852399,
      "calib/step_conf_rate": 0.99609375,
      "calib/step_q_c": 0.7197509829619922,
      "calib/step_q_c_n": 763.0,
      "calib/step_q_gap": 0.041674059885069115,
      "calib/step_q_w": 0.6780769230769231,
      "calib/step_q_w_n": 520.0,
      "clip_ratio/high_max": 0.0,
      "clip_ratio/high_mean": 0.0,
      "clip_ratio/low_mean": 0.0,
      "clip_ratio/low_min": 0.0,
      "clip_ratio/region_mean": 0.0,
      "completions/clipped_ratio": 0.00390625,
      "completions/max_length": 2989.0,
      "completions/max_terminated_length": 2989.0,
      "completions/mean_length": 550.1484375,
      "completions/mean_terminated_length": 552.305908203125,
      "completions/min_length": 0.0,
      "completions/min_terminated_length": 157.0,
      "epoch": 0.21333333333333335,
      "grad_norm": 0.004446312319487333,
      "kl": 0.05374908447265625,
      "learning_rate": 0.0,
      "loss": 0.0488,
      "num_tokens": 48476436.0,
      "reward": 0.671890377998352,
      "reward_std": 0.292317271232605,
      "rewards/accuracy_reward_step": 0.6171875,
      "rewards/final_brier_reward_step": 0.722089409828186,
      "rewards/format_reward_step": 0.984375,
      "rewards/step_correlation_reward": 0.3013787865638733,
      "step": 200
    },
    {
      "epoch": 0.21333333333333335,
      "step": 200,
      "total_flos": 0.0,
      "train_loss": 0.01753797926590778,
      "train_runtime": 12763.8849,
      "train_samples_per_second": 4.011,
      "train_steps_per_second": 0.016
    }
  ],
  "logging_steps": 1,
  "max_steps": 200,
  "num_input_tokens_seen": 48476436,
  "num_train_epochs": 1,
  "save_steps": 25,
  "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
}