10235 lines
397 KiB
JSON
10235 lines
397 KiB
JSON
{
|
|
"best_metric": null,
|
|
"best_model_checkpoint": null,
|
|
"epoch": 0.21333333333333335,
|
|
"eval_steps": 500,
|
|
"global_step": 200,
|
|
"is_hyper_param_search": false,
|
|
"is_local_process_zero": true,
|
|
"is_world_process_zero": true,
|
|
"log_history": [
|
|
{
|
|
"aux_distill/lambda": 0.1,
|
|
"aux_distill/loss": 1.4131654458386558,
|
|
"aux_distill/mean_u": 0.31677682190706,
|
|
"aux_distill/n_active_tok": 24.571428571428573,
|
|
"calib/answer_extract_rate": 0.08203125,
|
|
"calib/auroc": 0.6944444444444445,
|
|
"calib/avg_num_step_conf": 0.3359375,
|
|
"calib/ece": 0.6230769230769231,
|
|
"calib/final_conf_rate": 0.05078125,
|
|
"calib/format_rate": 0.04296875,
|
|
"calib/frac_conf_gt_0.9": 0.7692307692307693,
|
|
"calib/gap": 0.03861111111111115,
|
|
"calib/mean_conf": 0.9307692307692309,
|
|
"calib/mu_c": 0.9575,
|
|
"calib/mu_w": 0.9188888888888889,
|
|
"calib/nonempty_final_conf_rate": 0.05078125,
|
|
"calib/nonempty_reasoning_rate": 0.09765625,
|
|
"calib/nonempty_step_conf_rate": 0.0703125,
|
|
"calib/pce": 0.6230769230769231,
|
|
"calib/std_conf": 0.07965903671384378,
|
|
"calib/step_conf_rate": 0.0703125,
|
|
"calib/step_q_c": 0.8921052631578947,
|
|
"calib/step_q_c_n": 19.0,
|
|
"calib/step_q_gap": 0.19807541241162607,
|
|
"calib/step_q_w": 0.6940298507462687,
|
|
"calib/step_q_w_n": 67.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.08984375,
|
|
"completions/max_length": 2955.0,
|
|
"completions/max_terminated_length": 2955.0,
|
|
"completions/mean_length": 613.67578125,
|
|
"completions/mean_terminated_length": 674.2532348632812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0010666666666666667,
|
|
"grad_norm": 0.015507475472986698,
|
|
"learning_rate": 2.5000000000000004e-07,
|
|
"loss": 0.0932,
|
|
"num_tokens": 264685.0,
|
|
"reward": 0.037574999034404755,
|
|
"reward_std": 0.07449960708618164,
|
|
"rewards/accuracy_reward_step": 0.015625,
|
|
"rewards/final_brier_reward_step": 0.01655624993145466,
|
|
"rewards/format_reward_step": 0.04296875,
|
|
"step": 1
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000003,
|
|
"aux_distill/loss": 1.1092121005058289,
|
|
"aux_distill/mean_u": 0.2935626227740425,
|
|
"aux_distill/n_active_tok": 28.63157894736842,
|
|
"calib/answer_extract_rate": 0.13671875,
|
|
"calib/auroc": 0.5338345864661654,
|
|
"calib/avg_num_step_conf": 0.55078125,
|
|
"calib/ece": 0.6261538461538463,
|
|
"calib/final_conf_rate": 0.1015625,
|
|
"calib/format_rate": 0.08984375,
|
|
"calib/frac_conf_gt_0.9": 0.7692307692307693,
|
|
"calib/gap": 0.002406015037593856,
|
|
"calib/mean_conf": 0.8953846153846153,
|
|
"calib/mu_c": 0.897142857142857,
|
|
"calib/mu_w": 0.8947368421052632,
|
|
"calib/nonempty_final_conf_rate": 0.1015625,
|
|
"calib/nonempty_reasoning_rate": 0.14453125,
|
|
"calib/nonempty_step_conf_rate": 0.109375,
|
|
"calib/pce": 0.6261538461538463,
|
|
"calib/std_conf": 0.18653172073466937,
|
|
"calib/step_conf_rate": 0.109375,
|
|
"calib/step_q_c": 0.781,
|
|
"calib/step_q_c_n": 20.0,
|
|
"calib/step_q_gap": -0.042553719008264435,
|
|
"calib/step_q_w": 0.8235537190082645,
|
|
"calib/step_q_w_n": 121.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0546875,
|
|
"completions/max_length": 3001.0,
|
|
"completions/max_terminated_length": 3001.0,
|
|
"completions/mean_length": 646.4609375,
|
|
"completions/mean_terminated_length": 683.8594970703125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0021333333333333334,
|
|
"grad_norm": 0.01369641162455082,
|
|
"learning_rate": 5.000000000000001e-07,
|
|
"loss": 0.1313,
|
|
"num_tokens": 533467.0,
|
|
"reward": 0.07537207007408142,
|
|
"reward_std": 0.14035090804100037,
|
|
"rewards/accuracy_reward_step": 0.03125,
|
|
"rewards/final_brier_reward_step": 0.02965039201080799,
|
|
"rewards/format_reward_step": 0.08984375,
|
|
"step": 2
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.09999999999999999,
|
|
"aux_distill/loss": 1.2370983511209488,
|
|
"aux_distill/mean_u": 0.36669209650917667,
|
|
"aux_distill/n_active_tok": 19.333333333333332,
|
|
"calib/answer_extract_rate": 0.06640625,
|
|
"calib/auroc": 0.4074074074074074,
|
|
"calib/avg_num_step_conf": 0.2265625,
|
|
"calib/ece": 0.6466666666666668,
|
|
"calib/final_conf_rate": 0.046875,
|
|
"calib/format_rate": 0.03125,
|
|
"calib/frac_conf_gt_0.9": 0.9166666666666666,
|
|
"calib/gap": 0.06666666666666676,
|
|
"calib/mean_conf": 0.8966666666666666,
|
|
"calib/mu_c": 0.9466666666666667,
|
|
"calib/mu_w": 0.8799999999999999,
|
|
"calib/nonempty_final_conf_rate": 0.046875,
|
|
"calib/nonempty_reasoning_rate": 0.0859375,
|
|
"calib/nonempty_step_conf_rate": 0.0546875,
|
|
"calib/pce": 0.6466666666666668,
|
|
"calib/std_conf": 0.2110028962413128,
|
|
"calib/step_conf_rate": 0.0546875,
|
|
"calib/step_q_c": 0.91625,
|
|
"calib/step_q_c_n": 8.0,
|
|
"calib/step_q_gap": 0.18605000000000005,
|
|
"calib/step_q_w": 0.7302,
|
|
"calib/step_q_w_n": 50.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.1171875,
|
|
"completions/max_length": 2955.0,
|
|
"completions/max_terminated_length": 2955.0,
|
|
"completions/mean_length": 661.203125,
|
|
"completions/mean_terminated_length": 748.9734497070312,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0032,
|
|
"grad_norm": 0.010681693442165852,
|
|
"learning_rate": 7.5e-07,
|
|
"loss": 0.076,
|
|
"num_tokens": 807991.0,
|
|
"reward": 0.0260624997317791,
|
|
"reward_std": 0.05503210425376892,
|
|
"rewards/accuracy_reward_step": 0.01171875,
|
|
"rewards/final_brier_reward_step": 0.009156249463558197,
|
|
"rewards/format_reward_step": 0.03125,
|
|
"step": 3
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000002,
|
|
"aux_distill/loss": 1.386748492717743,
|
|
"aux_distill/mean_u": 0.2923102876422463,
|
|
"aux_distill/n_active_tok": 17.0,
|
|
"calib/answer_extract_rate": 0.09765625,
|
|
"calib/auroc": 0.3854166666666667,
|
|
"calib/avg_num_step_conf": 0.265625,
|
|
"calib/ece": 0.6525,
|
|
"calib/final_conf_rate": 0.0625,
|
|
"calib/format_rate": 0.04296875,
|
|
"calib/frac_conf_gt_0.9": 0.75,
|
|
"calib/gap": 0.043333333333333335,
|
|
"calib/mean_conf": 0.9025000000000001,
|
|
"calib/mu_c": 0.935,
|
|
"calib/mu_w": 0.8916666666666667,
|
|
"calib/nonempty_final_conf_rate": 0.0625,
|
|
"calib/nonempty_reasoning_rate": 0.11328125,
|
|
"calib/nonempty_step_conf_rate": 0.0625,
|
|
"calib/pce": 0.6525,
|
|
"calib/std_conf": 0.16532165617365438,
|
|
"calib/step_conf_rate": 0.0625,
|
|
"calib/step_q_c": 0.6809375,
|
|
"calib/step_q_c_n": 16.0,
|
|
"calib/step_q_gap": 0.021129807692307767,
|
|
"calib/step_q_w": 0.6598076923076922,
|
|
"calib/step_q_w_n": 52.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0859375,
|
|
"completions/max_length": 3017.0,
|
|
"completions/max_terminated_length": 3017.0,
|
|
"completions/mean_length": 736.09765625,
|
|
"completions/mean_terminated_length": 805.303466796875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 3.0,
|
|
"epoch": 0.004266666666666667,
|
|
"grad_norm": 0.01607530750334263,
|
|
"learning_rate": 1.0000000000000002e-06,
|
|
"loss": 0.0794,
|
|
"num_tokens": 1102600.0,
|
|
"reward": 0.03820156306028366,
|
|
"reward_std": 0.08263486623764038,
|
|
"rewards/accuracy_reward_step": 0.015625,
|
|
"rewards/final_brier_reward_step": 0.017809374257922173,
|
|
"rewards/format_reward_step": 0.04296875,
|
|
"step": 4
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.09999999999999999,
|
|
"aux_distill/loss": 1.2863253951072693,
|
|
"aux_distill/mean_u": 0.4327708304064644,
|
|
"aux_distill/n_active_tok": 22.0,
|
|
"calib/answer_extract_rate": 0.0390625,
|
|
"calib/avg_num_step_conf": 0.1484375,
|
|
"calib/ece": 0.9733333333333333,
|
|
"calib/final_conf_rate": 0.0234375,
|
|
"calib/format_rate": 0.0234375,
|
|
"calib/frac_conf_gt_0.9": 1.0,
|
|
"calib/mean_conf": 0.9733333333333333,
|
|
"calib/mu_c": NaN,
|
|
"calib/mu_w": 0.9733333333333333,
|
|
"calib/nonempty_final_conf_rate": 0.0234375,
|
|
"calib/nonempty_reasoning_rate": 0.046875,
|
|
"calib/nonempty_step_conf_rate": 0.03125,
|
|
"calib/pce": 0.9733333333333333,
|
|
"calib/std_conf": 0.019720265943665407,
|
|
"calib/step_conf_rate": 0.03125,
|
|
"calib/step_q_w": 0.8034210526315789,
|
|
"calib/step_q_w_n": 38.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.10546875,
|
|
"completions/max_length": 3010.0,
|
|
"completions/max_terminated_length": 3010.0,
|
|
"completions/mean_length": 674.91796875,
|
|
"completions/mean_terminated_length": 754.4934692382812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.005333333333333333,
|
|
"grad_norm": 0.006632169242948294,
|
|
"learning_rate": 1.25e-06,
|
|
"loss": 0.0243,
|
|
"num_tokens": 1382067.0,
|
|
"reward": 0.0123308589681983,
|
|
"reward_std": 0.02698388881981373,
|
|
"rewards/accuracy_reward_step": 0.0,
|
|
"rewards/final_brier_reward_step": 0.0012242187513038516,
|
|
"rewards/format_reward_step": 0.0234375,
|
|
"step": 5
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.09999999999999999,
|
|
"aux_distill/loss": 1.314890056848526,
|
|
"aux_distill/mean_u": 0.29406630225016933,
|
|
"aux_distill/n_active_tok": 16.666666666666668,
|
|
"calib/answer_extract_rate": 0.03515625,
|
|
"calib/auroc": 0.8333333333333334,
|
|
"calib/avg_num_step_conf": 0.1953125,
|
|
"calib/ece": 0.7971428571428572,
|
|
"calib/final_conf_rate": 0.02734375,
|
|
"calib/format_rate": 0.02734375,
|
|
"calib/frac_conf_gt_0.9": 0.8571428571428571,
|
|
"calib/gap": 0.03500000000000003,
|
|
"calib/mean_conf": 0.9399999999999998,
|
|
"calib/mu_c": 0.97,
|
|
"calib/mu_w": 0.9349999999999999,
|
|
"calib/nonempty_final_conf_rate": 0.02734375,
|
|
"calib/nonempty_reasoning_rate": 0.05859375,
|
|
"calib/nonempty_step_conf_rate": 0.0546875,
|
|
"calib/pce": 0.7971428571428572,
|
|
"calib/std_conf": 0.0427617987059879,
|
|
"calib/step_conf_rate": 0.0546875,
|
|
"calib/step_q_c": 0.884,
|
|
"calib/step_q_c_n": 5.0,
|
|
"calib/step_q_gap": 0.18955555555555548,
|
|
"calib/step_q_w": 0.6944444444444445,
|
|
"calib/step_q_w_n": 45.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.1015625,
|
|
"completions/max_length": 2984.0,
|
|
"completions/max_terminated_length": 2984.0,
|
|
"completions/mean_length": 685.56640625,
|
|
"completions/mean_terminated_length": 763.065185546875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0064,
|
|
"grad_norm": 0.012052838690578938,
|
|
"learning_rate": 1.5e-06,
|
|
"loss": 0.0596,
|
|
"num_tokens": 1663524.0,
|
|
"reward": 0.01902734488248825,
|
|
"reward_std": 0.04970274120569229,
|
|
"rewards/accuracy_reward_step": 0.00390625,
|
|
"rewards/final_brier_reward_step": 0.006804687902331352,
|
|
"rewards/format_reward_step": 0.02734375,
|
|
"step": 6
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.1,
|
|
"aux_distill/loss": 1.2238177359104156,
|
|
"aux_distill/mean_u": 0.4388083803543682,
|
|
"aux_distill/n_active_tok": 35.714285714285715,
|
|
"calib/answer_extract_rate": 0.07421875,
|
|
"calib/auroc": 0.9230769230769231,
|
|
"calib/avg_num_step_conf": 0.5234375,
|
|
"calib/ece": 0.6964428571428573,
|
|
"calib/final_conf_rate": 0.0546875,
|
|
"calib/format_rate": 0.03515625,
|
|
"calib/frac_conf_gt_0.9": 0.7857142857142857,
|
|
"calib/gap": 0.2176769230769231,
|
|
"calib/mean_conf": 0.7678714285714286,
|
|
"calib/mu_c": 0.97,
|
|
"calib/mu_w": 0.7523230769230769,
|
|
"calib/nonempty_final_conf_rate": 0.0546875,
|
|
"calib/nonempty_reasoning_rate": 0.11328125,
|
|
"calib/nonempty_step_conf_rate": 0.08203125,
|
|
"calib/pce": 0.6964428571428573,
|
|
"calib/std_conf": 0.3521092168805652,
|
|
"calib/step_conf_rate": 0.08203125,
|
|
"calib/step_q_c": 0.882,
|
|
"calib/step_q_c_n": 5.0,
|
|
"calib/step_q_gap": 0.059165891472868215,
|
|
"calib/step_q_w": 0.8228341085271318,
|
|
"calib/step_q_w_n": 129.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.11328125,
|
|
"completions/max_length": 2990.0,
|
|
"completions/max_terminated_length": 2990.0,
|
|
"completions/mean_length": 674.9375,
|
|
"completions/mean_terminated_length": 761.1629638671875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.007466666666666667,
|
|
"grad_norm": 0.009223215281963348,
|
|
"learning_rate": 1.75e-06,
|
|
"loss": 0.0493,
|
|
"num_tokens": 1943732.0,
|
|
"reward": 0.026374993845820427,
|
|
"reward_std": 0.05906704068183899,
|
|
"rewards/accuracy_reward_step": 0.00390625,
|
|
"rewards/final_brier_reward_step": 0.01368748489767313,
|
|
"rewards/format_reward_step": 0.03515625,
|
|
"step": 7
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000002,
|
|
"aux_distill/loss": 1.3814922496676445,
|
|
"aux_distill/mean_u": 0.30116149651308344,
|
|
"aux_distill/n_active_tok": 23.75,
|
|
"calib/answer_extract_rate": 0.0859375,
|
|
"calib/auroc": 0.6444444444444444,
|
|
"calib/avg_num_step_conf": 0.37109375,
|
|
"calib/ece": 0.7566722222222222,
|
|
"calib/final_conf_rate": 0.0703125,
|
|
"calib/format_rate": 0.046875,
|
|
"calib/frac_conf_gt_0.9": 0.8888888888888888,
|
|
"calib/gap": 0.05332666666666641,
|
|
"calib/mean_conf": 0.9088944444444444,
|
|
"calib/mu_c": 0.9533333333333333,
|
|
"calib/mu_w": 0.9000066666666668,
|
|
"calib/nonempty_final_conf_rate": 0.0703125,
|
|
"calib/nonempty_reasoning_rate": 0.1171875,
|
|
"calib/nonempty_step_conf_rate": 0.08984375,
|
|
"calib/pce": 0.74945,
|
|
"calib/std_conf": 0.2199744946635158,
|
|
"calib/step_conf_rate": 0.08984375,
|
|
"calib/step_q_c": 0.7177777777777777,
|
|
"calib/step_q_c_n": 9.0,
|
|
"calib/step_q_gap": -0.13652454780361756,
|
|
"calib/step_q_w": 0.8543023255813953,
|
|
"calib/step_q_w_n": 86.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0703125,
|
|
"completions/max_length": 3067.0,
|
|
"completions/max_terminated_length": 3067.0,
|
|
"completions/mean_length": 631.5,
|
|
"completions/mean_terminated_length": 679.2605590820312,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.008533333333333334,
|
|
"grad_norm": 0.01828446239233017,
|
|
"learning_rate": 2.0000000000000003e-06,
|
|
"loss": 0.0922,
|
|
"num_tokens": 2211908.0,
|
|
"reward": 0.03814628720283508,
|
|
"reward_std": 0.09642170369625092,
|
|
"rewards/accuracy_reward_step": 0.01171875,
|
|
"rewards/final_brier_reward_step": 0.017698820680379868,
|
|
"rewards/format_reward_step": 0.046875,
|
|
"step": 8
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000002,
|
|
"aux_distill/loss": 1.3213333524763584,
|
|
"aux_distill/mean_u": 0.27625369501018854,
|
|
"aux_distill/n_active_tok": 20.75,
|
|
"calib/answer_extract_rate": 0.09375,
|
|
"calib/auroc": 0.0,
|
|
"calib/avg_num_step_conf": 0.32421875,
|
|
"calib/ece": 0.9483333333333334,
|
|
"calib/final_conf_rate": 0.0703125,
|
|
"calib/format_rate": 0.0625,
|
|
"calib/frac_conf_gt_0.9": 0.7777777777777778,
|
|
"calib/gap": -0.9170588235294118,
|
|
"calib/mean_conf": 0.8961111111111113,
|
|
"calib/mu_c": 0.03,
|
|
"calib/mu_w": 0.9470588235294118,
|
|
"calib/nonempty_final_conf_rate": 0.0703125,
|
|
"calib/nonempty_reasoning_rate": 0.09765625,
|
|
"calib/nonempty_step_conf_rate": 0.0703125,
|
|
"calib/pce": 0.8944444444444445,
|
|
"calib/std_conf": 0.21587390581049057,
|
|
"calib/step_conf_rate": 0.0703125,
|
|
"calib/step_q_w": 0.7922891566265059,
|
|
"calib/step_q_w_n": 83.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0859375,
|
|
"completions/max_length": 3043.0,
|
|
"completions/max_terminated_length": 3043.0,
|
|
"completions/mean_length": 716.25,
|
|
"completions/mean_terminated_length": 783.5897827148438,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 3.0,
|
|
"epoch": 0.0096,
|
|
"grad_norm": 0.012646260671317577,
|
|
"learning_rate": 2.25e-06,
|
|
"loss": 0.0973,
|
|
"num_tokens": 2502804.0,
|
|
"reward": 0.03833867236971855,
|
|
"reward_std": 0.08556064963340759,
|
|
"rewards/accuracy_reward_step": 0.0078125,
|
|
"rewards/final_brier_reward_step": 0.006364844273775816,
|
|
"rewards/format_reward_step": 0.0625,
|
|
"step": 9
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.1,
|
|
"aux_distill/loss": 1.3350291166986739,
|
|
"aux_distill/mean_u": 0.2897335086370673,
|
|
"aux_distill/n_active_tok": 20.857142857142858,
|
|
"calib/answer_extract_rate": 0.08984375,
|
|
"calib/auroc": 0.9833333333333333,
|
|
"calib/avg_num_step_conf": 0.30078125,
|
|
"calib/ece": 0.766470588235294,
|
|
"calib/final_conf_rate": 0.06640625,
|
|
"calib/format_rate": 0.046875,
|
|
"calib/frac_conf_gt_0.9": 0.7647058823529411,
|
|
"calib/gap": 0.1256666666666666,
|
|
"calib/mean_conf": 0.8841176470588235,
|
|
"calib/mu_c": 0.995,
|
|
"calib/mu_w": 0.8693333333333334,
|
|
"calib/nonempty_final_conf_rate": 0.06640625,
|
|
"calib/nonempty_reasoning_rate": 0.10546875,
|
|
"calib/nonempty_step_conf_rate": 0.0703125,
|
|
"calib/pce": 0.766470588235294,
|
|
"calib/std_conf": 0.15751970272764243,
|
|
"calib/step_conf_rate": 0.0703125,
|
|
"calib/step_q_c": 0.3822222222222222,
|
|
"calib/step_q_c_n": 9.0,
|
|
"calib/step_q_gap": -0.31917777777777784,
|
|
"calib/step_q_w": 0.7014,
|
|
"calib/step_q_w_n": 68.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.046875,
|
|
"completions/max_length": 2752.0,
|
|
"completions/max_terminated_length": 2752.0,
|
|
"completions/mean_length": 672.25390625,
|
|
"completions/mean_terminated_length": 705.3155517578125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 3.0,
|
|
"epoch": 0.010666666666666666,
|
|
"grad_norm": 0.013075040653347969,
|
|
"learning_rate": 2.5e-06,
|
|
"loss": 0.0647,
|
|
"num_tokens": 2781701.0,
|
|
"reward": 0.03389336168766022,
|
|
"reward_std": 0.0721963569521904,
|
|
"rewards/accuracy_reward_step": 0.0078125,
|
|
"rewards/final_brier_reward_step": 0.013099219650030136,
|
|
"rewards/format_reward_step": 0.046875,
|
|
"step": 10
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000003,
|
|
"aux_distill/loss": 1.1985960386016152,
|
|
"aux_distill/mean_u": 0.28704307273318935,
|
|
"aux_distill/n_active_tok": 27.09090909090909,
|
|
"calib/answer_extract_rate": 0.12890625,
|
|
"calib/auroc": 0.5052631578947369,
|
|
"calib/avg_num_step_conf": 0.58203125,
|
|
"calib/ece": 0.6941666666666666,
|
|
"calib/final_conf_rate": 0.09375,
|
|
"calib/format_rate": 0.07421875,
|
|
"calib/frac_conf_gt_0.9": 0.7916666666666666,
|
|
"calib/gap": 0.054947368421052634,
|
|
"calib/mean_conf": 0.9025,
|
|
"calib/mu_c": 0.946,
|
|
"calib/mu_w": 0.8910526315789473,
|
|
"calib/nonempty_final_conf_rate": 0.09375,
|
|
"calib/nonempty_reasoning_rate": 0.171875,
|
|
"calib/nonempty_step_conf_rate": 0.12109375,
|
|
"calib/pce": 0.6941666666666666,
|
|
"calib/std_conf": 0.18166475534162738,
|
|
"calib/step_conf_rate": 0.12109375,
|
|
"calib/step_q_c": 0.83,
|
|
"calib/step_q_c_n": 16.0,
|
|
"calib/step_q_gap": 0.16220827067669163,
|
|
"calib/step_q_w": 0.6677917293233083,
|
|
"calib/step_q_w_n": 133.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.09375,
|
|
"completions/max_length": 3016.0,
|
|
"completions/max_terminated_length": 3016.0,
|
|
"completions/mean_length": 653.48828125,
|
|
"completions/mean_terminated_length": 721.0905151367188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.011733333333333333,
|
|
"grad_norm": 0.01548672839999199,
|
|
"learning_rate": 2.7500000000000004e-06,
|
|
"loss": 0.1246,
|
|
"num_tokens": 3053474.0,
|
|
"reward": 0.060481246560811996,
|
|
"reward_std": 0.1259399652481079,
|
|
"rewards/accuracy_reward_step": 0.01953125,
|
|
"rewards/final_brier_reward_step": 0.02721250057220459,
|
|
"rewards/format_reward_step": 0.07421875,
|
|
"step": 11
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000003,
|
|
"aux_distill/loss": 1.195373563390029,
|
|
"aux_distill/mean_u": 0.29857195077858234,
|
|
"aux_distill/n_active_tok": 25.68421052631579,
|
|
"calib/answer_extract_rate": 0.140625,
|
|
"calib/auroc": 0.5746753246753247,
|
|
"calib/avg_num_step_conf": 0.4765625,
|
|
"calib/ece": 0.5375999999999999,
|
|
"calib/final_conf_rate": 0.09765625,
|
|
"calib/format_rate": 0.08203125,
|
|
"calib/frac_conf_gt_0.9": 0.8,
|
|
"calib/gap": -0.05090909090909079,
|
|
"calib/mean_conf": 0.8976000000000001,
|
|
"calib/mu_c": 0.8690909090909091,
|
|
"calib/mu_w": 0.9199999999999999,
|
|
"calib/nonempty_final_conf_rate": 0.09765625,
|
|
"calib/nonempty_reasoning_rate": 0.1640625,
|
|
"calib/nonempty_step_conf_rate": 0.109375,
|
|
"calib/pce": 0.49759999999999993,
|
|
"calib/std_conf": 0.19996559704109104,
|
|
"calib/step_conf_rate": 0.109375,
|
|
"calib/step_q_c": 0.6972122448979592,
|
|
"calib/step_q_c_n": 49.0,
|
|
"calib/step_q_gap": -0.03566446743080787,
|
|
"calib/step_q_w": 0.732876712328767,
|
|
"calib/step_q_w_n": 73.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0703125,
|
|
"completions/max_length": 3038.0,
|
|
"completions/max_terminated_length": 3038.0,
|
|
"completions/mean_length": 608.76171875,
|
|
"completions/mean_terminated_length": 654.8025512695312,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 3.0,
|
|
"epoch": 0.0128,
|
|
"grad_norm": 0.013475374318659306,
|
|
"learning_rate": 3e-06,
|
|
"loss": 0.1106,
|
|
"num_tokens": 3313493.0,
|
|
"reward": 0.08334805071353912,
|
|
"reward_std": 0.15470468997955322,
|
|
"rewards/accuracy_reward_step": 0.04296875,
|
|
"rewards/final_brier_reward_step": 0.04169609397649765,
|
|
"rewards/format_reward_step": 0.08203125,
|
|
"step": 12
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000002,
|
|
"aux_distill/loss": 1.172689814120531,
|
|
"aux_distill/mean_u": 0.23420449887686123,
|
|
"aux_distill/n_active_tok": 32.25,
|
|
"calib/answer_extract_rate": 0.125,
|
|
"calib/auroc": 0.828125,
|
|
"calib/avg_num_step_conf": 0.5078125,
|
|
"calib/ece": 0.610505,
|
|
"calib/final_conf_rate": 0.078125,
|
|
"calib/format_rate": 0.0546875,
|
|
"calib/frac_conf_gt_0.9": 0.65,
|
|
"calib/gap": 0.21499374999999998,
|
|
"calib/mean_conf": 0.810505,
|
|
"calib/mu_c": 0.9824999999999999,
|
|
"calib/mu_w": 0.76750625,
|
|
"calib/nonempty_final_conf_rate": 0.078125,
|
|
"calib/nonempty_reasoning_rate": 0.16015625,
|
|
"calib/nonempty_step_conf_rate": 0.09765625,
|
|
"calib/pce": 0.610505,
|
|
"calib/std_conf": 0.2966255981452039,
|
|
"calib/step_conf_rate": 0.09765625,
|
|
"calib/step_q_c": 0.8213636363636364,
|
|
"calib/step_q_c_n": 22.0,
|
|
"calib/step_q_gap": 0.17941919191919187,
|
|
"calib/step_q_w": 0.6419444444444445,
|
|
"calib/step_q_w_n": 108.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0703125,
|
|
"completions/max_length": 3004.0,
|
|
"completions/max_terminated_length": 3004.0,
|
|
"completions/mean_length": 654.80859375,
|
|
"completions/mean_terminated_length": 704.3319702148438,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 3.0,
|
|
"epoch": 0.013866666666666666,
|
|
"grad_norm": 0.010491629131138325,
|
|
"learning_rate": 3.2500000000000002e-06,
|
|
"loss": 0.0957,
|
|
"num_tokens": 3585716.0,
|
|
"reward": 0.050288278609514236,
|
|
"reward_std": 0.11872527003288269,
|
|
"rewards/accuracy_reward_step": 0.015625,
|
|
"rewards/final_brier_reward_step": 0.030264053493738174,
|
|
"rewards/format_reward_step": 0.0546875,
|
|
"step": 13
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000003,
|
|
"aux_distill/loss": 1.0641672429831133,
|
|
"aux_distill/mean_u": 0.18141969879736022,
|
|
"aux_distill/n_active_tok": 26.608695652173914,
|
|
"calib/answer_extract_rate": 0.15234375,
|
|
"calib/auroc": 0.38461538461538464,
|
|
"calib/avg_num_step_conf": 0.6171875,
|
|
"calib/ece": 0.5858064516129031,
|
|
"calib/final_conf_rate": 0.12109375,
|
|
"calib/format_rate": 0.09375,
|
|
"calib/frac_conf_gt_0.9": 0.5161290322580645,
|
|
"calib/gap": 0.036615384615384605,
|
|
"calib/mean_conf": 0.7212903225806453,
|
|
"calib/mu_c": 0.752,
|
|
"calib/mu_w": 0.7153846153846154,
|
|
"calib/nonempty_final_conf_rate": 0.12109375,
|
|
"calib/nonempty_reasoning_rate": 0.1953125,
|
|
"calib/nonempty_step_conf_rate": 0.140625,
|
|
"calib/pce": 0.5729032258064515,
|
|
"calib/std_conf": 0.34307661079427937,
|
|
"calib/step_conf_rate": 0.140625,
|
|
"calib/step_q_c": 0.40700000000000003,
|
|
"calib/step_q_c_n": 10.0,
|
|
"calib/step_q_gap": -0.2302304054054054,
|
|
"calib/step_q_w": 0.6372304054054054,
|
|
"calib/step_q_w_n": 148.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.09375,
|
|
"completions/max_length": 3038.0,
|
|
"completions/max_terminated_length": 3038.0,
|
|
"completions/mean_length": 624.78515625,
|
|
"completions/mean_terminated_length": 689.4180908203125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.014933333333333333,
|
|
"grad_norm": 0.014551612548530102,
|
|
"learning_rate": 3.5e-06,
|
|
"loss": 0.1489,
|
|
"num_tokens": 3851061.0,
|
|
"reward": 0.07813125103712082,
|
|
"reward_std": 0.14755792915821075,
|
|
"rewards/accuracy_reward_step": 0.01953125,
|
|
"rewards/final_brier_reward_step": 0.04298124834895134,
|
|
"rewards/format_reward_step": 0.09375,
|
|
"step": 14
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000003,
|
|
"aux_distill/loss": 1.0421833992004395,
|
|
"aux_distill/mean_u": 0.31752556933961956,
|
|
"aux_distill/n_active_tok": 37.833333333333336,
|
|
"calib/answer_extract_rate": 0.1640625,
|
|
"calib/auroc": 0.7388888888888889,
|
|
"calib/avg_num_step_conf": 0.890625,
|
|
"calib/ece": 0.6410090909090909,
|
|
"calib/final_conf_rate": 0.12890625,
|
|
"calib/format_rate": 0.10546875,
|
|
"calib/frac_conf_gt_0.9": 0.5151515151515151,
|
|
"calib/gap": 0.2032233333333331,
|
|
"calib/mean_conf": 0.7319181818181819,
|
|
"calib/mu_c": 0.9166666666666666,
|
|
"calib/mu_w": 0.7134433333333335,
|
|
"calib/nonempty_final_conf_rate": 0.12890625,
|
|
"calib/nonempty_reasoning_rate": 0.23046875,
|
|
"calib/nonempty_step_conf_rate": 0.1796875,
|
|
"calib/pce": 0.6410090909090909,
|
|
"calib/std_conf": 0.32096754622551416,
|
|
"calib/step_conf_rate": 0.1796875,
|
|
"calib/step_q_c": 0.66375,
|
|
"calib/step_q_c_n": 16.0,
|
|
"calib/step_q_gap": 0.06805419811320756,
|
|
"calib/step_q_w": 0.5956958018867924,
|
|
"calib/step_q_w_n": 212.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0859375,
|
|
"completions/max_length": 2996.0,
|
|
"completions/max_terminated_length": 2996.0,
|
|
"completions/mean_length": 639.1953125,
|
|
"completions/mean_terminated_length": 699.2906494140625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.016,
|
|
"grad_norm": 0.012739659287035465,
|
|
"learning_rate": 3.7500000000000005e-06,
|
|
"loss": 0.1424,
|
|
"num_tokens": 4122575.0,
|
|
"reward": 0.0783761590719223,
|
|
"reward_std": 0.13948485255241394,
|
|
"rewards/accuracy_reward_step": 0.01171875,
|
|
"rewards/final_brier_reward_step": 0.0395648255944252,
|
|
"rewards/format_reward_step": 0.10546875,
|
|
"step": 15
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000003,
|
|
"aux_distill/loss": 1.1064559129568248,
|
|
"aux_distill/mean_u": 0.41832564748230244,
|
|
"aux_distill/n_active_tok": 35.53846153846154,
|
|
"calib/answer_extract_rate": 0.16796875,
|
|
"calib/auroc": 0.2633333333333333,
|
|
"calib/avg_num_step_conf": 0.90625,
|
|
"calib/ece": 0.6767142857142856,
|
|
"calib/final_conf_rate": 0.13671875,
|
|
"calib/format_rate": 0.125,
|
|
"calib/frac_conf_gt_0.9": 0.4857142857142857,
|
|
"calib/gap": -0.16349999999999998,
|
|
"calib/mean_conf": 0.7361428571428571,
|
|
"calib/mu_c": 0.596,
|
|
"calib/mu_w": 0.7595,
|
|
"calib/nonempty_final_conf_rate": 0.13671875,
|
|
"calib/nonempty_reasoning_rate": 0.22265625,
|
|
"calib/nonempty_step_conf_rate": 0.1875,
|
|
"calib/pce": 0.6349999999999999,
|
|
"calib/std_conf": 0.32873803229555143,
|
|
"calib/step_conf_rate": 0.1875,
|
|
"calib/step_q_c": 0.5223076923076923,
|
|
"calib/step_q_c_n": 13.0,
|
|
"calib/step_q_gap": -0.0027028099754127233,
|
|
"calib/step_q_w": 0.525010502283105,
|
|
"calib/step_q_w_n": 219.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0859375,
|
|
"completions/max_length": 2960.0,
|
|
"completions/max_terminated_length": 2960.0,
|
|
"completions/mean_length": 688.58984375,
|
|
"completions/mean_terminated_length": 753.3291015625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 4.0,
|
|
"epoch": 0.017066666666666667,
|
|
"grad_norm": 0.011877968907356262,
|
|
"learning_rate": 4.000000000000001e-06,
|
|
"loss": 0.1549,
|
|
"num_tokens": 4407702.0,
|
|
"reward": 0.09559077024459839,
|
|
"reward_std": 0.1968509405851364,
|
|
"rewards/accuracy_reward_step": 0.01953125,
|
|
"rewards/final_brier_reward_step": 0.04665029048919678,
|
|
"rewards/format_reward_step": 0.125,
|
|
"step": 16
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 1.0695910771687827,
|
|
"aux_distill/mean_u": 0.37407696141474706,
|
|
"aux_distill/n_active_tok": 40.53333333333333,
|
|
"calib/answer_extract_rate": 0.21875,
|
|
"calib/auroc": 0.5897435897435898,
|
|
"calib/avg_num_step_conf": 1.1875,
|
|
"calib/ece": 0.4930021739130434,
|
|
"calib/final_conf_rate": 0.1796875,
|
|
"calib/format_rate": 0.15234375,
|
|
"calib/frac_conf_gt_0.9": 0.30434782608695654,
|
|
"calib/gap": 0.1684736263736264,
|
|
"calib/mean_conf": 0.6243065217391305,
|
|
"calib/mu_c": 0.7671428571428571,
|
|
"calib/mu_w": 0.5986692307692307,
|
|
"calib/nonempty_final_conf_rate": 0.1796875,
|
|
"calib/nonempty_reasoning_rate": 0.28125,
|
|
"calib/nonempty_step_conf_rate": 0.234375,
|
|
"calib/pce": 0.4825673913043478,
|
|
"calib/std_conf": 0.3615274530451221,
|
|
"calib/step_conf_rate": 0.234375,
|
|
"calib/step_q_c": 0.43399999999999994,
|
|
"calib/step_q_c_n": 30.0,
|
|
"calib/step_q_gap": -0.11178827838827843,
|
|
"calib/step_q_w": 0.5457882783882784,
|
|
"calib/step_q_w_n": 273.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0390625,
|
|
"completions/max_length": 2863.0,
|
|
"completions/max_terminated_length": 2863.0,
|
|
"completions/mean_length": 667.96875,
|
|
"completions/mean_terminated_length": 695.1219482421875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 4.0,
|
|
"epoch": 0.018133333333333335,
|
|
"grad_norm": 0.012100381776690483,
|
|
"learning_rate": 4.25e-06,
|
|
"loss": 0.2091,
|
|
"num_tokens": 4682230.0,
|
|
"reward": 0.131727933883667,
|
|
"reward_std": 0.2430960088968277,
|
|
"rewards/accuracy_reward_step": 0.02734375,
|
|
"rewards/final_brier_reward_step": 0.08376838266849518,
|
|
"rewards/format_reward_step": 0.15234375,
|
|
"step": 17
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 1.0483861842325755,
|
|
"aux_distill/mean_u": 0.31695819481401377,
|
|
"aux_distill/n_active_tok": 40.285714285714285,
|
|
"calib/answer_extract_rate": 0.20703125,
|
|
"calib/auroc": 0.13,
|
|
"calib/avg_num_step_conf": 1.12109375,
|
|
"calib/ece": 0.62823214,
|
|
"calib/final_conf_rate": 0.17578125,
|
|
"calib/format_rate": 0.15234375,
|
|
"calib/frac_conf_gt_0.9": 0.2222222222222222,
|
|
"calib/gap": -0.4473961575,
|
|
"calib/mean_conf": 0.5786854733333333,
|
|
"calib/mu_c": 0.181,
|
|
"calib/mu_w": 0.6283961575,
|
|
"calib/nonempty_final_conf_rate": 0.17578125,
|
|
"calib/nonempty_reasoning_rate": 0.265625,
|
|
"calib/nonempty_step_conf_rate": 0.2265625,
|
|
"calib/pce": 0.547903251111111,
|
|
"calib/std_conf": 0.3337314558672589,
|
|
"calib/step_conf_rate": 0.2265625,
|
|
"calib/step_q_c": 0.5427272727272728,
|
|
"calib/step_q_c_n": 11.0,
|
|
"calib/step_q_gap": 0.03277944664031618,
|
|
"calib/step_q_w": 0.5099478260869567,
|
|
"calib/step_q_w_n": 276.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.078125,
|
|
"completions/max_length": 2990.0,
|
|
"completions/max_terminated_length": 2990.0,
|
|
"completions/mean_length": 634.51171875,
|
|
"completions/mean_terminated_length": 688.2838745117188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0192,
|
|
"grad_norm": 0.011530191637575626,
|
|
"learning_rate": 4.5e-06,
|
|
"loss": 0.2147,
|
|
"num_tokens": 4955385.0,
|
|
"reward": 0.125144824385643,
|
|
"reward_std": 0.23584964871406555,
|
|
"rewards/accuracy_reward_step": 0.01953125,
|
|
"rewards/final_brier_reward_step": 0.07841463387012482,
|
|
"rewards/format_reward_step": 0.15234375,
|
|
"step": 18
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9958725925534964,
|
|
"aux_distill/mean_u": 0.4017864188827994,
|
|
"aux_distill/n_active_tok": 75.375,
|
|
"calib/answer_extract_rate": 0.41796875,
|
|
"calib/auroc": 0.7012195121951219,
|
|
"calib/avg_num_step_conf": 2.375,
|
|
"calib/ece": 0.4502235294117647,
|
|
"calib/final_conf_rate": 0.33203125,
|
|
"calib/format_rate": 0.2734375,
|
|
"calib/frac_conf_gt_0.9": 0.27058823529411763,
|
|
"calib/gap": 0.33635365853658544,
|
|
"calib/mean_conf": 0.4855176470588235,
|
|
"calib/mu_c": 0.81,
|
|
"calib/mu_w": 0.4736463414634146,
|
|
"calib/nonempty_final_conf_rate": 0.33203125,
|
|
"calib/nonempty_reasoning_rate": 0.6015625,
|
|
"calib/nonempty_step_conf_rate": 0.49609375,
|
|
"calib/pce": 0.4502235294117647,
|
|
"calib/std_conf": 0.38048375776618304,
|
|
"calib/step_conf_rate": 0.49609375,
|
|
"calib/step_q_c": 0.7280555555555555,
|
|
"calib/step_q_c_n": 9.0,
|
|
"calib/step_q_gap": 0.3991562233351882,
|
|
"calib/step_q_w": 0.32889933222036727,
|
|
"calib/step_q_w_n": 599.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0234375,
|
|
"completions/max_length": 3063.0,
|
|
"completions/max_terminated_length": 3063.0,
|
|
"completions/mean_length": 574.859375,
|
|
"completions/mean_terminated_length": 588.656005859375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 7.0,
|
|
"epoch": 0.020266666666666665,
|
|
"grad_norm": 0.015343696810305119,
|
|
"learning_rate": 4.75e-06,
|
|
"loss": 0.2044,
|
|
"num_tokens": 5207309.0,
|
|
"reward": 0.22790691256523132,
|
|
"reward_std": 0.3663272261619568,
|
|
"rewards/accuracy_reward_step": 0.01171875,
|
|
"rewards/final_brier_reward_step": 0.17065757513046265,
|
|
"rewards/format_reward_step": 0.2734375,
|
|
"step": 19
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 1.0093897189944983,
|
|
"aux_distill/mean_u": 0.33957365230949094,
|
|
"aux_distill/n_active_tok": 107.375,
|
|
"calib/answer_extract_rate": 0.55078125,
|
|
"calib/auroc": 0.4638888888888889,
|
|
"calib/avg_num_step_conf": 3.390625,
|
|
"calib/ece": 0.3564631012784062,
|
|
"calib/final_conf_rate": 0.5,
|
|
"calib/format_rate": 0.40625,
|
|
"calib/frac_conf_gt_0.9": 0.171875,
|
|
"calib/gap": -0.056053082996629644,
|
|
"calib/mean_conf": 0.43178028877840624,
|
|
"calib/mu_c": 0.3844855,
|
|
"calib/mu_w": 0.4405385829966296,
|
|
"calib/nonempty_final_conf_rate": 0.5,
|
|
"calib/nonempty_reasoning_rate": 0.765625,
|
|
"calib/nonempty_step_conf_rate": 0.66796875,
|
|
"calib/pce": 0.3159966950284062,
|
|
"calib/std_conf": 0.3514771066323164,
|
|
"calib/step_conf_rate": 0.66796875,
|
|
"calib/step_q_c": 0.3547614457831325,
|
|
"calib/step_q_c_n": 83.0,
|
|
"calib/step_q_gap": 0.05394622469469712,
|
|
"calib/step_q_w": 0.3008152210884354,
|
|
"calib/step_q_w_n": 784.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.02734375,
|
|
"completions/max_length": 2984.0,
|
|
"completions/max_terminated_length": 2984.0,
|
|
"completions/mean_length": 487.01953125,
|
|
"completions/mean_terminated_length": 500.7108154296875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 3.0,
|
|
"epoch": 0.021333333333333333,
|
|
"grad_norm": 0.015337585471570492,
|
|
"learning_rate": 5e-06,
|
|
"loss": 0.2183,
|
|
"num_tokens": 5436858.0,
|
|
"reward": 0.38433361053466797,
|
|
"reward_std": 0.43774229288101196,
|
|
"rewards/accuracy_reward_step": 0.08203125,
|
|
"rewards/final_brier_reward_step": 0.28038597106933594,
|
|
"rewards/format_reward_step": 0.40625,
|
|
"step": 20
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9585423450917006,
|
|
"aux_distill/mean_u": 0.3275177277533846,
|
|
"aux_distill/n_active_tok": 124.125,
|
|
"calib/answer_extract_rate": 0.57421875,
|
|
"calib/auroc": 0.4688228438228439,
|
|
"calib/avg_num_step_conf": 3.88671875,
|
|
"calib/ece": 0.33572305747126435,
|
|
"calib/final_conf_rate": 0.56640625,
|
|
"calib/format_rate": 0.44140625,
|
|
"calib/frac_conf_gt_0.9": 0.09655172413793103,
|
|
"calib/gap": -0.04217678904428901,
|
|
"calib/mean_conf": 0.3948056781609196,
|
|
"calib/mu_c": 0.3564102564102564,
|
|
"calib/mu_w": 0.39858704545454543,
|
|
"calib/nonempty_final_conf_rate": 0.56640625,
|
|
"calib/nonempty_reasoning_rate": 0.7890625,
|
|
"calib/nonempty_step_conf_rate": 0.70703125,
|
|
"calib/pce": 0.3204367816091954,
|
|
"calib/std_conf": 0.3195025419416025,
|
|
"calib/step_conf_rate": 0.70703125,
|
|
"calib/step_q_c": 0.19203731343283584,
|
|
"calib/step_q_c_n": 67.0,
|
|
"calib/step_q_gap": -0.09968971847449173,
|
|
"calib/step_q_w": 0.29172703190732757,
|
|
"calib/step_q_w_n": 928.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.04296875,
|
|
"completions/max_length": 2962.0,
|
|
"completions/max_terminated_length": 2962.0,
|
|
"completions/mean_length": 458.6171875,
|
|
"completions/mean_terminated_length": 479.2081298828125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 10.0,
|
|
"epoch": 0.0224,
|
|
"grad_norm": 0.016379138454794884,
|
|
"learning_rate": 4.9722222222222224e-06,
|
|
"loss": 0.2659,
|
|
"num_tokens": 5657224.0,
|
|
"reward": 0.4006524085998535,
|
|
"reward_std": 0.446008563041687,
|
|
"rewards/accuracy_reward_step": 0.0546875,
|
|
"rewards/final_brier_reward_step": 0.30521100759506226,
|
|
"rewards/format_reward_step": 0.44140625,
|
|
"step": 21
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9724233224987984,
|
|
"aux_distill/mean_u": 0.38292050698782076,
|
|
"aux_distill/n_active_tok": 138.75,
|
|
"calib/answer_extract_rate": 0.73046875,
|
|
"calib/auroc": 0.5143292682926829,
|
|
"calib/avg_num_step_conf": 4.3671875,
|
|
"calib/ece": 0.28598071095066196,
|
|
"calib/final_conf_rate": 0.71875,
|
|
"calib/format_rate": 0.59375,
|
|
"calib/frac_conf_gt_0.9": 0.08152173913043478,
|
|
"calib/gap": 0.0046966513104987895,
|
|
"calib/mean_conf": 0.37781385426672937,
|
|
"calib/mu_c": 0.382,
|
|
"calib/mu_w": 0.3773033486895012,
|
|
"calib/nonempty_final_conf_rate": 0.71875,
|
|
"calib/nonempty_reasoning_rate": 0.91796875,
|
|
"calib/nonempty_step_conf_rate": 0.828125,
|
|
"calib/pce": 0.2775494565217391,
|
|
"calib/std_conf": 0.31317783992399195,
|
|
"calib/step_conf_rate": 0.828125,
|
|
"calib/step_q_c": 0.272159375,
|
|
"calib/step_q_c_n": 64.0,
|
|
"calib/step_q_gap": 0.029737965093538432,
|
|
"calib/step_q_w": 0.24242140990646155,
|
|
"calib/step_q_w_n": 1054.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2812.0,
|
|
"completions/max_terminated_length": 2812.0,
|
|
"completions/mean_length": 441.92578125,
|
|
"completions/mean_terminated_length": 448.94049072265625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 30.0,
|
|
"epoch": 0.023466666666666667,
|
|
"grad_norm": 0.015693727880716324,
|
|
"learning_rate": 4.944444444444445e-06,
|
|
"loss": 0.2656,
|
|
"num_tokens": 5872173.0,
|
|
"reward": 0.549543023109436,
|
|
"reward_std": 0.46734896302223206,
|
|
"rewards/accuracy_reward_step": 0.078125,
|
|
"rewards/final_brier_reward_step": 0.4272110164165497,
|
|
"rewards/format_reward_step": 0.59375,
|
|
"step": 22
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9368573725223541,
|
|
"aux_distill/mean_u": 0.32277779323490313,
|
|
"aux_distill/n_active_tok": 152.625,
|
|
"calib/answer_extract_rate": 0.76171875,
|
|
"calib/auroc": 0.38348849945235486,
|
|
"calib/avg_num_step_conf": 4.77734375,
|
|
"calib/ece": 0.3372117021276596,
|
|
"calib/final_conf_rate": 0.734375,
|
|
"calib/format_rate": 0.65234375,
|
|
"calib/frac_conf_gt_0.9": 0.0851063829787234,
|
|
"calib/gap": -0.1021048192771084,
|
|
"calib/mean_conf": 0.36075638297872337,
|
|
"calib/mu_c": 0.2706,
|
|
"calib/mu_w": 0.3727048192771084,
|
|
"calib/nonempty_final_conf_rate": 0.734375,
|
|
"calib/nonempty_reasoning_rate": 0.90625,
|
|
"calib/nonempty_step_conf_rate": 0.8515625,
|
|
"calib/pce": 0.29047340425531915,
|
|
"calib/std_conf": 0.3209408723162062,
|
|
"calib/step_conf_rate": 0.8515625,
|
|
"calib/step_q_c": 0.335420987654321,
|
|
"calib/step_q_c_n": 81.0,
|
|
"calib/step_q_gap": 0.054961840890472125,
|
|
"calib/step_q_w": 0.28045914676384887,
|
|
"calib/step_q_w_n": 1142.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/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": 444.546875,
|
|
"completions/mean_terminated_length": 446.29022216796875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.024533333333333334,
|
|
"grad_norm": 0.0145284878090024,
|
|
"learning_rate": 4.9166666666666665e-06,
|
|
"loss": 0.3628,
|
|
"num_tokens": 6089913.0,
|
|
"reward": 0.6050222516059875,
|
|
"reward_std": 0.4527014493942261,
|
|
"rewards/accuracy_reward_step": 0.08984375,
|
|
"rewards/final_brier_reward_step": 0.4678569436073303,
|
|
"rewards/format_reward_step": 0.65234375,
|
|
"step": 23
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9629884772002697,
|
|
"aux_distill/mean_u": 0.3936236706788197,
|
|
"aux_distill/n_active_tok": 147.75,
|
|
"calib/answer_extract_rate": 0.796875,
|
|
"calib/auroc": 0.5271164021164021,
|
|
"calib/avg_num_step_conf": 4.63671875,
|
|
"calib/ece": 0.2884106318407961,
|
|
"calib/final_conf_rate": 0.78515625,
|
|
"calib/format_rate": 0.71484375,
|
|
"calib/frac_conf_gt_0.9": 0.09950248756218906,
|
|
"calib/gap": 0.026388207142857167,
|
|
"calib/mean_conf": 0.3470116268656716,
|
|
"calib/mu_c": 0.37064285714285716,
|
|
"calib/mu_w": 0.34425465,
|
|
"calib/nonempty_final_conf_rate": 0.78515625,
|
|
"calib/nonempty_reasoning_rate": 0.94921875,
|
|
"calib/nonempty_step_conf_rate": 0.88671875,
|
|
"calib/pce": 0.2654723233830846,
|
|
"calib/std_conf": 0.3144549521779533,
|
|
"calib/step_conf_rate": 0.88671875,
|
|
"calib/step_q_c": 0.25758333333333333,
|
|
"calib/step_q_c_n": 72.0,
|
|
"calib/step_q_gap": -0.012828107025411095,
|
|
"calib/step_q_w": 0.2704114403587444,
|
|
"calib/step_q_w_n": 1115.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2254.0,
|
|
"completions/max_terminated_length": 2254.0,
|
|
"completions/mean_length": 396.0625,
|
|
"completions/mean_terminated_length": 399.18109130859375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 5.0,
|
|
"epoch": 0.0256,
|
|
"grad_norm": 0.013611423783004284,
|
|
"learning_rate": 4.888888888888889e-06,
|
|
"loss": 0.2628,
|
|
"num_tokens": 6295817.0,
|
|
"reward": 0.6710429787635803,
|
|
"reward_std": 0.4255039691925049,
|
|
"rewards/accuracy_reward_step": 0.0859375,
|
|
"rewards/final_brier_reward_step": 0.5413047075271606,
|
|
"rewards/format_reward_step": 0.71484375,
|
|
"step": 24
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9668833911418915,
|
|
"aux_distill/mean_u": 0.354028013155989,
|
|
"aux_distill/n_active_tok": 140.875,
|
|
"calib/answer_extract_rate": 0.84375,
|
|
"calib/auroc": 0.5017616580310882,
|
|
"calib/avg_num_step_conf": 4.41015625,
|
|
"calib/ece": 0.28692567340366976,
|
|
"calib/final_conf_rate": 0.8515625,
|
|
"calib/format_rate": 0.76171875,
|
|
"calib/frac_conf_gt_0.9": 0.08256880733944955,
|
|
"calib/gap": -0.007618006207253836,
|
|
"calib/mean_conf": 0.3615403816422018,
|
|
"calib/mu_c": 0.35479600000000006,
|
|
"calib/mu_w": 0.3624140062072539,
|
|
"calib/nonempty_final_conf_rate": 0.8515625,
|
|
"calib/nonempty_reasoning_rate": 0.9453125,
|
|
"calib/nonempty_step_conf_rate": 0.890625,
|
|
"calib/pce": 0.2668935779816514,
|
|
"calib/std_conf": 0.3098885909725723,
|
|
"calib/step_conf_rate": 0.890625,
|
|
"calib/step_q_c": 0.40253232323232324,
|
|
"calib/step_q_c_n": 99.0,
|
|
"calib/step_q_gap": 0.11097047857212905,
|
|
"calib/step_q_w": 0.2915618446601942,
|
|
"calib/step_q_w_n": 1030.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/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": 347.08203125,
|
|
"completions/mean_terminated_length": 347.08203125,
|
|
"completions/min_length": 38.0,
|
|
"completions/min_terminated_length": 38.0,
|
|
"epoch": 0.02666666666666667,
|
|
"grad_norm": 0.011474363505840302,
|
|
"learning_rate": 4.861111111111111e-06,
|
|
"loss": 0.2234,
|
|
"num_tokens": 6487894.0,
|
|
"reward": 0.7150354981422424,
|
|
"reward_std": 0.39984044432640076,
|
|
"rewards/accuracy_reward_step": 0.09765625,
|
|
"rewards/final_brier_reward_step": 0.5706959962844849,
|
|
"rewards/format_reward_step": 0.76171875,
|
|
"step": 25
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9821081515401602,
|
|
"aux_distill/mean_u": 0.43286420605477743,
|
|
"aux_distill/n_active_tok": 147.125,
|
|
"calib/answer_extract_rate": 0.8359375,
|
|
"calib/auroc": 0.4466071428571429,
|
|
"calib/avg_num_step_conf": 4.6015625,
|
|
"calib/ece": 0.313422429906542,
|
|
"calib/final_conf_rate": 0.8359375,
|
|
"calib/format_rate": 0.77734375,
|
|
"calib/frac_conf_gt_0.9": 0.07009345794392523,
|
|
"calib/gap": -0.04397699999999988,
|
|
"calib/mean_conf": 0.3434,
|
|
"calib/mu_c": 0.30230000000000007,
|
|
"calib/mu_w": 0.34627699999999995,
|
|
"calib/nonempty_final_conf_rate": 0.8359375,
|
|
"calib/nonempty_reasoning_rate": 0.98046875,
|
|
"calib/nonempty_step_conf_rate": 0.9453125,
|
|
"calib/pce": 0.29570093457943925,
|
|
"calib/std_conf": 0.30139188371635606,
|
|
"calib/step_conf_rate": 0.9453125,
|
|
"calib/step_q_c": 0.38596491228070184,
|
|
"calib/step_q_c_n": 57.0,
|
|
"calib/step_q_gap": 0.056750688783824055,
|
|
"calib/step_q_w": 0.3292142234968778,
|
|
"calib/step_q_w_n": 1121.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2283.0,
|
|
"completions/max_terminated_length": 2283.0,
|
|
"completions/mean_length": 334.86328125,
|
|
"completions/mean_terminated_length": 341.53387451171875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 25.0,
|
|
"epoch": 0.027733333333333332,
|
|
"grad_norm": 0.012159218080341816,
|
|
"learning_rate": 4.833333333333333e-06,
|
|
"loss": 0.1981,
|
|
"num_tokens": 6678859.0,
|
|
"reward": 0.7168387174606323,
|
|
"reward_std": 0.38704153895378113,
|
|
"rewards/accuracy_reward_step": 0.0546875,
|
|
"rewards/final_brier_reward_step": 0.6016461849212646,
|
|
"rewards/format_reward_step": 0.77734375,
|
|
"step": 26
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9431140720844269,
|
|
"aux_distill/mean_u": 0.36024294460778056,
|
|
"aux_distill/n_active_tok": 142.625,
|
|
"calib/answer_extract_rate": 0.94140625,
|
|
"calib/auroc": 0.44554772581678415,
|
|
"calib/avg_num_step_conf": 4.4609375,
|
|
"calib/ece": 0.2685700421940928,
|
|
"calib/final_conf_rate": 0.92578125,
|
|
"calib/format_rate": 0.87890625,
|
|
"calib/frac_conf_gt_0.9": 0.04219409282700422,
|
|
"calib/gap": -0.05232706598334408,
|
|
"calib/mean_conf": 0.3059717299578059,
|
|
"calib/mu_c": 0.25673571428571423,
|
|
"calib/mu_w": 0.3090627802690583,
|
|
"calib/nonempty_final_conf_rate": 0.92578125,
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
|
"calib/nonempty_step_conf_rate": 0.95703125,
|
|
"calib/pce": 0.2577350210970464,
|
|
"calib/std_conf": 0.26608733901787746,
|
|
"calib/step_conf_rate": 0.95703125,
|
|
"calib/step_q_c": 0.23749074074074072,
|
|
"calib/step_q_c_n": 54.0,
|
|
"calib/step_q_gap": -0.09427304602396519,
|
|
"calib/step_q_w": 0.3317637867647059,
|
|
"calib/step_q_w_n": 1088.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1761.0,
|
|
"completions/max_terminated_length": 1761.0,
|
|
"completions/mean_length": 287.59375,
|
|
"completions/mean_terminated_length": 288.7215881347656,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 26.0,
|
|
"epoch": 0.0288,
|
|
"grad_norm": 0.011444371193647385,
|
|
"learning_rate": 4.805555555555556e-06,
|
|
"loss": 0.065,
|
|
"num_tokens": 6857699.0,
|
|
"reward": 0.821548581123352,
|
|
"reward_std": 0.2998458445072174,
|
|
"rewards/accuracy_reward_step": 0.0546875,
|
|
"rewards/final_brier_reward_step": 0.7095034122467041,
|
|
"rewards/format_reward_step": 0.87890625,
|
|
"step": 27
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9115942157804966,
|
|
"aux_distill/mean_u": 0.3512940447289571,
|
|
"aux_distill/n_active_tok": 141.0,
|
|
"calib/answer_extract_rate": 0.921875,
|
|
"calib/auroc": 0.527336860670194,
|
|
"calib/avg_num_step_conf": 4.40625,
|
|
"calib/ece": 0.2601383152334178,
|
|
"calib/final_conf_rate": 0.92578125,
|
|
"calib/format_rate": 0.83203125,
|
|
"calib/frac_conf_gt_0.9": 0.029535864978902954,
|
|
"calib/gap": 0.033239774395301624,
|
|
"calib/mean_conf": 0.3483248131237131,
|
|
"calib/mu_c": 0.3777777777777778,
|
|
"calib/mu_w": 0.3445380033824762,
|
|
"calib/nonempty_final_conf_rate": 0.92578125,
|
|
"calib/nonempty_reasoning_rate": 0.98046875,
|
|
"calib/nonempty_step_conf_rate": 0.92578125,
|
|
"calib/pce": 0.24726953886210978,
|
|
"calib/std_conf": 0.25299117933780857,
|
|
"calib/step_conf_rate": 0.92578125,
|
|
"calib/step_q_c": 0.32133137254901956,
|
|
"calib/step_q_c_n": 102.0,
|
|
"calib/step_q_gap": -0.01909075220731571,
|
|
"calib/step_q_w": 0.34042212475633526,
|
|
"calib/step_q_w_n": 1026.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 3038.0,
|
|
"completions/max_terminated_length": 3038.0,
|
|
"completions/mean_length": 312.0546875,
|
|
"completions/mean_terminated_length": 313.2784423828125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 25.0,
|
|
"epoch": 0.029866666666666666,
|
|
"grad_norm": 0.010167010128498077,
|
|
"learning_rate": 4.777777777777778e-06,
|
|
"loss": 0.1532,
|
|
"num_tokens": 7044529.0,
|
|
"reward": 0.7973880171775818,
|
|
"reward_std": 0.3464619815349579,
|
|
"rewards/accuracy_reward_step": 0.10546875,
|
|
"rewards/final_brier_reward_step": 0.6572760343551636,
|
|
"rewards/format_reward_step": 0.83203125,
|
|
"step": 28
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9127364363521338,
|
|
"aux_distill/mean_u": 0.3823967146983576,
|
|
"aux_distill/n_active_tok": 157.375,
|
|
"calib/answer_extract_rate": 0.9140625,
|
|
"calib/auroc": 0.5223503965392934,
|
|
"calib/avg_num_step_conf": 4.91796875,
|
|
"calib/ece": 0.29551680672268904,
|
|
"calib/final_conf_rate": 0.9296875,
|
|
"calib/format_rate": 0.87890625,
|
|
"calib/frac_conf_gt_0.9": 0.037815126050420166,
|
|
"calib/gap": 0.03450781062244651,
|
|
"calib/mean_conf": 0.3666680672268907,
|
|
"calib/mu_c": 0.39842105263157895,
|
|
"calib/mu_w": 0.36391324200913244,
|
|
"calib/nonempty_final_conf_rate": 0.9296875,
|
|
"calib/nonempty_reasoning_rate": 0.984375,
|
|
"calib/nonempty_step_conf_rate": 0.96484375,
|
|
"calib/pce": 0.2911764705882353,
|
|
"calib/std_conf": 0.24589287461620496,
|
|
"calib/step_conf_rate": 0.96484375,
|
|
"calib/step_q_c": 0.38605633802816897,
|
|
"calib/step_q_c_n": 71.0,
|
|
"calib/step_q_gap": 0.01928975553658646,
|
|
"calib/step_q_w": 0.3667665824915825,
|
|
"calib/step_q_w_n": 1188.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 1309.0,
|
|
"completions/max_terminated_length": 1309.0,
|
|
"completions/mean_length": 291.05859375,
|
|
"completions/mean_terminated_length": 291.05859375,
|
|
"completions/min_length": 31.0,
|
|
"completions/min_terminated_length": 31.0,
|
|
"epoch": 0.030933333333333334,
|
|
"grad_norm": 0.009713099338114262,
|
|
"learning_rate": 4.75e-06,
|
|
"loss": 0.1081,
|
|
"num_tokens": 7226168.0,
|
|
"reward": 0.8235447406768799,
|
|
"reward_std": 0.3040723204612732,
|
|
"rewards/accuracy_reward_step": 0.07421875,
|
|
"rewards/final_brier_reward_step": 0.6939644813537598,
|
|
"rewards/format_reward_step": 0.87890625,
|
|
"step": 29
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9098640028387308,
|
|
"aux_distill/mean_u": 0.32866491871894127,
|
|
"aux_distill/n_active_tok": 177.625,
|
|
"calib/answer_extract_rate": 0.94140625,
|
|
"calib/auroc": 0.39525993883792054,
|
|
"calib/avg_num_step_conf": 5.68359375,
|
|
"calib/ece": 0.30045165289256204,
|
|
"calib/final_conf_rate": 0.9453125,
|
|
"calib/format_rate": 0.90234375,
|
|
"calib/frac_conf_gt_0.9": 0.049586776859504134,
|
|
"calib/gap": -0.10314159021406732,
|
|
"calib/mean_conf": 0.37582933884297526,
|
|
"calib/mu_c": 0.28291666666666665,
|
|
"calib/mu_w": 0.386058256880734,
|
|
"calib/nonempty_final_conf_rate": 0.9453125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.28855371900826443,
|
|
"calib/std_conf": 0.27210583928978066,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.3050420168067227,
|
|
"calib/step_q_c_n": 119.0,
|
|
"calib/step_q_gap": -0.08516090235495405,
|
|
"calib/step_q_w": 0.39020291916167676,
|
|
"calib/step_q_w_n": 1336.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1679.0,
|
|
"completions/max_terminated_length": 1679.0,
|
|
"completions/mean_length": 312.91796875,
|
|
"completions/mean_terminated_length": 314.1451110839844,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 53.0,
|
|
"epoch": 0.032,
|
|
"grad_norm": 0.011567425914108753,
|
|
"learning_rate": 4.722222222222222e-06,
|
|
"loss": 0.2423,
|
|
"num_tokens": 7413259.0,
|
|
"reward": 0.837209939956665,
|
|
"reward_std": 0.2785930931568146,
|
|
"rewards/accuracy_reward_step": 0.09765625,
|
|
"rewards/final_brier_reward_step": 0.6744199395179749,
|
|
"rewards/format_reward_step": 0.90234375,
|
|
"step": 30
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9276929814368486,
|
|
"aux_distill/mean_u": 0.3468866376900609,
|
|
"aux_distill/n_active_tok": 162.25,
|
|
"calib/answer_extract_rate": 0.9609375,
|
|
"calib/auroc": 0.5231801971890078,
|
|
"calib/avg_num_step_conf": 5.0703125,
|
|
"calib/ece": 0.28122983870967744,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.03225806451612903,
|
|
"calib/gap": 0.005270610446821866,
|
|
"calib/mean_conf": 0.3623185483870968,
|
|
"calib/mu_c": 0.36714285714285716,
|
|
"calib/mu_w": 0.3618722466960353,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.2794354838709677,
|
|
"calib/std_conf": 0.23781169558956333,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.38036585365853653,
|
|
"calib/step_q_c_n": 82.0,
|
|
"calib/step_q_gap": -0.0040372713414634775,
|
|
"calib/step_q_w": 0.384403125,
|
|
"calib/step_q_w_n": 1216.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 1924.0,
|
|
"completions/max_terminated_length": 1924.0,
|
|
"completions/mean_length": 276.140625,
|
|
"completions/mean_terminated_length": 276.140625,
|
|
"completions/min_length": 67.0,
|
|
"completions/min_terminated_length": 67.0,
|
|
"epoch": 0.03306666666666667,
|
|
"grad_norm": 0.010378521867096424,
|
|
"learning_rate": 4.694444444444445e-06,
|
|
"loss": 0.1865,
|
|
"num_tokens": 7589863.0,
|
|
"reward": 0.8974331617355347,
|
|
"reward_std": 0.2091233730316162,
|
|
"rewards/accuracy_reward_step": 0.08203125,
|
|
"rewards/final_brier_reward_step": 0.7558038234710693,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 31
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.961167685687542,
|
|
"aux_distill/mean_u": 0.37351661006632264,
|
|
"aux_distill/n_active_tok": 174.125,
|
|
"calib/answer_extract_rate": 0.96484375,
|
|
"calib/auroc": 0.466897233201581,
|
|
"calib/avg_num_step_conf": 5.44921875,
|
|
"calib/ece": 0.27502757201646094,
|
|
"calib/final_conf_rate": 0.94921875,
|
|
"calib/format_rate": 0.9375,
|
|
"calib/frac_conf_gt_0.9": 0.01646090534979424,
|
|
"calib/gap": -0.012935968379446605,
|
|
"calib/mean_conf": 0.34920288065843624,
|
|
"calib/mu_c": 0.3374913043478261,
|
|
"calib/mu_w": 0.3504272727272727,
|
|
"calib/nonempty_final_conf_rate": 0.94921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.26479012345679015,
|
|
"calib/std_conf": 0.22711269558032068,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.405546218487395,
|
|
"calib/step_q_c_n": 119.0,
|
|
"calib/step_q_gap": 0.03627239403598437,
|
|
"calib/step_q_w": 0.36927382445141066,
|
|
"calib/step_q_w_n": 1276.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1974.0,
|
|
"completions/max_terminated_length": 1974.0,
|
|
"completions/mean_length": 293.21484375,
|
|
"completions/mean_terminated_length": 294.3647155761719,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 76.0,
|
|
"epoch": 0.034133333333333335,
|
|
"grad_norm": 0.010933423414826393,
|
|
"learning_rate": 4.666666666666667e-06,
|
|
"loss": 0.1524,
|
|
"num_tokens": 7771630.0,
|
|
"reward": 0.885528028011322,
|
|
"reward_std": 0.22666269540786743,
|
|
"rewards/accuracy_reward_step": 0.08984375,
|
|
"rewards/final_brier_reward_step": 0.743712306022644,
|
|
"rewards/format_reward_step": 0.9375,
|
|
"step": 32
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.948054114356637,
|
|
"aux_distill/mean_u": 0.36492014468038125,
|
|
"aux_distill/n_active_tok": 166.625,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.565557939914163,
|
|
"calib/avg_num_step_conf": 5.33203125,
|
|
"calib/ece": 0.27930830039525695,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.953125,
|
|
"calib/frac_conf_gt_0.9": 0.023715415019762844,
|
|
"calib/gap": 0.04072746781115888,
|
|
"calib/mean_conf": 0.33899209486166015,
|
|
"calib/mu_c": 0.3765,
|
|
"calib/mu_w": 0.3357725321888411,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.9765625,
|
|
"calib/pce": 0.2696245059288538,
|
|
"calib/std_conf": 0.22427851812523084,
|
|
"calib/step_conf_rate": 0.9765625,
|
|
"calib/step_q_c": 0.4223333333333334,
|
|
"calib/step_q_c_n": 90.0,
|
|
"calib/step_q_gap": 0.046461960784313805,
|
|
"calib/step_q_w": 0.3758713725490196,
|
|
"calib/step_q_w_n": 1275.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1012.0,
|
|
"completions/max_terminated_length": 1012.0,
|
|
"completions/mean_length": 261.19140625,
|
|
"completions/mean_terminated_length": 262.2156982421875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 51.0,
|
|
"epoch": 0.0352,
|
|
"grad_norm": 0.010418594814836979,
|
|
"learning_rate": 4.638888888888889e-06,
|
|
"loss": 0.0628,
|
|
"num_tokens": 7945367.0,
|
|
"reward": 0.9025432467460632,
|
|
"reward_std": 0.20960542559623718,
|
|
"rewards/accuracy_reward_step": 0.078125,
|
|
"rewards/final_brier_reward_step": 0.7738364934921265,
|
|
"rewards/format_reward_step": 0.953125,
|
|
"step": 33
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9073459450155497,
|
|
"aux_distill/mean_u": 0.36159769528386776,
|
|
"aux_distill/n_active_tok": 198.25,
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/auroc": 0.5216346153846154,
|
|
"calib/avg_num_step_conf": 6.1953125,
|
|
"calib/ece": 0.217292,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.012,
|
|
"calib/gap": -0.005361950549450545,
|
|
"calib/mean_conf": 0.30761199999999994,
|
|
"calib/mu_c": 0.30280769230769233,
|
|
"calib/mu_w": 0.3081696428571429,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.210452,
|
|
"calib/std_conf": 0.2124196164576144,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.36953672316384184,
|
|
"calib/step_q_c_n": 177.0,
|
|
"calib/step_q_gap": 0.0358509176280008,
|
|
"calib/step_q_w": 0.33368580553584104,
|
|
"calib/step_q_w_n": 1409.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 1988.0,
|
|
"completions/max_terminated_length": 1988.0,
|
|
"completions/mean_length": 277.12109375,
|
|
"completions/mean_terminated_length": 277.12109375,
|
|
"completions/min_length": 49.0,
|
|
"completions/min_terminated_length": 49.0,
|
|
"epoch": 0.03626666666666667,
|
|
"grad_norm": 0.00895098876208067,
|
|
"learning_rate": 4.611111111111112e-06,
|
|
"loss": 0.0133,
|
|
"num_tokens": 8121422.0,
|
|
"reward": 0.923614501953125,
|
|
"reward_std": 0.18150296807289124,
|
|
"rewards/accuracy_reward_step": 0.1015625,
|
|
"rewards/final_brier_reward_step": 0.7847288846969604,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 34
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8795184995979071,
|
|
"aux_distill/mean_u": 0.37786130563642584,
|
|
"aux_distill/n_active_tok": 225.625,
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/auroc": 0.5231277533039648,
|
|
"calib/avg_num_step_conf": 7.15234375,
|
|
"calib/ece": 0.1823187250996016,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.00398406374501992,
|
|
"calib/gap": 0.01311233480176216,
|
|
"calib/mean_conf": 0.27564143426294824,
|
|
"calib/mu_c": 0.28750000000000003,
|
|
"calib/mu_w": 0.2743876651982379,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.18117131474103584,
|
|
"calib/std_conf": 0.18911776437217667,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.33661194029850744,
|
|
"calib/step_q_c_n": 134.0,
|
|
"calib/step_q_gap": 0.03802879356898475,
|
|
"calib/step_q_w": 0.2985831467295227,
|
|
"calib/step_q_w_n": 1697.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1376.0,
|
|
"completions/max_terminated_length": 1376.0,
|
|
"completions/mean_length": 327.74609375,
|
|
"completions/mean_terminated_length": 329.0314025878906,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 49.0,
|
|
"epoch": 0.037333333333333336,
|
|
"grad_norm": 0.007883084006607533,
|
|
"learning_rate": 4.583333333333333e-06,
|
|
"loss": 0.1154,
|
|
"num_tokens": 8314581.0,
|
|
"reward": 0.9347312450408936,
|
|
"reward_std": 0.163455069065094,
|
|
"rewards/accuracy_reward_step": 0.09375,
|
|
"rewards/final_brier_reward_step": 0.8147751092910767,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 35
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8844308815896511,
|
|
"aux_distill/mean_u": 0.3635496135825137,
|
|
"aux_distill/n_active_tok": 237.75,
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/auroc": 0.5120708748615725,
|
|
"calib/avg_num_step_conf": 7.7109375,
|
|
"calib/ece": 0.1540284584980237,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.011857707509881422,
|
|
"calib/gap": -0.0014955038759689732,
|
|
"calib/mean_conf": 0.28589249011857704,
|
|
"calib/mu_c": 0.2846511627906977,
|
|
"calib/mu_w": 0.28614666666666666,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.13498023715415022,
|
|
"calib/std_conf": 0.20678860428127496,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.33951417004048584,
|
|
"calib/step_q_c_n": 247.0,
|
|
"calib/step_q_gap": 0.03424879656046276,
|
|
"calib/step_q_w": 0.3052653734800231,
|
|
"calib/step_q_w_n": 1727.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1870.0,
|
|
"completions/max_terminated_length": 1870.0,
|
|
"completions/mean_length": 294.81640625,
|
|
"completions/mean_terminated_length": 297.1377868652344,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 49.0,
|
|
"epoch": 0.0384,
|
|
"grad_norm": 0.007582934107631445,
|
|
"learning_rate": 4.555555555555556e-06,
|
|
"loss": 0.0668,
|
|
"num_tokens": 8492766.0,
|
|
"reward": 0.9707470536231995,
|
|
"reward_std": 0.15500092506408691,
|
|
"rewards/accuracy_reward_step": 0.16796875,
|
|
"rewards/final_brier_reward_step": 0.7891504168510437,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 36
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8487446773797274,
|
|
"aux_distill/mean_u": 0.3279911799047748,
|
|
"aux_distill/n_active_tok": 246.25,
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/auroc": 0.42894611349546974,
|
|
"calib/avg_num_step_conf": 7.6953125,
|
|
"calib/ece": 0.20217011952191236,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.00796812749003984,
|
|
"calib/gap": -0.056092846924177375,
|
|
"calib/mean_conf": 0.25040358565737053,
|
|
"calib/mu_c": 0.19833333333333336,
|
|
"calib/mu_w": 0.25442618025751074,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.19043027888446215,
|
|
"calib/std_conf": 0.18634575676102816,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.21467289719626167,
|
|
"calib/step_q_c_n": 107.0,
|
|
"calib/step_q_gap": -0.06815055959386182,
|
|
"calib/step_q_w": 0.2828234567901235,
|
|
"calib/step_q_w_n": 1863.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2084.0,
|
|
"completions/max_terminated_length": 2084.0,
|
|
"completions/mean_length": 340.00390625,
|
|
"completions/mean_terminated_length": 340.00390625,
|
|
"completions/min_length": 58.0,
|
|
"completions/min_terminated_length": 58.0,
|
|
"epoch": 0.039466666666666664,
|
|
"grad_norm": 0.008167412132024765,
|
|
"learning_rate": 4.527777777777778e-06,
|
|
"loss": 0.1161,
|
|
"num_tokens": 8686903.0,
|
|
"reward": 0.9393041133880615,
|
|
"reward_std": 0.1312025487422943,
|
|
"rewards/accuracy_reward_step": 0.0703125,
|
|
"rewards/final_brier_reward_step": 0.835639476776123,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 37
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9289264902472496,
|
|
"aux_distill/mean_u": 0.41301154567567255,
|
|
"aux_distill/n_active_tok": 264.0,
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/auroc": 0.4389071757735352,
|
|
"calib/avg_num_step_conf": 8.25,
|
|
"calib/ece": 0.15761468253968253,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.04175622119815667,
|
|
"calib/mean_conf": 0.2345281746031746,
|
|
"calib/mu_c": 0.1985714285714286,
|
|
"calib/mu_w": 0.24032764976958526,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
|
"calib/pce": 0.12662698412698414,
|
|
"calib/std_conf": 0.16473101583857125,
|
|
"calib/step_conf_rate": 0.98828125,
|
|
"calib/step_q_c": 0.2512442396313364,
|
|
"calib/step_q_c_n": 217.0,
|
|
"calib/step_q_gap": -0.031052330289507935,
|
|
"calib/step_q_w": 0.28229656992084434,
|
|
"calib/step_q_w_n": 1895.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 1760.0,
|
|
"completions/max_terminated_length": 1760.0,
|
|
"completions/mean_length": 329.30859375,
|
|
"completions/mean_terminated_length": 329.30859375,
|
|
"completions/min_length": 10.0,
|
|
"completions/min_terminated_length": 10.0,
|
|
"epoch": 0.04053333333333333,
|
|
"grad_norm": 0.007806702516973019,
|
|
"learning_rate": 4.5e-06,
|
|
"loss": 0.1166,
|
|
"num_tokens": 8878094.0,
|
|
"reward": 0.955902099609375,
|
|
"reward_std": 0.1512955129146576,
|
|
"rewards/accuracy_reward_step": 0.13671875,
|
|
"rewards/final_brier_reward_step": 0.80633544921875,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 38
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8619994595646858,
|
|
"aux_distill/mean_u": 0.35136166090086507,
|
|
"aux_distill/n_active_tok": 290.75,
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/auroc": 0.4599693545297835,
|
|
"calib/avg_num_step_conf": 9.8828125,
|
|
"calib/ece": 0.14684,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.004,
|
|
"calib/gap": -0.027571346485347648,
|
|
"calib/mean_conf": 0.22460000000000002,
|
|
"calib/mu_c": 0.19956521739130434,
|
|
"calib/mu_w": 0.227136563876652,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
|
"calib/pce": 0.13971999999999998,
|
|
"calib/std_conf": 0.15636380655381857,
|
|
"calib/step_conf_rate": 0.9921875,
|
|
"calib/step_q_c": 0.2652631578947368,
|
|
"calib/step_q_c_n": 152.0,
|
|
"calib/step_q_gap": 0.010465849231995061,
|
|
"calib/step_q_w": 0.25479730866274175,
|
|
"calib/step_q_w_n": 2378.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1418.0,
|
|
"completions/max_terminated_length": 1418.0,
|
|
"completions/mean_length": 360.7890625,
|
|
"completions/mean_terminated_length": 363.6299133300781,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 49.0,
|
|
"epoch": 0.0416,
|
|
"grad_norm": 0.007670476567000151,
|
|
"learning_rate": 4.472222222222223e-06,
|
|
"loss": 0.0735,
|
|
"num_tokens": 9076544.0,
|
|
"reward": 0.9427025318145752,
|
|
"reward_std": 0.15323060750961304,
|
|
"rewards/accuracy_reward_step": 0.08984375,
|
|
"rewards/final_brier_reward_step": 0.8346238136291504,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 39
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8237579446285963,
|
|
"aux_distill/mean_u": 0.3452751043865746,
|
|
"aux_distill/n_active_tok": 307.0,
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/auroc": 0.5413419179939003,
|
|
"calib/avg_num_step_conf": 9.94140625,
|
|
"calib/ece": 0.13450592885375492,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.003952569169960474,
|
|
"calib/gap": 0.007600813283632657,
|
|
"calib/mean_conf": 0.2224110671936759,
|
|
"calib/mu_c": 0.22923076923076924,
|
|
"calib/mu_w": 0.22162995594713658,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.12707509881422924,
|
|
"calib/std_conf": 0.1768323523174813,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.3306122448979592,
|
|
"calib/step_q_c_n": 245.0,
|
|
"calib/step_q_gap": 0.08606224489795919,
|
|
"calib/step_q_w": 0.24455000000000002,
|
|
"calib/step_q_w_n": 2300.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2042.0,
|
|
"completions/max_terminated_length": 2042.0,
|
|
"completions/mean_length": 360.4921875,
|
|
"completions/mean_terminated_length": 361.9059143066406,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 50.0,
|
|
"epoch": 0.042666666666666665,
|
|
"grad_norm": 0.008977770805358887,
|
|
"learning_rate": 4.444444444444444e-06,
|
|
"loss": 0.0507,
|
|
"num_tokens": 9275590.0,
|
|
"reward": 0.9639776945114136,
|
|
"reward_std": 0.125410258769989,
|
|
"rewards/accuracy_reward_step": 0.1015625,
|
|
"rewards/final_brier_reward_step": 0.8459242582321167,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 40
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8463635481894016,
|
|
"aux_distill/mean_u": 0.3650670920135531,
|
|
"aux_distill/n_active_tok": 354.25,
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/auroc": 0.49706208425720616,
|
|
"calib/avg_num_step_conf": 11.0703125,
|
|
"calib/ece": 0.1395983935742972,
|
|
"calib/final_conf_rate": 0.97265625,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.004016064257028112,
|
|
"calib/gap": -0.008308203991130775,
|
|
"calib/mean_conf": 0.2163855421686747,
|
|
"calib/mu_c": 0.20954545454545456,
|
|
"calib/mu_w": 0.21785365853658534,
|
|
"calib/nonempty_final_conf_rate": 0.97265625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.08963855421686748,
|
|
"calib/std_conf": 0.16604720609034307,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2590909090909091,
|
|
"calib/step_q_c_n": 407.0,
|
|
"calib/step_q_gap": 0.012933925160130366,
|
|
"calib/step_q_w": 0.24615698393077873,
|
|
"calib/step_q_w_n": 2427.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 3049.0,
|
|
"completions/max_terminated_length": 3049.0,
|
|
"completions/mean_length": 413.68359375,
|
|
"completions/mean_terminated_length": 413.68359375,
|
|
"completions/min_length": 82.0,
|
|
"completions/min_terminated_length": 82.0,
|
|
"epoch": 0.04373333333333333,
|
|
"grad_norm": 0.006722410209476948,
|
|
"learning_rate": 4.416666666666667e-06,
|
|
"loss": 0.2131,
|
|
"num_tokens": 9488741.0,
|
|
"reward": 0.9724918007850647,
|
|
"reward_std": 0.14764532446861267,
|
|
"rewards/accuracy_reward_step": 0.171875,
|
|
"rewards/final_brier_reward_step": 0.8004522919654846,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 41
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8599336389452219,
|
|
"aux_distill/mean_u": 0.3952088352244836,
|
|
"aux_distill/n_active_tok": 315.375,
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/auroc": 0.5308812684365782,
|
|
"calib/avg_num_step_conf": 10.62109375,
|
|
"calib/ece": 0.1265,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.004,
|
|
"calib/gap": -0.00944690265486725,
|
|
"calib/mean_conf": 0.19854000000000002,
|
|
"calib/mu_c": 0.19000000000000003,
|
|
"calib/mu_w": 0.19944690265486728,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.11452000000000001,
|
|
"calib/std_conf": 0.17253048542214214,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24458100558659218,
|
|
"calib/step_q_c_n": 179.0,
|
|
"calib/step_q_gap": 0.01724714731887564,
|
|
"calib/step_q_w": 0.22733385826771654,
|
|
"calib/step_q_w_n": 2540.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2192.0,
|
|
"completions/max_terminated_length": 2192.0,
|
|
"completions/mean_length": 363.9921875,
|
|
"completions/mean_terminated_length": 368.3083190917969,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 52.0,
|
|
"epoch": 0.0448,
|
|
"grad_norm": 0.006656656973063946,
|
|
"learning_rate": 4.388888888888889e-06,
|
|
"loss": 0.1028,
|
|
"num_tokens": 9686291.0,
|
|
"reward": 0.9527983665466309,
|
|
"reward_std": 0.14238354563713074,
|
|
"rewards/accuracy_reward_step": 0.09375,
|
|
"rewards/final_brier_reward_step": 0.8430967926979065,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 42
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8323895744979382,
|
|
"aux_distill/mean_u": 0.3494921272171143,
|
|
"aux_distill/n_active_tok": 355.0,
|
|
"calib/answer_extract_rate": 0.95703125,
|
|
"calib/auroc": 0.49543650793650784,
|
|
"calib/avg_num_step_conf": 11.4765625,
|
|
"calib/ece": 0.14447073170731708,
|
|
"calib/final_conf_rate": 0.9609375,
|
|
"calib/format_rate": 0.953125,
|
|
"calib/frac_conf_gt_0.9": 0.0040650406504065045,
|
|
"calib/gap": -0.008338095238095228,
|
|
"calib/mean_conf": 0.20398455284552844,
|
|
"calib/mu_c": 0.19686666666666666,
|
|
"calib/mu_w": 0.2052047619047619,
|
|
"calib/nonempty_final_conf_rate": 0.9609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.10105691056910572,
|
|
"calib/std_conf": 0.17250708930054556,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2714287822878229,
|
|
"calib/step_q_c_n": 271.0,
|
|
"calib/step_q_gap": 0.049459800640653806,
|
|
"calib/step_q_w": 0.2219689816471691,
|
|
"calib/step_q_w_n": 2667.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2715.0,
|
|
"completions/max_terminated_length": 2715.0,
|
|
"completions/mean_length": 429.9296875,
|
|
"completions/mean_terminated_length": 431.61572265625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 72.0,
|
|
"epoch": 0.04586666666666667,
|
|
"grad_norm": 0.006840248592197895,
|
|
"learning_rate": 4.361111111111112e-06,
|
|
"loss": 0.2256,
|
|
"num_tokens": 9901577.0,
|
|
"reward": 0.9465239644050598,
|
|
"reward_std": 0.16663599014282227,
|
|
"rewards/accuracy_reward_step": 0.140625,
|
|
"rewards/final_brier_reward_step": 0.7992979288101196,
|
|
"rewards/format_reward_step": 0.953125,
|
|
"step": 43
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8238438032567501,
|
|
"aux_distill/mean_u": 0.3623975311474338,
|
|
"aux_distill/n_active_tok": 386.5,
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/auroc": 0.5369152046783625,
|
|
"calib/avg_num_step_conf": 13.3515625,
|
|
"calib/ece": 0.12897154471544714,
|
|
"calib/final_conf_rate": 0.9609375,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.013685672514619857,
|
|
"calib/mean_conf": 0.1984268292682927,
|
|
"calib/mu_c": 0.21111111111111108,
|
|
"calib/mu_w": 0.19742543859649123,
|
|
"calib/nonempty_final_conf_rate": 0.9609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.12711382113821137,
|
|
"calib/std_conf": 0.16426555170002113,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22874999999999998,
|
|
"calib/step_q_c_n": 152.0,
|
|
"calib/step_q_gap": 0.005412584200857301,
|
|
"calib/step_q_w": 0.22333741579914268,
|
|
"calib/step_q_w_n": 3266.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2949.0,
|
|
"completions/max_terminated_length": 2949.0,
|
|
"completions/mean_length": 419.015625,
|
|
"completions/mean_terminated_length": 425.66668701171875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 39.0,
|
|
"epoch": 0.046933333333333334,
|
|
"grad_norm": 0.006506130564957857,
|
|
"learning_rate": 4.333333333333334e-06,
|
|
"loss": 0.0528,
|
|
"num_tokens": 10115165.0,
|
|
"reward": 0.9438990950584412,
|
|
"reward_std": 0.15580600500106812,
|
|
"rewards/accuracy_reward_step": 0.0703125,
|
|
"rewards/final_brier_reward_step": 0.8565481901168823,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 44
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8242166340351105,
|
|
"aux_distill/mean_u": 0.3212903089373872,
|
|
"aux_distill/n_active_tok": 352.125,
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/auroc": 0.5975462590506838,
|
|
"calib/avg_num_step_conf": 12.609375,
|
|
"calib/ece": 0.10689516129032259,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.004032258064516129,
|
|
"calib/gap": 0.04069589702333065,
|
|
"calib/mean_conf": 0.18745967741935482,
|
|
"calib/mu_c": 0.22454545454545452,
|
|
"calib/mu_w": 0.18384955752212387,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.10282258064516128,
|
|
"calib/std_conf": 0.15124083012424303,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24678414096916299,
|
|
"calib/step_q_c_n": 227.0,
|
|
"calib/step_q_gap": 0.030857816410682443,
|
|
"calib/step_q_w": 0.21592632455848054,
|
|
"calib/step_q_w_n": 3001.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3004.0,
|
|
"completions/max_terminated_length": 3004.0,
|
|
"completions/mean_length": 400.49609375,
|
|
"completions/mean_terminated_length": 406.85321044921875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 45.0,
|
|
"epoch": 0.048,
|
|
"grad_norm": 0.0066505554132163525,
|
|
"learning_rate": 4.305555555555556e-06,
|
|
"loss": 0.0685,
|
|
"num_tokens": 10322740.0,
|
|
"reward": 0.9599459171295166,
|
|
"reward_std": 0.11950437724590302,
|
|
"rewards/accuracy_reward_step": 0.0859375,
|
|
"rewards/final_brier_reward_step": 0.8652043342590332,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 45
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8673719074577093,
|
|
"aux_distill/mean_u": 0.3893094594404342,
|
|
"aux_distill/n_active_tok": 310.625,
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/auroc": 0.5761150234741783,
|
|
"calib/avg_num_step_conf": 9.70703125,
|
|
"calib/ece": 0.09094861660079052,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.03984272300469485,
|
|
"calib/mean_conf": 0.17695652173913046,
|
|
"calib/mu_c": 0.2105,
|
|
"calib/mu_w": 0.17065727699530514,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.05490118577075099,
|
|
"calib/std_conf": 0.16107784828602306,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.32854938271604933,
|
|
"calib/step_q_c_n": 324.0,
|
|
"calib/step_q_gap": 0.08875299215612334,
|
|
"calib/step_q_w": 0.239796390559926,
|
|
"calib/step_q_w_n": 2161.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2444.0,
|
|
"completions/max_terminated_length": 2444.0,
|
|
"completions/mean_length": 341.765625,
|
|
"completions/mean_terminated_length": 343.10589599609375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 31.0,
|
|
"epoch": 0.04906666666666667,
|
|
"grad_norm": 0.007414902560412884,
|
|
"learning_rate": 4.277777777777778e-06,
|
|
"loss": 0.0844,
|
|
"num_tokens": 10515000.0,
|
|
"reward": 0.9855220317840576,
|
|
"reward_std": 0.11352653801441193,
|
|
"rewards/accuracy_reward_step": 0.15625,
|
|
"rewards/final_brier_reward_step": 0.8343254327774048,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 46
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8376510702073574,
|
|
"aux_distill/mean_u": 0.3701711814266018,
|
|
"aux_distill/n_active_tok": 377.625,
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/auroc": 0.4256296800544588,
|
|
"calib/avg_num_step_conf": 12.76171875,
|
|
"calib/ece": 0.1684920634920635,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.037726344452008204,
|
|
"calib/mean_conf": 0.1996031746031746,
|
|
"calib/mu_c": 0.16576923076923075,
|
|
"calib/mu_w": 0.20349557522123896,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.13246031746031744,
|
|
"calib/std_conf": 0.16099591236874838,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24721991701244814,
|
|
"calib/step_q_c_n": 241.0,
|
|
"calib/step_q_gap": 0.027494206503525437,
|
|
"calib/step_q_w": 0.2197257105089227,
|
|
"calib/step_q_w_n": 3026.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2882.0,
|
|
"completions/max_terminated_length": 2882.0,
|
|
"completions/mean_length": 398.42578125,
|
|
"completions/mean_terminated_length": 403.15020751953125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 44.0,
|
|
"epoch": 0.050133333333333335,
|
|
"grad_norm": 0.006663271691650152,
|
|
"learning_rate": 4.25e-06,
|
|
"loss": 0.0534,
|
|
"num_tokens": 10722973.0,
|
|
"reward": 0.9649396538734436,
|
|
"reward_std": 0.10768487304449081,
|
|
"rewards/accuracy_reward_step": 0.1015625,
|
|
"rewards/final_brier_reward_step": 0.8478480577468872,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 47
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8315467443317175,
|
|
"aux_distill/mean_u": 0.3491289536408796,
|
|
"aux_distill/n_active_tok": 358.375,
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/auroc": 0.5252850435949028,
|
|
"calib/avg_num_step_conf": 13.65625,
|
|
"calib/ece": 0.12555120967741937,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.013862441314553942,
|
|
"calib/mean_conf": 0.20009395161290322,
|
|
"calib/mu_c": 0.21199999999999997,
|
|
"calib/mu_w": 0.19813755868544602,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.09225806451612906,
|
|
"calib/std_conf": 0.1687304911175566,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.29284172661870506,
|
|
"calib/step_q_c_n": 278.0,
|
|
"calib/step_q_gap": 0.05798778006805247,
|
|
"calib/step_q_w": 0.2348539465506526,
|
|
"calib/step_q_w_n": 3218.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0234375,
|
|
"completions/max_length": 2289.0,
|
|
"completions/max_terminated_length": 2289.0,
|
|
"completions/mean_length": 368.98046875,
|
|
"completions/mean_terminated_length": 377.8360290527344,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 68.0,
|
|
"epoch": 0.0512,
|
|
"grad_norm": 0.006752731278538704,
|
|
"learning_rate": 4.222222222222223e-06,
|
|
"loss": 0.1196,
|
|
"num_tokens": 10921120.0,
|
|
"reward": 0.9645509719848633,
|
|
"reward_std": 0.14472366869449615,
|
|
"rewards/accuracy_reward_step": 0.13671875,
|
|
"rewards/final_brier_reward_step": 0.8236333131790161,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 48
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8337438236922026,
|
|
"aux_distill/mean_u": 0.39130343889937536,
|
|
"aux_distill/n_active_tok": 376.875,
|
|
"calib/answer_extract_rate": 0.96484375,
|
|
"calib/auroc": 0.49167733674775926,
|
|
"calib/avg_num_step_conf": 13.01171875,
|
|
"calib/ece": 0.11639837398373983,
|
|
"calib/final_conf_rate": 0.9609375,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.0040650406504065045,
|
|
"calib/gap": -0.011475032010243275,
|
|
"calib/mean_conf": 0.17872357723577237,
|
|
"calib/mu_c": 0.1687878787878788,
|
|
"calib/mu_w": 0.18026291079812207,
|
|
"calib/nonempty_final_conf_rate": 0.9609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.08048780487804877,
|
|
"calib/std_conf": 0.1658453977298707,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.27235294117647063,
|
|
"calib/step_q_c_n": 289.0,
|
|
"calib/step_q_gap": 0.04456563019685192,
|
|
"calib/step_q_w": 0.2277873109796187,
|
|
"calib/step_q_w_n": 3042.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2959.0,
|
|
"completions/max_terminated_length": 2959.0,
|
|
"completions/mean_length": 387.12890625,
|
|
"completions/mean_terminated_length": 393.2738342285156,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 91.0,
|
|
"epoch": 0.05226666666666667,
|
|
"grad_norm": 0.00637467298656702,
|
|
"learning_rate": 4.194444444444445e-06,
|
|
"loss": 0.0965,
|
|
"num_tokens": 11124761.0,
|
|
"reward": 0.9503391981124878,
|
|
"reward_std": 0.16382259130477905,
|
|
"rewards/accuracy_reward_step": 0.12890625,
|
|
"rewards/final_brier_reward_step": 0.8147409558296204,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 49
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8423834536224604,
|
|
"aux_distill/mean_u": 0.37696418118094266,
|
|
"aux_distill/n_active_tok": 362.125,
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/auroc": 0.543382167213943,
|
|
"calib/avg_num_step_conf": 12.73046875,
|
|
"calib/ece": 0.1179003984063745,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.00398406374501992,
|
|
"calib/gap": 0.022673276079818117,
|
|
"calib/mean_conf": 0.19553386454183264,
|
|
"calib/mu_c": 0.21486486486486486,
|
|
"calib/mu_w": 0.19219158878504675,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.08301195219123504,
|
|
"calib/std_conf": 0.1812651253439056,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.298502538071066,
|
|
"calib/step_q_c_n": 394.0,
|
|
"calib/step_q_gap": 0.0967322064829334,
|
|
"calib/step_q_w": 0.20177033158813262,
|
|
"calib/step_q_w_n": 2865.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1494.0,
|
|
"completions/max_terminated_length": 1494.0,
|
|
"completions/mean_length": 367.125,
|
|
"completions/mean_terminated_length": 374.4382629394531,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 75.0,
|
|
"epoch": 0.05333333333333334,
|
|
"grad_norm": 0.006909578572958708,
|
|
"learning_rate": 4.166666666666667e-06,
|
|
"loss": 0.0143,
|
|
"num_tokens": 11324105.0,
|
|
"reward": 0.9689640402793884,
|
|
"reward_std": 0.1383761614561081,
|
|
"rewards/accuracy_reward_step": 0.14453125,
|
|
"rewards/final_brier_reward_step": 0.8207405209541321,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 50
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.820130554959178,
|
|
"aux_distill/mean_u": 0.3369361976651867,
|
|
"aux_distill/n_active_tok": 380.25,
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/auroc": 0.5625835459958682,
|
|
"calib/avg_num_step_conf": 13.02734375,
|
|
"calib/ece": 0.1058,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.008,
|
|
"calib/gap": 0.06487908615870702,
|
|
"calib/mean_conf": 0.18395999999999998,
|
|
"calib/mu_c": 0.23871794871794877,
|
|
"calib/mu_w": 0.17383886255924175,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06688000000000001,
|
|
"calib/std_conf": 0.1822106429383311,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22159268929503917,
|
|
"calib/step_q_c_n": 383.0,
|
|
"calib/step_q_gap": -0.01511767655861937,
|
|
"calib/step_q_w": 0.23671036585365854,
|
|
"calib/step_q_w_n": 2952.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3036.0,
|
|
"completions/max_terminated_length": 3036.0,
|
|
"completions/mean_length": 367.9765625,
|
|
"completions/mean_terminated_length": 373.8174743652344,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 70.0,
|
|
"epoch": 0.0544,
|
|
"grad_norm": 0.006935219746083021,
|
|
"learning_rate": 4.138888888888889e-06,
|
|
"loss": 0.1055,
|
|
"num_tokens": 11527603.0,
|
|
"reward": 0.9762929677963257,
|
|
"reward_std": 0.13964669406414032,
|
|
"rewards/accuracy_reward_step": 0.15234375,
|
|
"rewards/final_brier_reward_step": 0.8275859355926514,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 51
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8263747384771705,
|
|
"aux_distill/mean_u": 0.3891820417880071,
|
|
"aux_distill/n_active_tok": 373.375,
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/auroc": 0.5281593406593407,
|
|
"calib/avg_num_step_conf": 13.015625,
|
|
"calib/ece": 0.1266,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.004,
|
|
"calib/gap": 0.008986950549450534,
|
|
"calib/mean_conf": 0.19963999999999998,
|
|
"calib/mu_c": 0.2076923076923077,
|
|
"calib/mu_w": 0.19870535714285717,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.11112,
|
|
"calib/std_conf": 0.18129608489981244,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2317586206896552,
|
|
"calib/step_q_c_n": 290.0,
|
|
"calib/step_q_gap": -0.01828118864630801,
|
|
"calib/step_q_w": 0.2500398093359632,
|
|
"calib/step_q_w_n": 3042.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2896.0,
|
|
"completions/max_terminated_length": 2896.0,
|
|
"completions/mean_length": 386.859375,
|
|
"completions/mean_terminated_length": 393.0000305175781,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 105.0,
|
|
"epoch": 0.055466666666666664,
|
|
"grad_norm": 0.006386070977896452,
|
|
"learning_rate": 4.111111111111111e-06,
|
|
"loss": 0.0668,
|
|
"num_tokens": 11734591.0,
|
|
"reward": 0.9621462821960449,
|
|
"reward_std": 0.13067379593849182,
|
|
"rewards/accuracy_reward_step": 0.1015625,
|
|
"rewards/final_brier_reward_step": 0.8461675643920898,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 52
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8546681217849255,
|
|
"aux_distill/mean_u": 0.3559310260113764,
|
|
"aux_distill/n_active_tok": 354.625,
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/auroc": 0.41209183673469385,
|
|
"calib/avg_num_step_conf": 12.4921875,
|
|
"calib/ece": 0.1763453815261044,
|
|
"calib/final_conf_rate": 0.97265625,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.004016064257028112,
|
|
"calib/gap": -0.03902142857142857,
|
|
"calib/mean_conf": 0.1827710843373494,
|
|
"calib/mu_c": 0.15142857142857144,
|
|
"calib/mu_w": 0.19045,
|
|
"calib/nonempty_final_conf_rate": 0.97265625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.08116465863453816,
|
|
"calib/std_conf": 0.17555863734895014,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23109442060085839,
|
|
"calib/step_q_c_n": 466.0,
|
|
"calib/step_q_gap": -0.002759898579229453,
|
|
"calib/step_q_w": 0.23385431918008784,
|
|
"calib/step_q_w_n": 2732.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1689.0,
|
|
"completions/max_terminated_length": 1689.0,
|
|
"completions/mean_length": 376.61328125,
|
|
"completions/mean_terminated_length": 382.5912780761719,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 88.0,
|
|
"epoch": 0.05653333333333333,
|
|
"grad_norm": 0.006896526087075472,
|
|
"learning_rate": 4.083333333333334e-06,
|
|
"loss": 0.0903,
|
|
"num_tokens": 11936828.0,
|
|
"reward": 0.9704056978225708,
|
|
"reward_std": 0.14509528875350952,
|
|
"rewards/accuracy_reward_step": 0.19140625,
|
|
"rewards/final_brier_reward_step": 0.7767488360404968,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 53
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8472011089324951,
|
|
"aux_distill/mean_u": 0.3172731910665095,
|
|
"aux_distill/n_active_tok": 324.75,
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/auroc": 0.4326,
|
|
"calib/avg_num_step_conf": 10.94140625,
|
|
"calib/ece": 0.19068,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.02390000000000006,
|
|
"calib/mean_conf": 0.17292000000000002,
|
|
"calib/mu_c": 0.15379999999999996,
|
|
"calib/mu_w": 0.17770000000000002,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0818,
|
|
"calib/std_conf": 0.17552171831428726,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.26541371158392435,
|
|
"calib/step_q_c_n": 423.0,
|
|
"calib/step_q_gap": -0.00824860969446084,
|
|
"calib/step_q_w": 0.2736623212783852,
|
|
"calib/step_q_w_n": 2378.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1353.0,
|
|
"completions/max_terminated_length": 1353.0,
|
|
"completions/mean_length": 342.5078125,
|
|
"completions/mean_terminated_length": 347.9444580078125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 95.0,
|
|
"epoch": 0.0576,
|
|
"grad_norm": 0.007074246183037758,
|
|
"learning_rate": 4.055555555555556e-06,
|
|
"loss": 0.0724,
|
|
"num_tokens": 12130742.0,
|
|
"reward": 0.9789115190505981,
|
|
"reward_std": 0.12483196705579758,
|
|
"rewards/accuracy_reward_step": 0.19921875,
|
|
"rewards/final_brier_reward_step": 0.7820417881011963,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 54
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8344004545360804,
|
|
"aux_distill/mean_u": 0.36634527799886907,
|
|
"aux_distill/n_active_tok": 361.875,
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/auroc": 0.4828042328042328,
|
|
"calib/avg_num_step_conf": 13.19921875,
|
|
"calib/ece": 0.13919678714859438,
|
|
"calib/final_conf_rate": 0.97265625,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.007305037957211852,
|
|
"calib/mean_conf": 0.17012048192771084,
|
|
"calib/mu_c": 0.16404761904761905,
|
|
"calib/mu_w": 0.1713526570048309,
|
|
"calib/nonempty_final_conf_rate": 0.97265625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.07032128514056223,
|
|
"calib/std_conf": 0.1639313254384552,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23919137466307278,
|
|
"calib/step_q_c_n": 371.0,
|
|
"calib/step_q_gap": 0.00885227891839191,
|
|
"calib/step_q_w": 0.23033909574468087,
|
|
"calib/step_q_w_n": 3008.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1977.0,
|
|
"completions/max_terminated_length": 1977.0,
|
|
"completions/mean_length": 377.52734375,
|
|
"completions/mean_terminated_length": 385.0478210449219,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 75.0,
|
|
"epoch": 0.058666666666666666,
|
|
"grad_norm": 0.007246608845889568,
|
|
"learning_rate": 4.027777777777779e-06,
|
|
"loss": 0.0711,
|
|
"num_tokens": 12335213.0,
|
|
"reward": 0.9685904383659363,
|
|
"reward_std": 0.13275596499443054,
|
|
"rewards/accuracy_reward_step": 0.1640625,
|
|
"rewards/final_brier_reward_step": 0.8043683767318726,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 55
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8279554862529039,
|
|
"aux_distill/mean_u": 0.37409573772743887,
|
|
"aux_distill/n_active_tok": 383.625,
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/auroc": 0.4842229635807618,
|
|
"calib/avg_num_step_conf": 12.7578125,
|
|
"calib/ece": 0.13058964143426297,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.023431470670002785,
|
|
"calib/mean_conf": 0.15004780876494023,
|
|
"calib/mu_c": 0.12969696969696973,
|
|
"calib/mu_w": 0.1531284403669725,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.07458167330677291,
|
|
"calib/std_conf": 0.15279055562966187,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.3069805194805195,
|
|
"calib/step_q_c_n": 308.0,
|
|
"calib/step_q_gap": 0.063362196289174,
|
|
"calib/step_q_w": 0.24361832319134552,
|
|
"calib/step_q_w_n": 2958.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2580.0,
|
|
"completions/max_terminated_length": 2580.0,
|
|
"completions/mean_length": 382.5546875,
|
|
"completions/mean_terminated_length": 387.0909118652344,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 87.0,
|
|
"epoch": 0.05973333333333333,
|
|
"grad_norm": 0.00718336459249258,
|
|
"learning_rate": 4.000000000000001e-06,
|
|
"loss": 0.1454,
|
|
"num_tokens": 12539987.0,
|
|
"reward": 0.9713689088821411,
|
|
"reward_std": 0.11349431425333023,
|
|
"rewards/accuracy_reward_step": 0.12890625,
|
|
"rewards/final_brier_reward_step": 0.8372691869735718,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 56
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8112405212596059,
|
|
"aux_distill/mean_u": 0.3368667943766801,
|
|
"aux_distill/n_active_tok": 393.25,
|
|
"calib/answer_extract_rate": 0.94140625,
|
|
"calib/auroc": 0.4852999340804219,
|
|
"calib/avg_num_step_conf": 15.98828125,
|
|
"calib/ece": 0.11681818181818181,
|
|
"calib/final_conf_rate": 0.9453125,
|
|
"calib/format_rate": 0.94140625,
|
|
"calib/frac_conf_gt_0.9": 0.004132231404958678,
|
|
"calib/gap": -0.029485827290705355,
|
|
"calib/mean_conf": 0.14524793388429755,
|
|
"calib/mu_c": 0.12027027027027028,
|
|
"calib/mu_w": 0.14975609756097563,
|
|
"calib/nonempty_final_conf_rate": 0.9453125,
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.054586776859504124,
|
|
"calib/std_conf": 0.15714702180656595,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.23159292035398227,
|
|
"calib/step_q_c_n": 339.0,
|
|
"calib/step_q_gap": 0.0018957972852555738,
|
|
"calib/step_q_w": 0.2296971230687267,
|
|
"calib/step_q_w_n": 3754.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.04296875,
|
|
"completions/max_length": 2962.0,
|
|
"completions/max_terminated_length": 2962.0,
|
|
"completions/mean_length": 399.31640625,
|
|
"completions/mean_terminated_length": 417.244873046875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 104.0,
|
|
"epoch": 0.0608,
|
|
"grad_norm": 0.006855270825326443,
|
|
"learning_rate": 3.972222222222223e-06,
|
|
"loss": 0.0401,
|
|
"num_tokens": 12749004.0,
|
|
"reward": 0.9371576309204102,
|
|
"reward_std": 0.19312262535095215,
|
|
"rewards/accuracy_reward_step": 0.14453125,
|
|
"rewards/final_brier_reward_step": 0.7883777618408203,
|
|
"rewards/format_reward_step": 0.94140625,
|
|
"step": 57
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8085120283067226,
|
|
"aux_distill/mean_u": 0.3060995207226738,
|
|
"aux_distill/n_active_tok": 358.125,
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/auroc": 0.5294642857142857,
|
|
"calib/avg_num_step_conf": 12.58203125,
|
|
"calib/ece": 0.11443999999999999,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.015035714285714291,
|
|
"calib/mean_conf": 0.15012,
|
|
"calib/mu_c": 0.16275,
|
|
"calib/mu_w": 0.14771428571428571,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.05228,
|
|
"calib/std_conf": 0.15823143050607866,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2920795107033639,
|
|
"calib/step_q_c_n": 327.0,
|
|
"calib/step_q_gap": 0.0573316530668746,
|
|
"calib/step_q_w": 0.2347478576364893,
|
|
"calib/step_q_w_n": 2894.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3068.0,
|
|
"completions/max_terminated_length": 3068.0,
|
|
"completions/mean_length": 385.15234375,
|
|
"completions/mean_terminated_length": 391.2658996582031,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 72.0,
|
|
"epoch": 0.06186666666666667,
|
|
"grad_norm": 0.007380124181509018,
|
|
"learning_rate": 3.944444444444445e-06,
|
|
"loss": 0.1052,
|
|
"num_tokens": 12953923.0,
|
|
"reward": 0.9748600125312805,
|
|
"reward_std": 0.12817591428756714,
|
|
"rewards/accuracy_reward_step": 0.15625,
|
|
"rewards/final_brier_reward_step": 0.8208136558532715,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 58
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8197278287261724,
|
|
"aux_distill/mean_u": 0.29762419152445097,
|
|
"aux_distill/n_active_tok": 332.25,
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/auroc": 0.3801643192488263,
|
|
"calib/avg_num_step_conf": 11.125,
|
|
"calib/ece": 0.15956521739130433,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.061835680751173724,
|
|
"calib/mean_conf": 0.13505928853754942,
|
|
"calib/mu_c": 0.083,
|
|
"calib/mu_w": 0.14483568075117373,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06826086956521739,
|
|
"calib/std_conf": 0.15493622442482427,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23215805471124618,
|
|
"calib/step_q_c_n": 329.0,
|
|
"calib/step_q_gap": -0.016218285106141678,
|
|
"calib/step_q_w": 0.24837633981738785,
|
|
"calib/step_q_w_n": 2519.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2126.0,
|
|
"completions/max_terminated_length": 2126.0,
|
|
"completions/mean_length": 369.046875,
|
|
"completions/mean_terminated_length": 373.4229431152344,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 88.0,
|
|
"epoch": 0.06293333333333333,
|
|
"grad_norm": 0.007437416817992926,
|
|
"learning_rate": 3.916666666666667e-06,
|
|
"loss": 0.0987,
|
|
"num_tokens": 13154647.0,
|
|
"reward": 0.9803744554519653,
|
|
"reward_std": 0.08235173672437668,
|
|
"rewards/accuracy_reward_step": 0.15625,
|
|
"rewards/final_brier_reward_step": 0.8162176012992859,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 59
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8610778823494911,
|
|
"aux_distill/mean_u": 0.31887360298002776,
|
|
"aux_distill/n_active_tok": 318.875,
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/auroc": 0.425182976416373,
|
|
"calib/avg_num_step_conf": 11.01953125,
|
|
"calib/ece": 0.11852589641434264,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0003103822174030796,
|
|
"calib/mean_conf": 0.09414342629482073,
|
|
"calib/mu_c": 0.09441176470588235,
|
|
"calib/mu_w": 0.09410138248847927,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.038605577689243026,
|
|
"calib/std_conf": 0.11483733232287055,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2889887640449438,
|
|
"calib/step_q_c_n": 267.0,
|
|
"calib/step_q_gap": 0.0642785056267762,
|
|
"calib/step_q_w": 0.22471025841816758,
|
|
"calib/step_q_w_n": 2554.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1242.0,
|
|
"completions/max_terminated_length": 1242.0,
|
|
"completions/mean_length": 341.07421875,
|
|
"completions/mean_terminated_length": 347.8685302734375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 102.0,
|
|
"epoch": 0.064,
|
|
"grad_norm": 0.0072919754311442375,
|
|
"learning_rate": 3.88888888888889e-06,
|
|
"loss": 0.0222,
|
|
"num_tokens": 13350818.0,
|
|
"reward": 0.9821978807449341,
|
|
"reward_std": 0.08395448327064514,
|
|
"rewards/accuracy_reward_step": 0.1328125,
|
|
"rewards/final_brier_reward_step": 0.8511144518852234,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 60
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8732533380389214,
|
|
"aux_distill/mean_u": 0.3002823455179606,
|
|
"aux_distill/n_active_tok": 284.75,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.4817826935588809,
|
|
"calib/avg_num_step_conf": 9.13671875,
|
|
"calib/ece": 0.17202734375000003,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.014920810484245768,
|
|
"calib/mean_conf": 0.06937890625000001,
|
|
"calib/mu_c": 0.05754716981132075,
|
|
"calib/mu_w": 0.07246798029556652,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.017187499999999998,
|
|
"calib/std_conf": 0.09771815030383921,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.20204999999999998,
|
|
"calib/step_q_c_n": 400.0,
|
|
"calib/step_q_gap": -0.04228935018050545,
|
|
"calib/step_q_w": 0.24433935018050543,
|
|
"calib/step_q_w_n": 1939.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1601.0,
|
|
"completions/max_terminated_length": 1601.0,
|
|
"completions/mean_length": 321.3125,
|
|
"completions/mean_terminated_length": 322.57257080078125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 60.0,
|
|
"epoch": 0.06506666666666666,
|
|
"grad_norm": 0.008903604932129383,
|
|
"learning_rate": 3.861111111111112e-06,
|
|
"loss": 0.0847,
|
|
"num_tokens": 13537138.0,
|
|
"reward": 1.0047328472137451,
|
|
"reward_std": 0.031484510749578476,
|
|
"rewards/accuracy_reward_step": 0.20703125,
|
|
"rewards/final_brier_reward_step": 0.8024345636367798,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 61
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8323895167559385,
|
|
"aux_distill/mean_u": 0.3086617885430675,
|
|
"aux_distill/n_active_tok": 315.0,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.4349730458221024,
|
|
"calib/avg_num_step_conf": 10.07421875,
|
|
"calib/ece": 0.13976377952755903,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.009975292003593893,
|
|
"calib/mean_conf": 0.05094488188976379,
|
|
"calib/mu_c": 0.042619047619047626,
|
|
"calib/mu_w": 0.05259433962264152,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.012677165354330711,
|
|
"calib/std_conf": 0.06526666162348307,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.20084745762711867,
|
|
"calib/step_q_c_n": 354.0,
|
|
"calib/step_q_gap": -0.021171868215577955,
|
|
"calib/step_q_w": 0.22201932584269662,
|
|
"calib/step_q_w_n": 2225.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 3002.0,
|
|
"completions/max_terminated_length": 3002.0,
|
|
"completions/mean_length": 364.2578125,
|
|
"completions/mean_terminated_length": 365.6863098144531,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 70.0,
|
|
"epoch": 0.06613333333333334,
|
|
"grad_norm": 0.007575937081128359,
|
|
"learning_rate": 3.833333333333334e-06,
|
|
"loss": 0.0982,
|
|
"num_tokens": 13737468.0,
|
|
"reward": 0.9957789182662964,
|
|
"reward_std": 0.03903381526470184,
|
|
"rewards/accuracy_reward_step": 0.1640625,
|
|
"rewards/final_brier_reward_step": 0.8353078365325928,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 62
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8655327968299389,
|
|
"aux_distill/mean_u": 0.3562405669229432,
|
|
"aux_distill/n_active_tok": 318.875,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.4805605247465712,
|
|
"calib/avg_num_step_conf": 9.96484375,
|
|
"calib/ece": 0.14362204724409453,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.00719618366129994,
|
|
"calib/mean_conf": 0.041732283464566935,
|
|
"calib/mu_c": 0.03564102564102564,
|
|
"calib/mu_w": 0.04283720930232558,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.01590551181102362,
|
|
"calib/std_conf": 0.07945279872663047,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22545270270270273,
|
|
"calib/step_q_c_n": 296.0,
|
|
"calib/step_q_gap": 0.015377048600707183,
|
|
"calib/step_q_w": 0.21007565410199555,
|
|
"calib/step_q_w_n": 2255.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2956.0,
|
|
"completions/max_terminated_length": 2956.0,
|
|
"completions/mean_length": 387.265625,
|
|
"completions/mean_terminated_length": 387.265625,
|
|
"completions/min_length": 51.0,
|
|
"completions/min_terminated_length": 51.0,
|
|
"epoch": 0.0672,
|
|
"grad_norm": 0.007319794036448002,
|
|
"learning_rate": 3.8055555555555556e-06,
|
|
"loss": 0.1671,
|
|
"num_tokens": 13945248.0,
|
|
"reward": 0.9936214089393616,
|
|
"reward_std": 0.039710745215415955,
|
|
"rewards/accuracy_reward_step": 0.15234375,
|
|
"rewards/final_brier_reward_step": 0.8427116870880127,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 63
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8316024951636791,
|
|
"aux_distill/mean_u": 0.2873626919249946,
|
|
"aux_distill/n_active_tok": 302.375,
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/auroc": 0.4885526921306949,
|
|
"calib/avg_num_step_conf": 9.80078125,
|
|
"calib/ece": 0.13798418972332016,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.004450069028992171,
|
|
"calib/mean_conf": 0.03324110671936759,
|
|
"calib/mu_c": 0.029512195121951225,
|
|
"calib/mu_w": 0.033962264150943396,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.00458498023715415,
|
|
"calib/std_conf": 0.048264025690469205,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.16411924119241192,
|
|
"calib/step_q_c_n": 369.0,
|
|
"calib/step_q_gap": -0.06580786161132643,
|
|
"calib/step_q_w": 0.22992710280373835,
|
|
"calib/step_q_w_n": 2140.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2231.0,
|
|
"completions/max_terminated_length": 2231.0,
|
|
"completions/mean_length": 334.80078125,
|
|
"completions/mean_terminated_length": 336.1137390136719,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 101.0,
|
|
"epoch": 0.06826666666666667,
|
|
"grad_norm": 0.008024153299629688,
|
|
"learning_rate": 3.777777777777778e-06,
|
|
"loss": 0.1567,
|
|
"num_tokens": 14134733.0,
|
|
"reward": 0.991310715675354,
|
|
"reward_std": 0.04444682598114014,
|
|
"rewards/accuracy_reward_step": 0.16015625,
|
|
"rewards/final_brier_reward_step": 0.8341840505599976,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 64
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8422551490366459,
|
|
"aux_distill/mean_u": 0.2814362829432169,
|
|
"aux_distill/n_active_tok": 255.75,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.44934605488850776,
|
|
"calib/avg_num_step_conf": 8.07421875,
|
|
"calib/ece": 0.14957031250000002,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.00016723842195540128,
|
|
"calib/mean_conf": 0.023320312500000002,
|
|
"calib/mu_c": 0.023181818181818182,
|
|
"calib/mu_w": 0.023349056603773583,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0005078125,
|
|
"calib/std_conf": 0.027191288125102565,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.1940531561461794,
|
|
"calib/step_q_c_n": 301.0,
|
|
"calib/step_q_gap": -0.005558961634115039,
|
|
"calib/step_q_w": 0.19961211778029445,
|
|
"calib/step_q_w_n": 1766.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1142.0,
|
|
"completions/max_terminated_length": 1142.0,
|
|
"completions/mean_length": 298.44921875,
|
|
"completions/mean_terminated_length": 299.61962890625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 85.0,
|
|
"epoch": 0.06933333333333333,
|
|
"grad_norm": 0.008111841976642609,
|
|
"learning_rate": 3.7500000000000005e-06,
|
|
"loss": 0.0778,
|
|
"num_tokens": 14316160.0,
|
|
"reward": 1.0033427476882935,
|
|
"reward_std": 0.00751079898327589,
|
|
"rewards/accuracy_reward_step": 0.171875,
|
|
"rewards/final_brier_reward_step": 0.8348105549812317,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 65
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8705679681152105,
|
|
"aux_distill/mean_u": 0.3264017254864122,
|
|
"aux_distill/n_active_tok": 282.25,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.4394767441860465,
|
|
"calib/avg_num_step_conf": 8.8203125,
|
|
"calib/ece": 0.14352941176470588,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0030523255813953484,
|
|
"calib/mean_conf": 0.014823529411764708,
|
|
"calib/mu_c": 0.01225,
|
|
"calib/mu_w": 0.015302325581395349,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0007450980392156863,
|
|
"calib/std_conf": 0.019937850379098822,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.20936305732484078,
|
|
"calib/step_q_c_n": 314.0,
|
|
"calib/step_q_gap": 0.043277666378338714,
|
|
"calib/step_q_w": 0.16608539094650207,
|
|
"calib/step_q_w_n": 1944.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2218.0,
|
|
"completions/max_terminated_length": 2218.0,
|
|
"completions/mean_length": 342.12109375,
|
|
"completions/mean_terminated_length": 342.12109375,
|
|
"completions/min_length": 92.0,
|
|
"completions/min_terminated_length": 92.0,
|
|
"epoch": 0.0704,
|
|
"grad_norm": 0.007121607661247253,
|
|
"learning_rate": 3.7222222222222225e-06,
|
|
"loss": 0.1123,
|
|
"num_tokens": 14510095.0,
|
|
"reward": 0.9977003931999207,
|
|
"reward_std": 0.014767561107873917,
|
|
"rewards/accuracy_reward_step": 0.15625,
|
|
"rewards/final_brier_reward_step": 0.8430570363998413,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 66
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8964165635406971,
|
|
"aux_distill/mean_u": 0.30745739046953435,
|
|
"aux_distill/n_active_tok": 243.0,
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/auroc": 0.4806829268292683,
|
|
"calib/avg_num_step_conf": 7.59375,
|
|
"calib/ece": 0.1858270588235294,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.004717560975609755,
|
|
"calib/mean_conf": 0.015192549019607844,
|
|
"calib/mu_c": 0.0114,
|
|
"calib/mu_w": 0.016117560975609756,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0024705882352941176,
|
|
"calib/std_conf": 0.032740294526154796,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.17552380952380955,
|
|
"calib/step_q_c_n": 315.0,
|
|
"calib/step_q_gap": 0.0008153994562834499,
|
|
"calib/step_q_w": 0.1747084100675261,
|
|
"calib/step_q_w_n": 1629.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2851.0,
|
|
"completions/max_terminated_length": 2851.0,
|
|
"completions/mean_length": 320.25,
|
|
"completions/mean_terminated_length": 320.25,
|
|
"completions/min_length": 81.0,
|
|
"completions/min_terminated_length": 81.0,
|
|
"epoch": 0.07146666666666666,
|
|
"grad_norm": 0.008333339355885983,
|
|
"learning_rate": 3.694444444444445e-06,
|
|
"loss": 0.1222,
|
|
"num_tokens": 14697087.0,
|
|
"reward": 0.9898937344551086,
|
|
"reward_std": 0.038719628006219864,
|
|
"rewards/accuracy_reward_step": 0.1953125,
|
|
"rewards/final_brier_reward_step": 0.7961937189102173,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 67
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8725739866495132,
|
|
"aux_distill/mean_u": 0.29536300696172413,
|
|
"aux_distill/n_active_tok": 247.375,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.511605415860735,
|
|
"calib/avg_num_step_conf": 7.7734375,
|
|
"calib/ece": 0.17578125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.000989514404967931,
|
|
"calib/mean_conf": 0.00953125,
|
|
"calib/mu_c": 0.00872340425531915,
|
|
"calib/mu_w": 0.009712918660287081,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.000859375,
|
|
"calib/std_conf": 0.01784741643592988,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.17542372881355936,
|
|
"calib/step_q_c_n": 236.0,
|
|
"calib/step_q_gap": 0.02437241752507588,
|
|
"calib/step_q_w": 0.15105131128848348,
|
|
"calib/step_q_w_n": 1754.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 791.0,
|
|
"completions/max_terminated_length": 791.0,
|
|
"completions/mean_length": 297.390625,
|
|
"completions/mean_terminated_length": 298.556884765625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 82.0,
|
|
"epoch": 0.07253333333333334,
|
|
"grad_norm": 0.007754533085972071,
|
|
"learning_rate": 3.6666666666666666e-06,
|
|
"loss": 0.1273,
|
|
"num_tokens": 14877307.0,
|
|
"reward": 0.9974908828735352,
|
|
"reward_std": 0.014132047072052956,
|
|
"rewards/accuracy_reward_step": 0.18359375,
|
|
"rewards/final_brier_reward_step": 0.8152941465377808,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 68
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8729984704405069,
|
|
"aux_distill/mean_u": 0.29370034451770377,
|
|
"aux_distill/n_active_tok": 250.0,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.646583850931677,
|
|
"calib/avg_num_step_conf": 8.0625,
|
|
"calib/ece": 0.17317226395737534,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.001344695271032011,
|
|
"calib/mean_conf": 0.009119402709291341,
|
|
"calib/mu_c": 0.010222473048809789,
|
|
"calib/mu_w": 0.008877777777777778,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0013020833333333333,
|
|
"calib/std_conf": 0.023883040880635267,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2219811320754717,
|
|
"calib/step_q_c_n": 318.0,
|
|
"calib/step_q_gap": 0.08574401867340986,
|
|
"calib/step_q_w": 0.13623711340206185,
|
|
"calib/step_q_w_n": 1746.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1646.0,
|
|
"completions/max_terminated_length": 1646.0,
|
|
"completions/mean_length": 318.09375,
|
|
"completions/mean_terminated_length": 319.3411865234375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 77.0,
|
|
"epoch": 0.0736,
|
|
"grad_norm": 0.00929983053356409,
|
|
"learning_rate": 3.638888888888889e-06,
|
|
"loss": 0.0899,
|
|
"num_tokens": 15063235.0,
|
|
"reward": 1.0015100240707397,
|
|
"reward_std": 0.003858521580696106,
|
|
"rewards/accuracy_reward_step": 0.1796875,
|
|
"rewards/final_brier_reward_step": 0.823332667350769,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 69
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9363727066665888,
|
|
"aux_distill/mean_u": 0.3052057791038958,
|
|
"aux_distill/n_active_tok": 213.5,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5160601976639713,
|
|
"calib/avg_num_step_conf": 6.83984375,
|
|
"calib/ece": 0.1601968503937008,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.00018867924528301935,
|
|
"calib/mean_conf": 0.00515748031496063,
|
|
"calib/mu_c": 0.005,
|
|
"calib/mu_w": 0.0051886792452830195,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.009504184902402799,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.14195833333333335,
|
|
"calib/step_q_c_n": 240.0,
|
|
"calib/step_q_gap": -0.004583228546216622,
|
|
"calib/step_q_w": 0.14654156187954998,
|
|
"calib/step_q_w_n": 1511.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1461.0,
|
|
"completions/max_terminated_length": 1461.0,
|
|
"completions/mean_length": 272.34375,
|
|
"completions/mean_terminated_length": 273.4117736816406,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 68.0,
|
|
"epoch": 0.07466666666666667,
|
|
"grad_norm": 0.008598068729043007,
|
|
"learning_rate": 3.6111111111111115e-06,
|
|
"loss": 0.0989,
|
|
"num_tokens": 15239947.0,
|
|
"reward": 0.9929498434066772,
|
|
"reward_std": 0.023966234177350998,
|
|
"rewards/accuracy_reward_step": 0.1640625,
|
|
"rewards/final_brier_reward_step": 0.8296495676040649,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 70
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8899203259497881,
|
|
"aux_distill/mean_u": 0.3015708150804158,
|
|
"aux_distill/n_active_tok": 220.25,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.542351892213196,
|
|
"calib/avg_num_step_conf": 7.19921875,
|
|
"calib/ece": 0.18811764705882353,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0013156330493362387,
|
|
"calib/mean_conf": 0.004039215686274511,
|
|
"calib/mu_c": 0.00510204081632653,
|
|
"calib/mu_w": 0.0037864077669902915,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006842659683318667,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2603875968992248,
|
|
"calib/step_q_c_n": 258.0,
|
|
"calib/step_q_gap": 0.12090494705695351,
|
|
"calib/step_q_w": 0.1394826498422713,
|
|
"calib/step_q_w_n": 1585.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 953.0,
|
|
"completions/max_terminated_length": 953.0,
|
|
"completions/mean_length": 279.7421875,
|
|
"completions/mean_terminated_length": 280.8392333984375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 66.0,
|
|
"epoch": 0.07573333333333333,
|
|
"grad_norm": 0.00749945780262351,
|
|
"learning_rate": 3.5833333333333335e-06,
|
|
"loss": 0.0561,
|
|
"num_tokens": 15415969.0,
|
|
"reward": 0.9931343793869019,
|
|
"reward_std": 0.02386919967830181,
|
|
"rewards/accuracy_reward_step": 0.19140625,
|
|
"rewards/final_brier_reward_step": 0.8026750087738037,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 71
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9215385001152754,
|
|
"aux_distill/mean_u": 0.2936664698069263,
|
|
"aux_distill/n_active_tok": 237.0,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.4305194805194805,
|
|
"calib/avg_num_step_conf": 7.515625,
|
|
"calib/ece": 0.13246666666666668,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0015759740259740258,
|
|
"calib/mean_conf": 0.004788235294117647,
|
|
"calib/mu_c": 0.0034285714285714284,
|
|
"calib/mu_w": 0.005004545454545454,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007296135078374823,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.18448888888888892,
|
|
"calib/step_q_c_n": 225.0,
|
|
"calib/step_q_gap": -0.006087920999280599,
|
|
"calib/step_q_w": 0.19057680988816952,
|
|
"calib/step_q_w_n": 1699.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 966.0,
|
|
"completions/max_terminated_length": 966.0,
|
|
"completions/mean_length": 289.51953125,
|
|
"completions/mean_terminated_length": 290.6549072265625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 94.0,
|
|
"epoch": 0.0768,
|
|
"grad_norm": 0.006771552842110395,
|
|
"learning_rate": 3.555555555555556e-06,
|
|
"loss": 0.0924,
|
|
"num_tokens": 15594494.0,
|
|
"reward": 0.9965245127677917,
|
|
"reward_std": 0.012207376770675182,
|
|
"rewards/accuracy_reward_step": 0.13671875,
|
|
"rewards/final_brier_reward_step": 0.860236644744873,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 72
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8665466215461493,
|
|
"aux_distill/mean_u": 0.23089176820613588,
|
|
"aux_distill/n_active_tok": 229.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5551916148051098,
|
|
"calib/avg_num_step_conf": 7.23828125,
|
|
"calib/ece": 0.16430078125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0022997052079921387,
|
|
"calib/mean_conf": 0.0036679687500000002,
|
|
"calib/mu_c": 0.005581395348837209,
|
|
"calib/mu_w": 0.0032816901408450707,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.0071032500659221785,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22144396551724138,
|
|
"calib/step_q_c_n": 232.0,
|
|
"calib/step_q_gap": -0.001768804377885097,
|
|
"calib/step_q_w": 0.22321276989512648,
|
|
"calib/step_q_w_n": 1621.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1205.0,
|
|
"completions/max_terminated_length": 1205.0,
|
|
"completions/mean_length": 296.3671875,
|
|
"completions/mean_terminated_length": 297.5294189453125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 81.0,
|
|
"epoch": 0.07786666666666667,
|
|
"grad_norm": 0.007758422289043665,
|
|
"learning_rate": 3.5277777777777784e-06,
|
|
"loss": 0.0508,
|
|
"num_tokens": 15777396.0,
|
|
"reward": 1.0009055137634277,
|
|
"reward_std": 0.002186012454330921,
|
|
"rewards/accuracy_reward_step": 0.16796875,
|
|
"rewards/final_brier_reward_step": 0.8338423371315002,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 73
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9035820104181767,
|
|
"aux_distill/mean_u": 0.2673429256586252,
|
|
"aux_distill/n_active_tok": 233.875,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5504227053140097,
|
|
"calib/avg_num_step_conf": 7.3828125,
|
|
"calib/ece": 0.18501176470588235,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0016751207729468602,
|
|
"calib/mean_conf": 0.0032235294117647063,
|
|
"calib/mu_c": 0.004583333333333333,
|
|
"calib/mu_w": 0.002908212560386473,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.0063103583763024305,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.21533112582781458,
|
|
"calib/step_q_c_n": 302.0,
|
|
"calib/step_q_gap": 0.024556566633859883,
|
|
"calib/step_q_w": 0.1907745591939547,
|
|
"calib/step_q_w_n": 1588.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 993.0,
|
|
"completions/max_terminated_length": 993.0,
|
|
"completions/mean_length": 289.82421875,
|
|
"completions/mean_terminated_length": 290.9608154296875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 73.0,
|
|
"epoch": 0.07893333333333333,
|
|
"grad_norm": 0.007173248566687107,
|
|
"learning_rate": 3.5e-06,
|
|
"loss": 0.0995,
|
|
"num_tokens": 15955519.0,
|
|
"reward": 0.9930218458175659,
|
|
"reward_std": 0.02364840731024742,
|
|
"rewards/accuracy_reward_step": 0.1875,
|
|
"rewards/final_brier_reward_step": 0.8063561916351318,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 74
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8751150369644165,
|
|
"aux_distill/mean_u": 0.2630927631062383,
|
|
"aux_distill/n_active_tok": 224.375,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5070944972630366,
|
|
"calib/avg_num_step_conf": 7.0625,
|
|
"calib/ece": 0.30078125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 9.795447997695264e-05,
|
|
"calib/mean_conf": 0.00390625,
|
|
"calib/mu_c": 0.003974358974358975,
|
|
"calib/mu_w": 0.0038764044943820227,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006463161837483261,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24687037037037038,
|
|
"calib/step_q_c_n": 540.0,
|
|
"calib/step_q_gap": 0.029165323051758396,
|
|
"calib/step_q_w": 0.21770504731861198,
|
|
"calib/step_q_w_n": 1268.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 964.0,
|
|
"completions/max_terminated_length": 964.0,
|
|
"completions/mean_length": 283.9296875,
|
|
"completions/mean_terminated_length": 285.04315185546875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 53.0,
|
|
"epoch": 0.08,
|
|
"grad_norm": 0.010993538424372673,
|
|
"learning_rate": 3.4722222222222224e-06,
|
|
"loss": 0.0414,
|
|
"num_tokens": 16132957.0,
|
|
"reward": 1.0011824369430542,
|
|
"reward_std": 0.0027043407317250967,
|
|
"rewards/accuracy_reward_step": 0.3046875,
|
|
"rewards/final_brier_reward_step": 0.6976773738861084,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 75
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9287750497460365,
|
|
"aux_distill/mean_u": 0.2989380853039391,
|
|
"aux_distill/n_active_tok": 217.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5268634622082897,
|
|
"calib/avg_num_step_conf": 6.8671875,
|
|
"calib/ece": 0.222890625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0006025774991292232,
|
|
"calib/mean_conf": 0.003671875,
|
|
"calib/mu_c": 0.004137931034482759,
|
|
"calib/mu_w": 0.0035353535353535356,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006658065333441465,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22013698630136985,
|
|
"calib/step_q_c_n": 365.0,
|
|
"calib/step_q_gap": -0.016203717216218133,
|
|
"calib/step_q_w": 0.23634070351758799,
|
|
"calib/step_q_w_n": 1393.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1078.0,
|
|
"completions/max_terminated_length": 1078.0,
|
|
"completions/mean_length": 271.57421875,
|
|
"completions/mean_terminated_length": 272.63922119140625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 77.0,
|
|
"epoch": 0.08106666666666666,
|
|
"grad_norm": 0.008762000128626823,
|
|
"learning_rate": 3.444444444444445e-06,
|
|
"loss": 0.0686,
|
|
"num_tokens": 16305536.0,
|
|
"reward": 1.000908613204956,
|
|
"reward_std": 0.002230893587693572,
|
|
"rewards/accuracy_reward_step": 0.2265625,
|
|
"rewards/final_brier_reward_step": 0.7752546668052673,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 76
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8958236277103424,
|
|
"aux_distill/mean_u": 0.27076952840991,
|
|
"aux_distill/n_active_tok": 247.125,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5445985708280789,
|
|
"calib/avg_num_step_conf": 7.78515625,
|
|
"calib/ece": 0.2348828125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0015645229087852043,
|
|
"calib/mean_conf": 0.0033984375,
|
|
"calib/mu_c": 0.00459016393442623,
|
|
"calib/mu_w": 0.003025641025641026,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006478136117633972,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24567200000000003,
|
|
"calib/step_q_c_n": 375.0,
|
|
"calib/step_q_gap": 0.000832692212608177,
|
|
"calib/step_q_w": 0.24483930778739185,
|
|
"calib/step_q_w_n": 1618.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1003.0,
|
|
"completions/max_terminated_length": 1003.0,
|
|
"completions/mean_length": 303.20703125,
|
|
"completions/mean_terminated_length": 304.3960876464844,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 63.0,
|
|
"epoch": 0.08213333333333334,
|
|
"grad_norm": 0.00724209938198328,
|
|
"learning_rate": 3.416666666666667e-06,
|
|
"loss": 0.1019,
|
|
"num_tokens": 16487821.0,
|
|
"reward": 1.0010669231414795,
|
|
"reward_std": 0.002557475585490465,
|
|
"rewards/accuracy_reward_step": 0.23828125,
|
|
"rewards/final_brier_reward_step": 0.7638527154922485,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 77
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.886054253205657,
|
|
"aux_distill/mean_u": 0.2860843198626096,
|
|
"aux_distill/n_active_tok": 292.375,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.48472781375809554,
|
|
"calib/avg_num_step_conf": 9.13671875,
|
|
"calib/ece": 0.22411764705882353,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0009670925958340628,
|
|
"calib/mean_conf": 0.0033333333333333335,
|
|
"calib/mu_c": 0.0025862068965517245,
|
|
"calib/mu_w": 0.0035532994923857873,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007212915068878163,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2432278481012658,
|
|
"calib/step_q_c_n": 474.0,
|
|
"calib/step_q_gap": -0.02152657549122744,
|
|
"calib/step_q_w": 0.26475442359249324,
|
|
"calib/step_q_w_n": 1865.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2417.0,
|
|
"completions/max_terminated_length": 2417.0,
|
|
"completions/mean_length": 373.08984375,
|
|
"completions/mean_terminated_length": 373.08984375,
|
|
"completions/min_length": 97.0,
|
|
"completions/min_terminated_length": 97.0,
|
|
"epoch": 0.0832,
|
|
"grad_norm": 0.008203305304050446,
|
|
"learning_rate": 3.3888888888888893e-06,
|
|
"loss": 0.1184,
|
|
"num_tokens": 16691356.0,
|
|
"reward": 0.9966482520103455,
|
|
"reward_std": 0.012584471143782139,
|
|
"rewards/accuracy_reward_step": 0.2265625,
|
|
"rewards/final_brier_reward_step": 0.7706402540206909,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 78
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8530756682157516,
|
|
"aux_distill/mean_u": 0.27669342164705013,
|
|
"aux_distill/n_active_tok": 306.5,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.4914740431981811,
|
|
"calib/avg_num_step_conf": 9.67578125,
|
|
"calib/ece": 0.20047058823529412,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0004698749526335739,
|
|
"calib/mean_conf": 0.003450980392156863,
|
|
"calib/mu_c": 0.003076923076923077,
|
|
"calib/mu_w": 0.003546798029556651,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006250690081087976,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.2234047619047619,
|
|
"calib/step_q_c_n": 420.0,
|
|
"calib/step_q_gap": -0.022195627011135033,
|
|
"calib/step_q_w": 0.24560038891589694,
|
|
"calib/step_q_w_n": 2057.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1474.0,
|
|
"completions/max_terminated_length": 1474.0,
|
|
"completions/mean_length": 372.7421875,
|
|
"completions/mean_terminated_length": 374.2039489746094,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 83.0,
|
|
"epoch": 0.08426666666666667,
|
|
"grad_norm": 0.0066146100871264935,
|
|
"learning_rate": 3.3611111111111117e-06,
|
|
"loss": 0.0719,
|
|
"num_tokens": 16893154.0,
|
|
"reward": 0.9927873611450195,
|
|
"reward_std": 0.023470385000109673,
|
|
"rewards/accuracy_reward_step": 0.203125,
|
|
"rewards/final_brier_reward_step": 0.7902621030807495,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 79
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8846382517367601,
|
|
"aux_distill/mean_u": 0.3161391397071147,
|
|
"aux_distill/n_active_tok": 302.375,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5348426933254925,
|
|
"calib/avg_num_step_conf": 9.72265625,
|
|
"calib/ece": 0.29423529411764704,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0013290208762128776,
|
|
"calib/mean_conf": 0.003803921568627451,
|
|
"calib/mu_c": 0.004736842105263157,
|
|
"calib/mu_w": 0.0034078212290502797,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006747605553063313,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2645139028475712,
|
|
"calib/step_q_c_n": 597.0,
|
|
"calib/step_q_gap": 0.03888441024714839,
|
|
"calib/step_q_w": 0.2256294926004228,
|
|
"calib/step_q_w_n": 1892.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1072.0,
|
|
"completions/max_terminated_length": 1072.0,
|
|
"completions/mean_length": 355.96875,
|
|
"completions/mean_terminated_length": 357.3647155761719,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 85.0,
|
|
"epoch": 0.08533333333333333,
|
|
"grad_norm": 0.007952683605253696,
|
|
"learning_rate": 3.3333333333333333e-06,
|
|
"loss": 0.1073,
|
|
"num_tokens": 17086442.0,
|
|
"reward": 0.9974700808525085,
|
|
"reward_std": 0.013819643296301365,
|
|
"rewards/accuracy_reward_step": 0.296875,
|
|
"rewards/final_brier_reward_step": 0.7019714713096619,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 80
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8901407215744257,
|
|
"aux_distill/mean_u": 0.285899711043199,
|
|
"aux_distill/n_active_tok": 327.125,
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/auroc": 0.5391246684350133,
|
|
"calib/avg_num_step_conf": 11.09765625,
|
|
"calib/ece": 0.3053174603174603,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0004067197170645447,
|
|
"calib/mean_conf": 0.004206349206349207,
|
|
"calib/mu_c": 0.004487179487179487,
|
|
"calib/mu_w": 0.004080459770114942,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.00815068159611388,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2041506646971935,
|
|
"calib/step_q_c_n": 677.0,
|
|
"calib/step_q_gap": -0.04997927984994144,
|
|
"calib/step_q_w": 0.25412994454713494,
|
|
"calib/step_q_w_n": 2164.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2698.0,
|
|
"completions/max_terminated_length": 2698.0,
|
|
"completions/mean_length": 390.67578125,
|
|
"completions/mean_terminated_length": 395.3083190917969,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 71.0,
|
|
"epoch": 0.0864,
|
|
"grad_norm": 0.00825242418795824,
|
|
"learning_rate": 3.3055555555555558e-06,
|
|
"loss": 0.091,
|
|
"num_tokens": 17292703.0,
|
|
"reward": 0.9857007265090942,
|
|
"reward_std": 0.03914933651685715,
|
|
"rewards/accuracy_reward_step": 0.3046875,
|
|
"rewards/final_brier_reward_step": 0.6823390126228333,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 81
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8487306144088507,
|
|
"aux_distill/mean_u": 0.265551814483768,
|
|
"aux_distill/n_active_tok": 319.5,
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/auroc": 0.472996254681648,
|
|
"calib/avg_num_step_conf": 10.27734375,
|
|
"calib/ece": 0.29284584980237155,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0011325842696629207,
|
|
"calib/mean_conf": 0.0035968379446640318,
|
|
"calib/mu_c": 0.0028,
|
|
"calib/mu_w": 0.003932584269662921,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006661457030128761,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.27258389261744964,
|
|
"calib/step_q_c_n": 596.0,
|
|
"calib/step_q_gap": 0.020334261167818224,
|
|
"calib/step_q_w": 0.2522496314496314,
|
|
"calib/step_q_w_n": 2035.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2420.0,
|
|
"completions/max_terminated_length": 2420.0,
|
|
"completions/mean_length": 392.7109375,
|
|
"completions/mean_terminated_length": 395.80316162109375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 96.0,
|
|
"epoch": 0.08746666666666666,
|
|
"grad_norm": 0.006242407485842705,
|
|
"learning_rate": 3.277777777777778e-06,
|
|
"loss": 0.111,
|
|
"num_tokens": 17498789.0,
|
|
"reward": 0.9851671457290649,
|
|
"reward_std": 0.045587096363306046,
|
|
"rewards/accuracy_reward_step": 0.29296875,
|
|
"rewards/final_brier_reward_step": 0.6929906010627747,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 82
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.844496738165617,
|
|
"aux_distill/mean_u": 0.27586103866241396,
|
|
"aux_distill/n_active_tok": 350.875,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.5069444444444444,
|
|
"calib/avg_num_step_conf": 11.12109375,
|
|
"calib/ece": 0.28003937007874014,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0003220390720390723,
|
|
"calib/mean_conf": 0.003425196850393701,
|
|
"calib/mu_c": 0.003194444444444444,
|
|
"calib/mu_w": 0.0035164835164835165,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006122142187278071,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.21304761904761907,
|
|
"calib/step_q_c_n": 651.0,
|
|
"calib/step_q_gap": -0.028655104085349964,
|
|
"calib/step_q_w": 0.24170272313296903,
|
|
"calib/step_q_w_n": 2196.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1536.0,
|
|
"completions/max_terminated_length": 1536.0,
|
|
"completions/mean_length": 437.68359375,
|
|
"completions/mean_terminated_length": 439.4000244140625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 99.0,
|
|
"epoch": 0.08853333333333334,
|
|
"grad_norm": 0.006423268001526594,
|
|
"learning_rate": 3.2500000000000002e-06,
|
|
"loss": 0.0654,
|
|
"num_tokens": 17718100.0,
|
|
"reward": 0.9891560077667236,
|
|
"reward_std": 0.035011157393455505,
|
|
"rewards/accuracy_reward_step": 0.28125,
|
|
"rewards/final_brier_reward_step": 0.7087808847427368,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 83
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8821609672158957,
|
|
"aux_distill/mean_u": 0.29654691763738605,
|
|
"aux_distill/n_active_tok": 311.625,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5858699411968177,
|
|
"calib/avg_num_step_conf": 9.92578125,
|
|
"calib/ece": 0.22686274509803922,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.002512106537530266,
|
|
"calib/mean_conf": 0.004509803921568627,
|
|
"calib/mu_c": 0.006440677966101695,
|
|
"calib/mu_w": 0.003928571428571429,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007171218478847034,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24171568627450982,
|
|
"calib/step_q_c_n": 408.0,
|
|
"calib/step_q_gap": 0.006632329500013773,
|
|
"calib/step_q_w": 0.23508335677449604,
|
|
"calib/step_q_w_n": 2133.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1109.0,
|
|
"completions/max_terminated_length": 1109.0,
|
|
"completions/mean_length": 377.25390625,
|
|
"completions/mean_terminated_length": 378.7333679199219,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 61.0,
|
|
"epoch": 0.0896,
|
|
"grad_norm": 0.007551021408289671,
|
|
"learning_rate": 3.2222222222222227e-06,
|
|
"loss": 0.0945,
|
|
"num_tokens": 17920597.0,
|
|
"reward": 0.9975423812866211,
|
|
"reward_std": 0.013799784705042839,
|
|
"rewards/accuracy_reward_step": 0.23046875,
|
|
"rewards/final_brier_reward_step": 0.7685222625732422,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 84
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8072050474584103,
|
|
"aux_distill/mean_u": 0.27407098827109644,
|
|
"aux_distill/n_active_tok": 381.0,
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/auroc": 0.4070313968791126,
|
|
"calib/avg_num_step_conf": 12.80078125,
|
|
"calib/ece": 0.21127487250996016,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.002328480729460425,
|
|
"calib/mean_conf": 0.003864569721115538,
|
|
"calib/mu_c": 0.002037037037037037,
|
|
"calib/mu_w": 0.004365517766497462,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.0061037968614311745,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24305845511482252,
|
|
"calib/step_q_c_n": 479.0,
|
|
"calib/step_q_gap": 0.02982529535785325,
|
|
"calib/step_q_w": 0.21323315975696927,
|
|
"calib/step_q_w_n": 2798.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1740.0,
|
|
"completions/max_terminated_length": 1740.0,
|
|
"completions/mean_length": 446.6015625,
|
|
"completions/mean_terminated_length": 455.4980163574219,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 96.0,
|
|
"epoch": 0.09066666666666667,
|
|
"grad_norm": 0.005789844784885645,
|
|
"learning_rate": 3.1944444444444443e-06,
|
|
"loss": 0.0193,
|
|
"num_tokens": 18142751.0,
|
|
"reward": 0.9769666194915771,
|
|
"reward_std": 0.05974484235048294,
|
|
"rewards/accuracy_reward_step": 0.2109375,
|
|
"rewards/final_brier_reward_step": 0.7664331793785095,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 85
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8167195729911327,
|
|
"aux_distill/mean_u": 0.27448563426857864,
|
|
"aux_distill/n_active_tok": 428.875,
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/auroc": 0.5223398001175779,
|
|
"calib/avg_num_step_conf": 13.859375,
|
|
"calib/ece": 0.24583333333333332,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0010052910052910057,
|
|
"calib/mean_conf": 0.0041666666666666675,
|
|
"calib/mu_c": 0.004920634920634922,
|
|
"calib/mu_w": 0.003915343915343916,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006147976825559694,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.24684210526315786,
|
|
"calib/step_q_c_n": 608.0,
|
|
"calib/step_q_gap": 0.03063003043322593,
|
|
"calib/step_q_w": 0.21621207482993193,
|
|
"calib/step_q_w_n": 2940.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2912.0,
|
|
"completions/max_terminated_length": 2912.0,
|
|
"completions/mean_length": 503.32421875,
|
|
"completions/mean_terminated_length": 507.28741455078125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 98.0,
|
|
"epoch": 0.09173333333333333,
|
|
"grad_norm": 0.0065255239605903625,
|
|
"learning_rate": 3.1666666666666667e-06,
|
|
"loss": 0.1452,
|
|
"num_tokens": 18377114.0,
|
|
"reward": 0.9777464270591736,
|
|
"reward_std": 0.06083172559738159,
|
|
"rewards/accuracy_reward_step": 0.24609375,
|
|
"rewards/final_brier_reward_step": 0.7328367233276367,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 86
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7748824749141932,
|
|
"aux_distill/mean_u": 0.25857830306465757,
|
|
"aux_distill/n_active_tok": 381.625,
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/auroc": 0.4614916286149163,
|
|
"calib/avg_num_step_conf": 12.015625,
|
|
"calib/ece": 0.2851383399209486,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0029421613394216128,
|
|
"calib/mean_conf": 0.005928853754940711,
|
|
"calib/mu_c": 0.0038356164383561648,
|
|
"calib/mu_w": 0.0067777777777777775,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0012648221343873518,
|
|
"calib/std_conf": 0.021050167759847967,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.1767248322147651,
|
|
"calib/step_q_c_n": 745.0,
|
|
"calib/step_q_gap": -0.07141240502247212,
|
|
"calib/step_q_w": 0.2481372372372372,
|
|
"calib/step_q_w_n": 2331.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3010.0,
|
|
"completions/max_terminated_length": 3010.0,
|
|
"completions/mean_length": 451.01171875,
|
|
"completions/mean_terminated_length": 454.56298828125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 71.0,
|
|
"epoch": 0.0928,
|
|
"grad_norm": 0.0069386824034154415,
|
|
"learning_rate": 3.138888888888889e-06,
|
|
"loss": 0.0785,
|
|
"num_tokens": 18598069.0,
|
|
"reward": 0.9891386032104492,
|
|
"reward_std": 0.03575124591588974,
|
|
"rewards/accuracy_reward_step": 0.28515625,
|
|
"rewards/final_brier_reward_step": 0.704839825630188,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 87
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7960807308554649,
|
|
"aux_distill/mean_u": 0.26279148734187663,
|
|
"aux_distill/n_active_tok": 447.125,
|
|
"calib/answer_extract_rate": 0.95703125,
|
|
"calib/auroc": 0.4537837837837837,
|
|
"calib/avg_num_step_conf": 16.36328125,
|
|
"calib/ece": 0.23917551020408162,
|
|
"calib/final_conf_rate": 0.95703125,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.001574774774774774,
|
|
"calib/mean_conf": 0.005722448979591837,
|
|
"calib/mu_c": 0.004533333333333334,
|
|
"calib/mu_w": 0.006108108108108108,
|
|
"calib/nonempty_final_conf_rate": 0.95703125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008126832987742656,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.17887025139664806,
|
|
"calib/step_q_c_n": 716.0,
|
|
"calib/step_q_gap": -0.027347686985154412,
|
|
"calib/step_q_w": 0.20621793838180247,
|
|
"calib/step_q_w_n": 3473.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.03515625,
|
|
"completions/max_length": 2677.0,
|
|
"completions/max_terminated_length": 2677.0,
|
|
"completions/mean_length": 515.9921875,
|
|
"completions/mean_terminated_length": 534.7935180664062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 109.0,
|
|
"epoch": 0.09386666666666667,
|
|
"grad_norm": 0.005010412540286779,
|
|
"learning_rate": 3.1111111111111116e-06,
|
|
"loss": 0.0542,
|
|
"num_tokens": 18840011.0,
|
|
"reward": 0.9580464959144592,
|
|
"reward_std": 0.11547387391328812,
|
|
"rewards/accuracy_reward_step": 0.234375,
|
|
"rewards/final_brier_reward_step": 0.7246866822242737,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 88
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8258982449769974,
|
|
"aux_distill/mean_u": 0.32058147032185685,
|
|
"aux_distill/n_active_tok": 455.375,
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/auroc": 0.5010034904013961,
|
|
"calib/avg_num_step_conf": 15.20703125,
|
|
"calib/ece": 0.23350996015936257,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0006125654450261789,
|
|
"calib/mean_conf": 0.0055338645418326685,
|
|
"calib/mu_c": 0.006,
|
|
"calib/mu_w": 0.005387434554973821,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007922548925671525,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2176074498567335,
|
|
"calib/step_q_c_n": 698.0,
|
|
"calib/step_q_gap": 0.012401215520893905,
|
|
"calib/step_q_w": 0.2052062343358396,
|
|
"calib/step_q_w_n": 3192.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3050.0,
|
|
"completions/max_terminated_length": 3050.0,
|
|
"completions/mean_length": 543.171875,
|
|
"completions/mean_terminated_length": 551.793701171875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 103.0,
|
|
"epoch": 0.09493333333333333,
|
|
"grad_norm": 0.005261671729385853,
|
|
"learning_rate": 3.0833333333333336e-06,
|
|
"loss": 0.087,
|
|
"num_tokens": 19087951.0,
|
|
"reward": 0.9740170836448669,
|
|
"reward_std": 0.07242131233215332,
|
|
"rewards/accuracy_reward_step": 0.234375,
|
|
"rewards/final_brier_reward_step": 0.7410029172897339,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 89
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8276430629193783,
|
|
"aux_distill/mean_u": 0.29844895285457557,
|
|
"aux_distill/n_active_tok": 487.0,
|
|
"calib/answer_extract_rate": 0.9609375,
|
|
"calib/auroc": 0.5064856711915535,
|
|
"calib/avg_num_step_conf": 16.74609375,
|
|
"calib/ece": 0.20158536585365852,
|
|
"calib/final_conf_rate": 0.9609375,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0009321266968325788,
|
|
"calib/mean_conf": 0.0057317073170731715,
|
|
"calib/mu_c": 0.006470588235294118,
|
|
"calib/mu_w": 0.005538461538461539,
|
|
"calib/nonempty_final_conf_rate": 0.9609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007916816690814732,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2181012658227848,
|
|
"calib/step_q_c_n": 553.0,
|
|
"calib/step_q_gap": 0.013298587729587141,
|
|
"calib/step_q_w": 0.20480267809319766,
|
|
"calib/step_q_w_n": 3734.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.03515625,
|
|
"completions/max_length": 2536.0,
|
|
"completions/max_terminated_length": 2536.0,
|
|
"completions/mean_length": 537.6484375,
|
|
"completions/mean_terminated_length": 557.2388916015625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 76.0,
|
|
"epoch": 0.096,
|
|
"grad_norm": 0.004481129348278046,
|
|
"learning_rate": 3.055555555555556e-06,
|
|
"loss": 0.0262,
|
|
"num_tokens": 19328909.0,
|
|
"reward": 0.9621806144714355,
|
|
"reward_std": 0.10468724370002747,
|
|
"rewards/accuracy_reward_step": 0.19921875,
|
|
"rewards/final_brier_reward_step": 0.7642050981521606,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 90
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.783676678314805,
|
|
"aux_distill/mean_u": 0.2469408792428226,
|
|
"aux_distill/n_active_tok": 464.875,
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/auroc": 0.5252176139272914,
|
|
"calib/avg_num_step_conf": 15.4453125,
|
|
"calib/ece": 0.2402788844621514,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 5.461682881037662e-05,
|
|
"calib/mean_conf": 0.0067330677290836655,
|
|
"calib/mu_c": 0.006774193548387097,
|
|
"calib/mu_w": 0.006719576719576721,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008014209377325996,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22713694267515924,
|
|
"calib/step_q_c_n": 785.0,
|
|
"calib/step_q_gap": -0.005854758176844527,
|
|
"calib/step_q_w": 0.23299170085200377,
|
|
"calib/step_q_w_n": 3169.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2617.0,
|
|
"completions/max_terminated_length": 2617.0,
|
|
"completions/mean_length": 543.29296875,
|
|
"completions/mean_terminated_length": 551.9166870117188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 120.0,
|
|
"epoch": 0.09706666666666666,
|
|
"grad_norm": 0.0064388904720544815,
|
|
"learning_rate": 3.0277777777777776e-06,
|
|
"loss": 0.1083,
|
|
"num_tokens": 19575704.0,
|
|
"reward": 0.9781495332717896,
|
|
"reward_std": 0.06891413033008575,
|
|
"rewards/accuracy_reward_step": 0.2421875,
|
|
"rewards/final_brier_reward_step": 0.7375491857528687,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 91
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7937708795070648,
|
|
"aux_distill/mean_u": 0.25537118199335607,
|
|
"aux_distill/n_active_tok": 453.75,
|
|
"calib/answer_extract_rate": 0.9609375,
|
|
"calib/auroc": 0.5101160862354893,
|
|
"calib/avg_num_step_conf": 16.390625,
|
|
"calib/ece": 0.17495934959349593,
|
|
"calib/final_conf_rate": 0.9609375,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.00656716417910448,
|
|
"calib/mean_conf": 0.007967479674796748,
|
|
"calib/mu_c": 0.013333333333333336,
|
|
"calib/mu_w": 0.006766169154228856,
|
|
"calib/nonempty_final_conf_rate": 0.9609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.020756401680472734,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.2312367088607595,
|
|
"calib/step_q_c_n": 474.0,
|
|
"calib/step_q_gap": 0.03461878838789548,
|
|
"calib/step_q_w": 0.19661792047286403,
|
|
"calib/step_q_w_n": 3722.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.03515625,
|
|
"completions/max_length": 2862.0,
|
|
"completions/max_terminated_length": 2862.0,
|
|
"completions/mean_length": 502.58203125,
|
|
"completions/mean_terminated_length": 520.894775390625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 77.0,
|
|
"epoch": 0.09813333333333334,
|
|
"grad_norm": 0.005982478614896536,
|
|
"learning_rate": 3e-06,
|
|
"loss": 0.0326,
|
|
"num_tokens": 19811085.0,
|
|
"reward": 0.9591376781463623,
|
|
"reward_std": 0.11132695525884628,
|
|
"rewards/accuracy_reward_step": 0.17578125,
|
|
"rewards/final_brier_reward_step": 0.7854628562927246,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 92
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8032534569501877,
|
|
"aux_distill/mean_u": 0.26527682188902096,
|
|
"aux_distill/n_active_tok": 496.875,
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/auroc": 0.5151528013582343,
|
|
"calib/avg_num_step_conf": 16.046875,
|
|
"calib/ece": 0.2391825396825397,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.000724278438030559,
|
|
"calib/mean_conf": 0.007642857142857143,
|
|
"calib/mu_c": 0.007096774193548388,
|
|
"calib/mu_w": 0.007821052631578947,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0003968253968253968,
|
|
"calib/std_conf": 0.010570221938858933,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24839263657957242,
|
|
"calib/step_q_c_n": 842.0,
|
|
"calib/step_q_gap": -0.004466824534940711,
|
|
"calib/step_q_w": 0.25285946111451313,
|
|
"calib/step_q_w_n": 3266.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2701.0,
|
|
"completions/max_terminated_length": 2701.0,
|
|
"completions/mean_length": 580.0625,
|
|
"completions/mean_terminated_length": 584.6299438476562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 74.0,
|
|
"epoch": 0.0992,
|
|
"grad_norm": 0.00656322855502367,
|
|
"learning_rate": 2.9722222222222225e-06,
|
|
"loss": 0.0868,
|
|
"num_tokens": 20065357.0,
|
|
"reward": 0.9761672019958496,
|
|
"reward_std": 0.06974552571773529,
|
|
"rewards/accuracy_reward_step": 0.2421875,
|
|
"rewards/final_brier_reward_step": 0.7374906539916992,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 93
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7575046606361866,
|
|
"aux_distill/mean_u": 0.23041620430958304,
|
|
"aux_distill/n_active_tok": 437.75,
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/auroc": 0.49822957839262183,
|
|
"calib/avg_num_step_conf": 14.8671875,
|
|
"calib/ece": 0.25604,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0003162055335968377,
|
|
"calib/mean_conf": 0.00796,
|
|
"calib/mu_c": 0.007727272727272728,
|
|
"calib/mu_w": 0.008043478260869565,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007812707597241817,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.28975,
|
|
"calib/step_q_c_n": 760.0,
|
|
"calib/step_q_gap": 0.060996979645436655,
|
|
"calib/step_q_w": 0.22875302035456335,
|
|
"calib/step_q_w_n": 3046.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2178.0,
|
|
"completions/max_terminated_length": 2178.0,
|
|
"completions/mean_length": 499.359375,
|
|
"completions/mean_terminated_length": 509.3067932128906,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 87.0,
|
|
"epoch": 0.10026666666666667,
|
|
"grad_norm": 0.005905228201299906,
|
|
"learning_rate": 2.944444444444445e-06,
|
|
"loss": 0.0575,
|
|
"num_tokens": 20301873.0,
|
|
"reward": 0.9745877981185913,
|
|
"reward_std": 0.08027815073728561,
|
|
"rewards/accuracy_reward_step": 0.2578125,
|
|
"rewards/final_brier_reward_step": 0.7187070250511169,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 94
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8056490551680326,
|
|
"aux_distill/mean_u": 0.3068893133746946,
|
|
"aux_distill/n_active_tok": 406.875,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.46144943730545135,
|
|
"calib/avg_num_step_conf": 13.46875,
|
|
"calib/ece": 0.25535433070866137,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.001104637241599489,
|
|
"calib/mean_conf": 0.008425196850393703,
|
|
"calib/mu_c": 0.007611940298507463,
|
|
"calib/mu_w": 0.008716577540106952,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.009257913988471854,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.26467741935483874,
|
|
"calib/step_q_c_n": 620.0,
|
|
"calib/step_q_gap": 0.011791492905050882,
|
|
"calib/step_q_w": 0.25288592644978786,
|
|
"calib/step_q_w_n": 2828.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2769.0,
|
|
"completions/max_terminated_length": 2769.0,
|
|
"completions/mean_length": 467.4765625,
|
|
"completions/mean_terminated_length": 471.157470703125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 71.0,
|
|
"epoch": 0.10133333333333333,
|
|
"grad_norm": 0.007395831868052483,
|
|
"learning_rate": 2.916666666666667e-06,
|
|
"loss": 0.0508,
|
|
"num_tokens": 20527675.0,
|
|
"reward": 0.9941019415855408,
|
|
"reward_std": 0.025779156014323235,
|
|
"rewards/accuracy_reward_step": 0.26171875,
|
|
"rewards/final_brier_reward_step": 0.7342976331710815,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 95
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.800974678248167,
|
|
"aux_distill/mean_u": 0.28526942786374343,
|
|
"aux_distill/n_active_tok": 428.125,
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/auroc": 0.5111559139784947,
|
|
"calib/avg_num_step_conf": 13.9921875,
|
|
"calib/ece": 0.3588537549407115,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.00012970430107526847,
|
|
"calib/mean_conf": 0.008735177865612648,
|
|
"calib/mu_c": 0.008817204301075268,
|
|
"calib/mu_w": 0.008687499999999999,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.00824411562022684,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2628656374501992,
|
|
"calib/step_q_c_n": 1004.0,
|
|
"calib/step_q_gap": 0.021554543578981233,
|
|
"calib/step_q_w": 0.24131109387121796,
|
|
"calib/step_q_w_n": 2578.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2726.0,
|
|
"completions/max_terminated_length": 2726.0,
|
|
"completions/mean_length": 481.1796875,
|
|
"completions/mean_terminated_length": 486.8854064941406,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 82.0,
|
|
"epoch": 0.1024,
|
|
"grad_norm": 0.005970444995909929,
|
|
"learning_rate": 2.888888888888889e-06,
|
|
"loss": 0.0611,
|
|
"num_tokens": 20756673.0,
|
|
"reward": 0.9836015105247498,
|
|
"reward_std": 0.0583636611700058,
|
|
"rewards/accuracy_reward_step": 0.36328125,
|
|
"rewards/final_brier_reward_step": 0.6234530806541443,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 96
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8248364944010973,
|
|
"aux_distill/mean_u": 0.27616341024908203,
|
|
"aux_distill/n_active_tok": 411.0,
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/auroc": 0.48250511818351016,
|
|
"calib/avg_num_step_conf": 13.3828125,
|
|
"calib/ece": 0.20478260869565215,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0011166945840312666,
|
|
"calib/mean_conf": 0.00865612648221344,
|
|
"calib/mu_c": 0.007777777777777778,
|
|
"calib/mu_w": 0.008894472361809045,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.010357843069824,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24534420289855075,
|
|
"calib/step_q_c_n": 552.0,
|
|
"calib/step_q_gap": -0.014249499258721376,
|
|
"calib/step_q_w": 0.2595937021572721,
|
|
"calib/step_q_w_n": 2874.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1860.0,
|
|
"completions/max_terminated_length": 1860.0,
|
|
"completions/mean_length": 468.1328125,
|
|
"completions/mean_terminated_length": 471.81890869140625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 111.0,
|
|
"epoch": 0.10346666666666667,
|
|
"grad_norm": 0.005742794368416071,
|
|
"learning_rate": 2.861111111111111e-06,
|
|
"loss": 0.067,
|
|
"num_tokens": 20981587.0,
|
|
"reward": 0.985925555229187,
|
|
"reward_std": 0.04699475318193436,
|
|
"rewards/accuracy_reward_step": 0.2109375,
|
|
"rewards/final_brier_reward_step": 0.776538610458374,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 97
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8426261711865664,
|
|
"aux_distill/mean_u": 0.3216190219877432,
|
|
"aux_distill/n_active_tok": 404.5,
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/auroc": 0.5656523357528382,
|
|
"calib/avg_num_step_conf": 13.57421875,
|
|
"calib/ece": 0.2075494071146245,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.001636888144425832,
|
|
"calib/mean_conf": 0.012213438735177865,
|
|
"calib/mu_c": 0.010925925925925927,
|
|
"calib/mu_w": 0.01256281407035176,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0031620553359683794,
|
|
"calib/std_conf": 0.0381053498214233,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2518213058419244,
|
|
"calib/step_q_c_n": 582.0,
|
|
"calib/step_q_gap": -0.015493094572869903,
|
|
"calib/step_q_w": 0.2673144004147943,
|
|
"calib/step_q_w_n": 2893.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1944.0,
|
|
"completions/max_terminated_length": 1944.0,
|
|
"completions/mean_length": 462.1640625,
|
|
"completions/mean_terminated_length": 467.644287109375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 45.0,
|
|
"epoch": 0.10453333333333334,
|
|
"grad_norm": 0.0061998069286346436,
|
|
"learning_rate": 2.8333333333333335e-06,
|
|
"loss": 0.0432,
|
|
"num_tokens": 21206085.0,
|
|
"reward": 0.9897946715354919,
|
|
"reward_std": 0.03887510299682617,
|
|
"rewards/accuracy_reward_step": 0.2109375,
|
|
"rewards/final_brier_reward_step": 0.7803707122802734,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 98
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7924968916922808,
|
|
"aux_distill/mean_u": 0.26154136502969894,
|
|
"aux_distill/n_active_tok": 433.0,
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/auroc": 0.5699200913242009,
|
|
"calib/avg_num_step_conf": 13.8828125,
|
|
"calib/ece": 0.11856573705179282,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0008319063926940633,
|
|
"calib/mean_conf": 0.010836653386454183,
|
|
"calib/mu_c": 0.0115625,
|
|
"calib/mu_w": 0.010730593607305937,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0009561752988047808,
|
|
"calib/std_conf": 0.01761974431133978,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2841832669322709,
|
|
"calib/step_q_c_n": 251.0,
|
|
"calib/step_q_gap": -0.01624967281946993,
|
|
"calib/step_q_w": 0.30043293975174085,
|
|
"calib/step_q_w_n": 3303.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2834.0,
|
|
"completions/max_terminated_length": 2834.0,
|
|
"completions/mean_length": 512.203125,
|
|
"completions/mean_terminated_length": 514.2117919921875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 61.0,
|
|
"epoch": 0.1056,
|
|
"grad_norm": 0.005434286780655384,
|
|
"learning_rate": 2.805555555555556e-06,
|
|
"loss": 0.1352,
|
|
"num_tokens": 21443009.0,
|
|
"reward": 0.9817042350769043,
|
|
"reward_std": 0.057376354932785034,
|
|
"rewards/accuracy_reward_step": 0.125,
|
|
"rewards/final_brier_reward_step": 0.8579398393630981,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 99
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8403134858235717,
|
|
"aux_distill/mean_u": 0.304650107976595,
|
|
"aux_distill/n_active_tok": 362.75,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.4799983638743455,
|
|
"calib/avg_num_step_conf": 11.4609375,
|
|
"calib/ece": 0.24011764705882355,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.00105693717277487,
|
|
"calib/mean_conf": 0.012666666666666668,
|
|
"calib/mu_c": 0.011875,
|
|
"calib/mu_w": 0.01293193717277487,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0009019607843137256,
|
|
"calib/std_conf": 0.023194545223059995,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2982043343653251,
|
|
"calib/step_q_c_n": 646.0,
|
|
"calib/step_q_gap": 0.013900397357451133,
|
|
"calib/step_q_w": 0.284303937007874,
|
|
"calib/step_q_w_n": 2286.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1673.0,
|
|
"completions/max_terminated_length": 1673.0,
|
|
"completions/mean_length": 432.72265625,
|
|
"completions/mean_terminated_length": 434.4196472167969,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 59.0,
|
|
"epoch": 0.10666666666666667,
|
|
"grad_norm": 0.007272300310432911,
|
|
"learning_rate": 2.7777777777777783e-06,
|
|
"loss": 0.0793,
|
|
"num_tokens": 21661194.0,
|
|
"reward": 0.9948787689208984,
|
|
"reward_std": 0.020361021161079407,
|
|
"rewards/accuracy_reward_step": 0.25,
|
|
"rewards/final_brier_reward_step": 0.747570276260376,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 100
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.820907037705183,
|
|
"aux_distill/mean_u": 0.2527867314052802,
|
|
"aux_distill/n_active_tok": 374.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.4725410341848698,
|
|
"calib/avg_num_step_conf": 11.85546875,
|
|
"calib/ece": 0.1340234375,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0012254720473898542,
|
|
"calib/mean_conf": 0.010507812500000002,
|
|
"calib/mu_c": 0.00945945945945946,
|
|
"calib/mu_w": 0.010684931506849314,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.011221140827244071,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.2470157068062827,
|
|
"calib/step_q_c_n": 382.0,
|
|
"calib/step_q_gap": -0.00756171498037389,
|
|
"calib/step_q_w": 0.2545774217866566,
|
|
"calib/step_q_w_n": 2653.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1342.0,
|
|
"completions/max_terminated_length": 1342.0,
|
|
"completions/mean_length": 440.0703125,
|
|
"completions/mean_terminated_length": 441.7961120605469,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 74.0,
|
|
"epoch": 0.10773333333333333,
|
|
"grad_norm": 0.0064960652962327,
|
|
"learning_rate": 2.7500000000000004e-06,
|
|
"loss": 0.0664,
|
|
"num_tokens": 21880844.0,
|
|
"reward": 0.9973434805870056,
|
|
"reward_std": 0.013668889179825783,
|
|
"rewards/accuracy_reward_step": 0.14453125,
|
|
"rewards/final_brier_reward_step": 0.8540620803833008,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 101
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7962197810411453,
|
|
"aux_distill/mean_u": 0.2730796349751205,
|
|
"aux_distill/n_active_tok": 331.125,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.5503753753753755,
|
|
"calib/avg_num_step_conf": 10.765625,
|
|
"calib/ece": 0.28129921259842516,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0004399399399399387,
|
|
"calib/mean_conf": 0.011850393700787403,
|
|
"calib/mu_c": 0.012162162162162163,
|
|
"calib/mu_w": 0.011722222222222224,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0009055118110236221,
|
|
"calib/std_conf": 0.017838990623327684,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2971952554744526,
|
|
"calib/step_q_c_n": 548.0,
|
|
"calib/step_q_gap": -0.0031182861922141036,
|
|
"calib/step_q_w": 0.3003135416666667,
|
|
"calib/step_q_w_n": 2208.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2663.0,
|
|
"completions/max_terminated_length": 2663.0,
|
|
"completions/mean_length": 389.7109375,
|
|
"completions/mean_terminated_length": 392.779541015625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 43.0,
|
|
"epoch": 0.1088,
|
|
"grad_norm": 0.006796353962272406,
|
|
"learning_rate": 2.7222222222222224e-06,
|
|
"loss": 0.062,
|
|
"num_tokens": 22087306.0,
|
|
"reward": 0.9954755902290344,
|
|
"reward_std": 0.027535689994692802,
|
|
"rewards/accuracy_reward_step": 0.2890625,
|
|
"rewards/final_brier_reward_step": 0.7097011804580688,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 102
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8647203091531992,
|
|
"aux_distill/mean_u": 0.29944187739858286,
|
|
"aux_distill/n_active_tok": 331.25,
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/auroc": 0.527927927927928,
|
|
"calib/avg_num_step_conf": 10.58984375,
|
|
"calib/ece": 0.28141732283464566,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0006831831831831826,
|
|
"calib/mean_conf": 0.009921259842519688,
|
|
"calib/mu_c": 0.010405405405405405,
|
|
"calib/mu_w": 0.009722222222222222,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.009306340484634464,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24324257274119448,
|
|
"calib/step_q_c_n": 653.0,
|
|
"calib/step_q_gap": 0.026392718513789243,
|
|
"calib/step_q_w": 0.21684985422740524,
|
|
"calib/step_q_w_n": 2058.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1834.0,
|
|
"completions/max_terminated_length": 1834.0,
|
|
"completions/mean_length": 404.06640625,
|
|
"completions/mean_terminated_length": 405.6510009765625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 46.0,
|
|
"epoch": 0.10986666666666667,
|
|
"grad_norm": 0.00689524132758379,
|
|
"learning_rate": 2.6944444444444444e-06,
|
|
"loss": 0.062,
|
|
"num_tokens": 22295299.0,
|
|
"reward": 0.9911972284317017,
|
|
"reward_std": 0.037496358156204224,
|
|
"rewards/accuracy_reward_step": 0.2890625,
|
|
"rewards/final_brier_reward_step": 0.7050508260726929,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 103
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8144241627305746,
|
|
"aux_distill/mean_u": 0.2616677400131344,
|
|
"aux_distill/n_active_tok": 343.0,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.5434120335110435,
|
|
"calib/avg_num_step_conf": 10.99609375,
|
|
"calib/ece": 0.19433070866141733,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0009558263518659564,
|
|
"calib/mean_conf": 0.010393700787401573,
|
|
"calib/mu_c": 0.011153846153846155,
|
|
"calib/mu_w": 0.010198019801980198,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008685364176078153,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.275157399103139,
|
|
"calib/step_q_c_n": 446.0,
|
|
"calib/step_q_gap": 0.06505085625805673,
|
|
"calib/step_q_w": 0.2101065428450823,
|
|
"calib/step_q_w_n": 2369.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2965.0,
|
|
"completions/max_terminated_length": 2965.0,
|
|
"completions/mean_length": 419.90234375,
|
|
"completions/mean_terminated_length": 421.5490417480469,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 74.0,
|
|
"epoch": 0.11093333333333333,
|
|
"grad_norm": 0.006455368362367153,
|
|
"learning_rate": 2.666666666666667e-06,
|
|
"loss": 0.1269,
|
|
"num_tokens": 22509474.0,
|
|
"reward": 0.9943621158599854,
|
|
"reward_std": 0.025546476244926453,
|
|
"rewards/accuracy_reward_step": 0.203125,
|
|
"rewards/final_brier_reward_step": 0.7934117317199707,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 104
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8212667172774673,
|
|
"aux_distill/mean_u": 0.2547904974645424,
|
|
"aux_distill/n_active_tok": 321.875,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5416195324283559,
|
|
"calib/avg_num_step_conf": 10.12109375,
|
|
"calib/ece": 0.197734375,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.00390625,
|
|
"calib/gap": -0.003054298642533936,
|
|
"calib/mean_conf": 0.013203125000000001,
|
|
"calib/mu_c": 0.01076923076923077,
|
|
"calib/mu_w": 0.013823529411764707,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.00390625,
|
|
"calib/std_conf": 0.06247416458212447,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.25754342431761784,
|
|
"calib/step_q_c_n": 403.0,
|
|
"calib/step_q_gap": 0.007443195798422231,
|
|
"calib/step_q_w": 0.2501002285191956,
|
|
"calib/step_q_w_n": 2188.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1128.0,
|
|
"completions/max_terminated_length": 1128.0,
|
|
"completions/mean_length": 380.5625,
|
|
"completions/mean_terminated_length": 382.054931640625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 87.0,
|
|
"epoch": 0.112,
|
|
"grad_norm": 0.00898874830454588,
|
|
"learning_rate": 2.6388888888888893e-06,
|
|
"loss": 0.0861,
|
|
"num_tokens": 22712658.0,
|
|
"reward": 1.0001487731933594,
|
|
"reward_std": 0.009183384478092194,
|
|
"rewards/accuracy_reward_step": 0.203125,
|
|
"rewards/final_brier_reward_step": 0.7971726059913635,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 105
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8393920790404081,
|
|
"aux_distill/mean_u": 0.2885217277018143,
|
|
"aux_distill/n_active_tok": 308.625,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5007281553398057,
|
|
"calib/avg_num_step_conf": 9.7421875,
|
|
"calib/ece": 0.1848828125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -3.689320388349307e-05,
|
|
"calib/mean_conf": 0.0104296875,
|
|
"calib/mu_c": 0.010400000000000003,
|
|
"calib/mu_w": 0.010436893203883496,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.009239168991437691,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2901785714285714,
|
|
"calib/step_q_c_n": 392.0,
|
|
"calib/step_q_gap": 0.009724099497077587,
|
|
"calib/step_q_w": 0.2804544719314938,
|
|
"calib/step_q_w_n": 2102.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1085.0,
|
|
"completions/max_terminated_length": 1085.0,
|
|
"completions/mean_length": 375.60546875,
|
|
"completions/mean_terminated_length": 377.0784606933594,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 86.0,
|
|
"epoch": 0.11306666666666666,
|
|
"grad_norm": 0.00921398214995861,
|
|
"learning_rate": 2.6111111111111113e-06,
|
|
"loss": 0.105,
|
|
"num_tokens": 22913397.0,
|
|
"reward": 1.0019341707229614,
|
|
"reward_std": 0.00327928620390594,
|
|
"rewards/accuracy_reward_step": 0.1953125,
|
|
"rewards/final_brier_reward_step": 0.8085559010505676,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 106
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8549458589404821,
|
|
"aux_distill/mean_u": 0.2930654655301045,
|
|
"aux_distill/n_active_tok": 320.375,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5131324442634633,
|
|
"calib/avg_num_step_conf": 10.0703125,
|
|
"calib/ece": 0.17445312500000001,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0002606128473989621,
|
|
"calib/mean_conf": 0.01,
|
|
"calib/mu_c": 0.01021276595744681,
|
|
"calib/mu_w": 0.009952153110047848,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0004296875,
|
|
"calib/std_conf": 0.010307764064044152,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.31333333333333335,
|
|
"calib/step_q_c_n": 453.0,
|
|
"calib/step_q_gap": 0.06561130980392157,
|
|
"calib/step_q_w": 0.24772202352941178,
|
|
"calib/step_q_w_n": 2125.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1170.0,
|
|
"completions/max_terminated_length": 1170.0,
|
|
"completions/mean_length": 376.2109375,
|
|
"completions/mean_terminated_length": 377.6863098144531,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 69.0,
|
|
"epoch": 0.11413333333333334,
|
|
"grad_norm": 0.01009163074195385,
|
|
"learning_rate": 2.5833333333333337e-06,
|
|
"loss": 0.0553,
|
|
"num_tokens": 23114323.0,
|
|
"reward": 1.0017719268798828,
|
|
"reward_std": 0.0037272428162395954,
|
|
"rewards/accuracy_reward_step": 0.18359375,
|
|
"rewards/final_brier_reward_step": 0.8199499845504761,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 107
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.826457666233182,
|
|
"aux_distill/mean_u": 0.25843460005882996,
|
|
"aux_distill/n_active_tok": 287.0,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5170957918050941,
|
|
"calib/avg_num_step_conf": 9.03125,
|
|
"calib/ece": 0.3167578125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 9.136212624584819e-05,
|
|
"calib/mean_conf": 0.0113671875,
|
|
"calib/mu_c": 0.01142857142857143,
|
|
"calib/mu_w": 0.011337209302325582,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.009646381359859445,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.30450741839762613,
|
|
"calib/step_q_c_n": 674.0,
|
|
"calib/step_q_gap": 0.0063376992279069455,
|
|
"calib/step_q_w": 0.2981697191697192,
|
|
"calib/step_q_w_n": 1638.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 851.0,
|
|
"completions/max_terminated_length": 851.0,
|
|
"completions/mean_length": 349.85546875,
|
|
"completions/mean_terminated_length": 351.22747802734375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 81.0,
|
|
"epoch": 0.1152,
|
|
"grad_norm": 0.008953594602644444,
|
|
"learning_rate": 2.5555555555555557e-06,
|
|
"loss": 0.1039,
|
|
"num_tokens": 23307118.0,
|
|
"reward": 1.003638744354248,
|
|
"reward_std": 0.005494957789778709,
|
|
"rewards/accuracy_reward_step": 0.328125,
|
|
"rewards/final_brier_reward_step": 0.6791527271270752,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 108
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8436379954218864,
|
|
"aux_distill/mean_u": 0.25006158266790723,
|
|
"aux_distill/n_active_tok": 290.125,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.501933585540143,
|
|
"calib/avg_num_step_conf": 9.08984375,
|
|
"calib/ece": 0.2273046875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 9.247583018078681e-06,
|
|
"calib/mean_conf": 0.0109765625,
|
|
"calib/mu_c": 0.01098360655737705,
|
|
"calib/mu_w": 0.010974358974358972,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.009282009517534107,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.28361872146118716,
|
|
"calib/step_q_c_n": 438.0,
|
|
"calib/step_q_gap": 0.008033491180615415,
|
|
"calib/step_q_w": 0.27558523028057175,
|
|
"calib/step_q_w_n": 1889.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 962.0,
|
|
"completions/max_terminated_length": 962.0,
|
|
"completions/mean_length": 350.03515625,
|
|
"completions/mean_terminated_length": 351.4078674316406,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 88.0,
|
|
"epoch": 0.11626666666666667,
|
|
"grad_norm": 0.00752226123586297,
|
|
"learning_rate": 2.5277777777777778e-06,
|
|
"loss": 0.1031,
|
|
"num_tokens": 23501327.0,
|
|
"reward": 1.0025138854980469,
|
|
"reward_std": 0.003788697300478816,
|
|
"rewards/accuracy_reward_step": 0.23828125,
|
|
"rewards/final_brier_reward_step": 0.766746461391449,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 109
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8580736331641674,
|
|
"aux_distill/mean_u": 0.24811984237804047,
|
|
"aux_distill/n_active_tok": 255.375,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.48946814112690895,
|
|
"calib/avg_num_step_conf": 8.02734375,
|
|
"calib/ece": 0.1649609375,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.00045602948920484346,
|
|
"calib/mean_conf": 0.010820312500000002,
|
|
"calib/mu_c": 0.010444444444444445,
|
|
"calib/mu_w": 0.010900473933649289,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008689013891250475,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2304778156996587,
|
|
"calib/step_q_c_n": 293.0,
|
|
"calib/step_q_gap": 0.00432571581316607,
|
|
"calib/step_q_w": 0.22615209988649262,
|
|
"calib/step_q_w_n": 1762.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1089.0,
|
|
"completions/max_terminated_length": 1089.0,
|
|
"completions/mean_length": 326.859375,
|
|
"completions/mean_terminated_length": 328.1412048339844,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 49.0,
|
|
"epoch": 0.11733333333333333,
|
|
"grad_norm": 0.00968889519572258,
|
|
"learning_rate": 2.5e-06,
|
|
"loss": 0.079,
|
|
"num_tokens": 23689923.0,
|
|
"reward": 1.0017396211624146,
|
|
"reward_std": 0.0027383025735616684,
|
|
"rewards/accuracy_reward_step": 0.17578125,
|
|
"rewards/final_brier_reward_step": 0.8276980519294739,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 110
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8748779203742743,
|
|
"aux_distill/mean_u": 0.2795663521799445,
|
|
"aux_distill/n_active_tok": 249.125,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5199227799227799,
|
|
"calib/avg_num_step_conf": 8.17578125,
|
|
"calib/ece": 0.263843137254902,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0008262548262548269,
|
|
"calib/mean_conf": 0.01231372549019608,
|
|
"calib/mu_c": 0.011714285714285715,
|
|
"calib/mu_w": 0.012540540540540542,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0008235294117647058,
|
|
"calib/std_conf": 0.014916092370688742,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.33807439824945296,
|
|
"calib/step_q_c_n": 457.0,
|
|
"calib/step_q_gap": 0.03397293125678802,
|
|
"calib/step_q_w": 0.30410146699266494,
|
|
"calib/step_q_w_n": 1636.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2087.0,
|
|
"completions/max_terminated_length": 2087.0,
|
|
"completions/mean_length": 326.27734375,
|
|
"completions/mean_terminated_length": 327.556884765625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 66.0,
|
|
"epoch": 0.1184,
|
|
"grad_norm": 0.00978804100304842,
|
|
"learning_rate": 2.4722222222222226e-06,
|
|
"loss": 0.0978,
|
|
"num_tokens": 23880858.0,
|
|
"reward": 0.9991105198860168,
|
|
"reward_std": 0.015512233600020409,
|
|
"rewards/accuracy_reward_step": 0.2734375,
|
|
"rewards/final_brier_reward_step": 0.7286898493766785,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 111
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8446307703852654,
|
|
"aux_distill/mean_u": 0.278147945651484,
|
|
"aux_distill/n_active_tok": 252.5,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.4851499423298732,
|
|
"calib/avg_num_step_conf": 7.890625,
|
|
"calib/ece": 0.19054901960784315,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.001470588235294116,
|
|
"calib/mean_conf": 0.011568627450980393,
|
|
"calib/mu_c": 0.010392156862745099,
|
|
"calib/mu_w": 0.011862745098039215,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0010588235294117648,
|
|
"calib/std_conf": 0.01792933285282756,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.28216374269005845,
|
|
"calib/step_q_c_n": 342.0,
|
|
"calib/step_q_gap": 0.03338179990102391,
|
|
"calib/step_q_w": 0.24878194278903454,
|
|
"calib/step_q_w_n": 1678.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 1947.0,
|
|
"completions/max_terminated_length": 1947.0,
|
|
"completions/mean_length": 327.9296875,
|
|
"completions/mean_terminated_length": 327.9296875,
|
|
"completions/min_length": 65.0,
|
|
"completions/min_terminated_length": 65.0,
|
|
"epoch": 0.11946666666666667,
|
|
"grad_norm": 0.009700233116745949,
|
|
"learning_rate": 2.4444444444444447e-06,
|
|
"loss": 0.1246,
|
|
"num_tokens": 24072728.0,
|
|
"reward": 0.9979373216629028,
|
|
"reward_std": 0.014979146420955658,
|
|
"rewards/accuracy_reward_step": 0.19921875,
|
|
"rewards/final_brier_reward_step": 0.8005621433258057,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 112
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8938411138951778,
|
|
"aux_distill/mean_u": 0.27519478460141256,
|
|
"aux_distill/n_active_tok": 228.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.4992364857986359,
|
|
"calib/avg_num_step_conf": 7.3125,
|
|
"calib/ece": 0.1726953125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -5.8027079303675164e-05,
|
|
"calib/mean_conf": 0.0108984375,
|
|
"calib/mu_c": 0.010851063829787235,
|
|
"calib/mu_w": 0.01090909090909091,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007928725311082088,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24974921630094044,
|
|
"calib/step_q_c_n": 319.0,
|
|
"calib/step_q_gap": -0.016225413447932657,
|
|
"calib/step_q_w": 0.2659746297488731,
|
|
"calib/step_q_w_n": 1553.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 926.0,
|
|
"completions/max_terminated_length": 926.0,
|
|
"completions/mean_length": 276.69921875,
|
|
"completions/mean_terminated_length": 277.7843322753906,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 84.0,
|
|
"epoch": 0.12053333333333334,
|
|
"grad_norm": 0.011297706514596939,
|
|
"learning_rate": 2.4166666666666667e-06,
|
|
"loss": 0.0632,
|
|
"num_tokens": 24248763.0,
|
|
"reward": 1.0019012689590454,
|
|
"reward_std": 0.003506625071167946,
|
|
"rewards/accuracy_reward_step": 0.18359375,
|
|
"rewards/final_brier_reward_step": 0.8202089667320251,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 113
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8577989656478167,
|
|
"aux_distill/mean_u": 0.24188379755471998,
|
|
"aux_distill/n_active_tok": 210.75,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.534063701661467,
|
|
"calib/avg_num_step_conf": 6.75390625,
|
|
"calib/ece": 0.2883203125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.00019516796053108826,
|
|
"calib/mean_conf": 0.012460937500000002,
|
|
"calib/mu_c": 0.012597402597402598,
|
|
"calib/mu_w": 0.01240223463687151,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008088767620663469,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.2853211009174312,
|
|
"calib/step_q_c_n": 436.0,
|
|
"calib/step_q_gap": 0.02676093077048608,
|
|
"calib/step_q_w": 0.2585601701469451,
|
|
"calib/step_q_w_n": 1293.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1168.0,
|
|
"completions/max_terminated_length": 1168.0,
|
|
"completions/mean_length": 273.59765625,
|
|
"completions/mean_terminated_length": 274.67059326171875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 55.0,
|
|
"epoch": 0.1216,
|
|
"grad_norm": 0.010183519683778286,
|
|
"learning_rate": 2.388888888888889e-06,
|
|
"loss": 0.0621,
|
|
"num_tokens": 24423828.0,
|
|
"reward": 1.0017255544662476,
|
|
"reward_std": 0.010000040754675865,
|
|
"rewards/accuracy_reward_step": 0.30078125,
|
|
"rewards/final_brier_reward_step": 0.7065761685371399,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 114
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8538937643170357,
|
|
"aux_distill/mean_u": 0.22364562285076225,
|
|
"aux_distill/n_active_tok": 222.0,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5282455603184323,
|
|
"calib/avg_num_step_conf": 7.2578125,
|
|
"calib/ece": 0.2669019607843137,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0015890998162890382,
|
|
"calib/mean_conf": 0.011529411764705884,
|
|
"calib/mu_c": 0.01267605633802817,
|
|
"calib/mu_w": 0.011086956521739131,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008044856939895886,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.3118958333333334,
|
|
"calib/step_q_c_n": 480.0,
|
|
"calib/step_q_gap": 0.02730221940009686,
|
|
"calib/step_q_w": 0.28459361393323657,
|
|
"calib/step_q_w_n": 1378.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2678.0,
|
|
"completions/max_terminated_length": 2678.0,
|
|
"completions/mean_length": 286.85546875,
|
|
"completions/mean_terminated_length": 287.98040771484375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 50.0,
|
|
"epoch": 0.12266666666666666,
|
|
"grad_norm": 0.010189507156610489,
|
|
"learning_rate": 2.361111111111111e-06,
|
|
"loss": 0.0982,
|
|
"num_tokens": 24602527.0,
|
|
"reward": 0.9995108842849731,
|
|
"reward_std": 0.016104616224765778,
|
|
"rewards/accuracy_reward_step": 0.27734375,
|
|
"rewards/final_brier_reward_step": 0.7255843877792358,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 115
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8547423658892512,
|
|
"aux_distill/mean_u": 0.24822352620147983,
|
|
"aux_distill/n_active_tok": 233.125,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5182397959183673,
|
|
"calib/avg_num_step_conf": 7.3671875,
|
|
"calib/ece": 0.22429687499999998,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0011020408163265275,
|
|
"calib/mean_conf": 0.01234375,
|
|
"calib/mu_c": 0.011500000000000002,
|
|
"calib/mu_w": 0.012602040816326529,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0011328125,
|
|
"calib/std_conf": 0.019303674156426802,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24276649746192894,
|
|
"calib/step_q_c_n": 394.0,
|
|
"calib/step_q_gap": 0.007699473333242585,
|
|
"calib/step_q_w": 0.23506702412868635,
|
|
"calib/step_q_w_n": 1492.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1080.0,
|
|
"completions/max_terminated_length": 1080.0,
|
|
"completions/mean_length": 301.54296875,
|
|
"completions/mean_terminated_length": 302.7254943847656,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 49.0,
|
|
"epoch": 0.12373333333333333,
|
|
"grad_norm": 0.009150901809334755,
|
|
"learning_rate": 2.3333333333333336e-06,
|
|
"loss": 0.1132,
|
|
"num_tokens": 24784242.0,
|
|
"reward": 1.0024328231811523,
|
|
"reward_std": 0.0038933088071644306,
|
|
"rewards/accuracy_reward_step": 0.234375,
|
|
"rewards/final_brier_reward_step": 0.7704906463623047,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 116
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8299000021070242,
|
|
"aux_distill/mean_u": 0.21773052726680453,
|
|
"aux_distill/n_active_tok": 250.875,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.4671406003159558,
|
|
"calib/avg_num_step_conf": 7.89453125,
|
|
"calib/ece": 0.1650390625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0011700895208004176,
|
|
"calib/mean_conf": 0.0107421875,
|
|
"calib/mu_c": 0.00977777777777778,
|
|
"calib/mu_w": 0.010947867298578197,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007436970667875715,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.26386627906976745,
|
|
"calib/step_q_c_n": 344.0,
|
|
"calib/step_q_gap": 0.023079159212880146,
|
|
"calib/step_q_w": 0.2407871198568873,
|
|
"calib/step_q_w_n": 1677.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 991.0,
|
|
"completions/max_terminated_length": 991.0,
|
|
"completions/mean_length": 310.97265625,
|
|
"completions/mean_terminated_length": 312.1921691894531,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 69.0,
|
|
"epoch": 0.1248,
|
|
"grad_norm": 0.008661650121212006,
|
|
"learning_rate": 2.305555555555556e-06,
|
|
"loss": 0.1199,
|
|
"num_tokens": 24970451.0,
|
|
"reward": 0.9977273344993591,
|
|
"reward_std": 0.013147260062396526,
|
|
"rewards/accuracy_reward_step": 0.17578125,
|
|
"rewards/final_brier_reward_step": 0.8235796689987183,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 117
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8391965217888355,
|
|
"aux_distill/mean_u": 0.1979998792369977,
|
|
"aux_distill/n_active_tok": 211.0,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5883572567783094,
|
|
"calib/avg_num_step_conf": 6.74609375,
|
|
"calib/ece": 0.24605468749999998,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.004776714513556619,
|
|
"calib/mean_conf": 0.011757812500000001,
|
|
"calib/mu_c": 0.015303030303030303,
|
|
"calib/mu_w": 0.010526315789473684,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.012264816354713338,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.20689156626506025,
|
|
"calib/step_q_c_n": 415.0,
|
|
"calib/step_q_gap": -0.04102001910079339,
|
|
"calib/step_q_w": 0.24791158536585364,
|
|
"calib/step_q_w_n": 1312.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 887.0,
|
|
"completions/max_terminated_length": 887.0,
|
|
"completions/mean_length": 271.43359375,
|
|
"completions/mean_terminated_length": 272.498046875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 49.0,
|
|
"epoch": 0.12586666666666665,
|
|
"grad_norm": 0.011321449652314186,
|
|
"learning_rate": 2.277777777777778e-06,
|
|
"loss": 0.0597,
|
|
"num_tokens": 25143946.0,
|
|
"reward": 1.003800868988037,
|
|
"reward_std": 0.005998665001243353,
|
|
"rewards/accuracy_reward_step": 0.2578125,
|
|
"rewards/final_brier_reward_step": 0.7497894763946533,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 118
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.9129964718595147,
|
|
"aux_distill/mean_u": 0.28269734278456216,
|
|
"aux_distill/n_active_tok": 216.25,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5326227636086792,
|
|
"calib/avg_num_step_conf": 6.8359375,
|
|
"calib/ece": 0.2666015625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0011168633422154566,
|
|
"calib/mean_conf": 0.0107421875,
|
|
"calib/mu_c": 0.011549295774647889,
|
|
"calib/mu_w": 0.010432432432432432,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.0065428229927794734,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.25570806100217863,
|
|
"calib/step_q_c_n": 459.0,
|
|
"calib/step_q_gap": 0.028099230638119782,
|
|
"calib/step_q_w": 0.22760883036405885,
|
|
"calib/step_q_w_n": 1291.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 850.0,
|
|
"completions/max_terminated_length": 850.0,
|
|
"completions/mean_length": 276.4296875,
|
|
"completions/mean_terminated_length": 277.51373291015625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 48.0,
|
|
"epoch": 0.12693333333333334,
|
|
"grad_norm": 0.012318076565861702,
|
|
"learning_rate": 2.25e-06,
|
|
"loss": 0.0955,
|
|
"num_tokens": 25319776.0,
|
|
"reward": 0.9992179274559021,
|
|
"reward_std": 0.015154972672462463,
|
|
"rewards/accuracy_reward_step": 0.27734375,
|
|
"rewards/final_brier_reward_step": 0.7249984741210938,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 119
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8805275009945035,
|
|
"aux_distill/mean_u": 0.2454018055368668,
|
|
"aux_distill/n_active_tok": 206.875,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5360638210379143,
|
|
"calib/avg_num_step_conf": 6.48046875,
|
|
"calib/ece": 0.234609375,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0013997861666255433,
|
|
"calib/mean_conf": 0.011484375,
|
|
"calib/mu_c": 0.01253968253968254,
|
|
"calib/mu_w": 0.011139896373056997,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008205356839246846,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22673239436619716,
|
|
"calib/step_q_c_n": 355.0,
|
|
"calib/step_q_gap": 0.002174879028773863,
|
|
"calib/step_q_w": 0.2245575153374233,
|
|
"calib/step_q_w_n": 1304.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1198.0,
|
|
"completions/max_terminated_length": 1198.0,
|
|
"completions/mean_length": 265.234375,
|
|
"completions/mean_terminated_length": 266.2745361328125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 60.0,
|
|
"epoch": 0.128,
|
|
"grad_norm": 0.010012347251176834,
|
|
"learning_rate": 2.222222222222222e-06,
|
|
"loss": 0.0672,
|
|
"num_tokens": 25494364.0,
|
|
"reward": 1.0029863119125366,
|
|
"reward_std": 0.003938643261790276,
|
|
"rewards/accuracy_reward_step": 0.24609375,
|
|
"rewards/final_brier_reward_step": 0.759878933429718,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 120
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8603787571191788,
|
|
"aux_distill/mean_u": 0.23333242359569997,
|
|
"aux_distill/n_active_tok": 240.0,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.514964216005205,
|
|
"calib/avg_num_step_conf": 7.625,
|
|
"calib/ece": 0.1973046875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0003448275862068962,
|
|
"calib/mean_conf": 0.0097265625,
|
|
"calib/mu_c": 0.01,
|
|
"calib/mu_w": 0.009655172413793104,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007572787263194032,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.17681052631578947,
|
|
"calib/step_q_c_n": 342.0,
|
|
"calib/step_q_gap": -0.02709630598234719,
|
|
"calib/step_q_w": 0.20390683229813666,
|
|
"calib/step_q_w_n": 1610.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 897.0,
|
|
"completions/max_terminated_length": 897.0,
|
|
"completions/mean_length": 294.94140625,
|
|
"completions/mean_terminated_length": 296.0980529785156,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 63.0,
|
|
"epoch": 0.12906666666666666,
|
|
"grad_norm": 0.010293123312294483,
|
|
"learning_rate": 2.1944444444444445e-06,
|
|
"loss": 0.0489,
|
|
"num_tokens": 25674925.0,
|
|
"reward": 1.0019943714141846,
|
|
"reward_std": 0.0032473872415721416,
|
|
"rewards/accuracy_reward_step": 0.20703125,
|
|
"rewards/final_brier_reward_step": 0.7969574332237244,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 121
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.908028569072485,
|
|
"aux_distill/mean_u": 0.2910248312232852,
|
|
"aux_distill/n_active_tok": 231.75,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.5142921754862053,
|
|
"calib/avg_num_step_conf": 7.296875,
|
|
"calib/ece": 0.2037890625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0002777023971053845,
|
|
"calib/mean_conf": 0.0110546875,
|
|
"calib/mu_c": 0.011272727272727275,
|
|
"calib/mu_w": 0.01099502487562189,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007451728610016856,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23934999999999998,
|
|
"calib/step_q_c_n": 400.0,
|
|
"calib/step_q_gap": 0.024211852861035388,
|
|
"calib/step_q_w": 0.2151381471389646,
|
|
"calib/step_q_w_n": 1468.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 885.0,
|
|
"completions/max_terminated_length": 885.0,
|
|
"completions/mean_length": 283.5390625,
|
|
"completions/mean_terminated_length": 284.6510009765625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 69.0,
|
|
"epoch": 0.13013333333333332,
|
|
"grad_norm": 0.011241834610700607,
|
|
"learning_rate": 2.166666666666667e-06,
|
|
"loss": 0.0677,
|
|
"num_tokens": 25854855.0,
|
|
"reward": 0.9945212602615356,
|
|
"reward_std": 0.025330815464258194,
|
|
"rewards/accuracy_reward_step": 0.21484375,
|
|
"rewards/final_brier_reward_step": 0.7820113301277161,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 122
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8685710206627846,
|
|
"aux_distill/mean_u": 0.277867029052038,
|
|
"aux_distill/n_active_tok": 246.5,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5124842370744009,
|
|
"calib/avg_num_step_conf": 7.74609375,
|
|
"calib/ece": 0.227421875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.00026733921815888864,
|
|
"calib/mean_conf": 0.010859375000000001,
|
|
"calib/mu_c": 0.010655737704918034,
|
|
"calib/mu_w": 0.010923076923076923,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007957636245102876,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22950338600451467,
|
|
"calib/step_q_c_n": 443.0,
|
|
"calib/step_q_gap": -0.003756354255225608,
|
|
"calib/step_q_w": 0.23325974025974028,
|
|
"calib/step_q_w_n": 1540.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/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": 321.84765625,
|
|
"completions/mean_terminated_length": 323.1098327636719,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 77.0,
|
|
"epoch": 0.1312,
|
|
"grad_norm": 0.009935425594449043,
|
|
"learning_rate": 2.138888888888889e-06,
|
|
"loss": 0.0973,
|
|
"num_tokens": 26042536.0,
|
|
"reward": 1.0024484395980835,
|
|
"reward_std": 0.0032734833657741547,
|
|
"rewards/accuracy_reward_step": 0.23828125,
|
|
"rewards/final_brier_reward_step": 0.766615629196167,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 123
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8649204336106777,
|
|
"aux_distill/mean_u": 0.24206068871074204,
|
|
"aux_distill/n_active_tok": 224.5,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5431367711160457,
|
|
"calib/avg_num_step_conf": 7.0234375,
|
|
"calib/ece": 0.2354296875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0010140636565507035,
|
|
"calib/mean_conf": 0.0106640625,
|
|
"calib/mu_c": 0.011428571428571429,
|
|
"calib/mu_w": 0.010414507772020725,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007122562459964373,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.26356481481481486,
|
|
"calib/step_q_c_n": 432.0,
|
|
"calib/step_q_gap": 0.02929834336532733,
|
|
"calib/step_q_w": 0.23426647144948753,
|
|
"calib/step_q_w_n": 1366.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 768.0,
|
|
"completions/max_terminated_length": 768.0,
|
|
"completions/mean_length": 281.875,
|
|
"completions/mean_terminated_length": 282.98040771484375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 46.0,
|
|
"epoch": 0.13226666666666667,
|
|
"grad_norm": 0.012049200013279915,
|
|
"learning_rate": 2.1111111111111114e-06,
|
|
"loss": 0.103,
|
|
"num_tokens": 26221512.0,
|
|
"reward": 1.0027302503585815,
|
|
"reward_std": 0.0034669036976993084,
|
|
"rewards/accuracy_reward_step": 0.24609375,
|
|
"rewards/final_brier_reward_step": 0.7593668103218079,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 124
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.830241059884429,
|
|
"aux_distill/mean_u": 0.21613227087091394,
|
|
"aux_distill/n_active_tok": 228.5,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.46915446020305357,
|
|
"calib/avg_num_step_conf": 7.2109375,
|
|
"calib/ece": 0.2573046875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0008656901495776182,
|
|
"calib/mean_conf": 0.0122265625,
|
|
"calib/mu_c": 0.011594202898550725,
|
|
"calib/mu_w": 0.012459893048128343,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007559880583289246,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2907912087912088,
|
|
"calib/step_q_c_n": 455.0,
|
|
"calib/step_q_gap": 0.03646338708020952,
|
|
"calib/step_q_w": 0.25432782171099927,
|
|
"calib/step_q_w_n": 1391.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1133.0,
|
|
"completions/max_terminated_length": 1133.0,
|
|
"completions/mean_length": 308.4921875,
|
|
"completions/mean_terminated_length": 309.70196533203125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 64.0,
|
|
"epoch": 0.13333333333333333,
|
|
"grad_norm": 0.009387078694999218,
|
|
"learning_rate": 2.0833333333333334e-06,
|
|
"loss": 0.1047,
|
|
"num_tokens": 26405294.0,
|
|
"reward": 1.0030217170715332,
|
|
"reward_std": 0.004005182534456253,
|
|
"rewards/accuracy_reward_step": 0.26953125,
|
|
"rewards/final_brier_reward_step": 0.7365121245384216,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 125
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8431950993835926,
|
|
"aux_distill/mean_u": 0.2633687595939001,
|
|
"aux_distill/n_active_tok": 250.75,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5391230790920626,
|
|
"calib/avg_num_step_conf": 7.8984375,
|
|
"calib/ece": 0.3106666666666667,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.000827576483857324,
|
|
"calib/mean_conf": 0.010901960784313727,
|
|
"calib/mu_c": 0.011463414634146341,
|
|
"calib/mu_w": 0.010635838150289017,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006535738558744991,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.28768749999999993,
|
|
"calib/step_q_c_n": 480.0,
|
|
"calib/step_q_gap": 0.01721947146562902,
|
|
"calib/step_q_w": 0.2704680285343709,
|
|
"calib/step_q_w_n": 1542.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 851.0,
|
|
"completions/max_terminated_length": 851.0,
|
|
"completions/mean_length": 314.47265625,
|
|
"completions/mean_terminated_length": 315.7059020996094,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 69.0,
|
|
"epoch": 0.1344,
|
|
"grad_norm": 0.011555805802345276,
|
|
"learning_rate": 2.0555555555555555e-06,
|
|
"loss": 0.0752,
|
|
"num_tokens": 26591263.0,
|
|
"reward": 0.9976931810379028,
|
|
"reward_std": 0.02042119950056076,
|
|
"rewards/accuracy_reward_step": 0.3203125,
|
|
"rewards/final_brier_reward_step": 0.6828863024711609,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 126
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8271699827164412,
|
|
"aux_distill/mean_u": 0.23429490647847387,
|
|
"aux_distill/n_active_tok": 241.375,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5146700294241278,
|
|
"calib/avg_num_step_conf": 7.6640625,
|
|
"calib/ece": 0.2266796875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.00048003362757460984,
|
|
"calib/mean_conf": 0.011601562499999999,
|
|
"calib/mu_c": 0.011967213114754098,
|
|
"calib/mu_w": 0.011487179487179488,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006567961826822211,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2443217665615142,
|
|
"calib/step_q_c_n": 317.0,
|
|
"calib/step_q_gap": -0.042304373256115,
|
|
"calib/step_q_w": 0.2866261398176292,
|
|
"calib/step_q_w_n": 1645.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 972.0,
|
|
"completions/max_terminated_length": 972.0,
|
|
"completions/mean_length": 295.76171875,
|
|
"completions/mean_terminated_length": 296.9216003417969,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 60.0,
|
|
"epoch": 0.13546666666666668,
|
|
"grad_norm": 0.008244852535426617,
|
|
"learning_rate": 2.027777777777778e-06,
|
|
"loss": 0.0643,
|
|
"num_tokens": 26770650.0,
|
|
"reward": 1.0027626752853394,
|
|
"reward_std": 0.002931135706603527,
|
|
"rewards/accuracy_reward_step": 0.23828125,
|
|
"rewards/final_brier_reward_step": 0.7672441601753235,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 127
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8970307148993015,
|
|
"aux_distill/mean_u": 0.2829727085306536,
|
|
"aux_distill/n_active_tok": 237.5,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5422732132576692,
|
|
"calib/avg_num_step_conf": 7.44921875,
|
|
"calib/ece": 0.2336328125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0017887984209227752,
|
|
"calib/mean_conf": 0.012460937500000002,
|
|
"calib/mu_c": 0.013809523809523811,
|
|
"calib/mu_w": 0.012020725388601036,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008136916591749836,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.3202928870292887,
|
|
"calib/step_q_c_n": 478.0,
|
|
"calib/step_q_gap": 0.06881493041627262,
|
|
"calib/step_q_w": 0.2514779566130161,
|
|
"calib/step_q_w_n": 1429.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 955.0,
|
|
"completions/max_terminated_length": 955.0,
|
|
"completions/mean_length": 309.6953125,
|
|
"completions/mean_terminated_length": 310.9098205566406,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 47.0,
|
|
"epoch": 0.13653333333333334,
|
|
"grad_norm": 0.008724896237254143,
|
|
"learning_rate": 2.0000000000000003e-06,
|
|
"loss": 0.1234,
|
|
"num_tokens": 26956596.0,
|
|
"reward": 1.003287672996521,
|
|
"reward_std": 0.004546608775854111,
|
|
"rewards/accuracy_reward_step": 0.24609375,
|
|
"rewards/final_brier_reward_step": 0.7604816555976868,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 128
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8841156912967563,
|
|
"aux_distill/mean_u": 0.2761511465104415,
|
|
"aux_distill/n_active_tok": 231.75,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.6039580725907384,
|
|
"calib/avg_num_step_conf": 7.3046875,
|
|
"calib/ece": 0.2541015625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0025312891113892354,
|
|
"calib/mean_conf": 0.0115234375,
|
|
"calib/mu_c": 0.01338235294117647,
|
|
"calib/mu_w": 0.010851063829787235,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.00788240212014039,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.32403940886699506,
|
|
"calib/step_q_c_n": 406.0,
|
|
"calib/step_q_gap": 0.06931782416754151,
|
|
"calib/step_q_w": 0.25472158469945355,
|
|
"calib/step_q_w_n": 1464.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 943.0,
|
|
"completions/max_terminated_length": 943.0,
|
|
"completions/mean_length": 278.8203125,
|
|
"completions/mean_terminated_length": 279.9137268066406,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 79.0,
|
|
"epoch": 0.1376,
|
|
"grad_norm": 0.008412109687924385,
|
|
"learning_rate": 1.9722222222222224e-06,
|
|
"loss": 0.1057,
|
|
"num_tokens": 27130358.0,
|
|
"reward": 1.0034571886062622,
|
|
"reward_std": 0.004877117462456226,
|
|
"rewards/accuracy_reward_step": 0.265625,
|
|
"rewards/final_brier_reward_step": 0.7412894368171692,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 129
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8594199605286121,
|
|
"aux_distill/mean_u": 0.2653886585378518,
|
|
"aux_distill/n_active_tok": 261.125,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5633802816901409,
|
|
"calib/avg_num_step_conf": 8.16015625,
|
|
"calib/ece": 0.2662745098039216,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.00130511328842621,
|
|
"calib/mean_conf": 0.01215686274509804,
|
|
"calib/mu_c": 0.013098591549295775,
|
|
"calib/mu_w": 0.011793478260869565,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007171218478847034,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.38084444444444443,
|
|
"calib/step_q_c_n": 450.0,
|
|
"calib/step_q_gap": 0.11618690936207715,
|
|
"calib/step_q_w": 0.2646575350823673,
|
|
"calib/step_q_w_n": 1639.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/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": 321.6640625,
|
|
"completions/mean_terminated_length": 321.6640625,
|
|
"completions/min_length": 53.0,
|
|
"completions/min_terminated_length": 53.0,
|
|
"epoch": 0.13866666666666666,
|
|
"grad_norm": 0.012677663937211037,
|
|
"learning_rate": 1.944444444444445e-06,
|
|
"loss": 0.0936,
|
|
"num_tokens": 27317992.0,
|
|
"reward": 0.9996272921562195,
|
|
"reward_std": 0.015145928598940372,
|
|
"rewards/accuracy_reward_step": 0.27734375,
|
|
"rewards/final_brier_reward_step": 0.7258172035217285,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 130
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.864942922256887,
|
|
"aux_distill/mean_u": 0.23996074333237236,
|
|
"aux_distill/n_active_tok": 243.0,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.45621744791666674,
|
|
"calib/avg_num_step_conf": 7.70703125,
|
|
"calib/ece": 0.2376171875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0017187499999999998,
|
|
"calib/mean_conf": 0.0123828125,
|
|
"calib/mu_c": 0.01109375,
|
|
"calib/mu_w": 0.0128125,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008160136615881119,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2754566326530612,
|
|
"calib/step_q_c_n": 392.0,
|
|
"calib/step_q_gap": -0.0505262262969704,
|
|
"calib/step_q_w": 0.3259828589500316,
|
|
"calib/step_q_w_n": 1581.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1511.0,
|
|
"completions/max_terminated_length": 1511.0,
|
|
"completions/mean_length": 291.46875,
|
|
"completions/mean_terminated_length": 292.6117858886719,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 54.0,
|
|
"epoch": 0.13973333333333332,
|
|
"grad_norm": 0.015411583706736565,
|
|
"learning_rate": 1.916666666666667e-06,
|
|
"loss": 0.0912,
|
|
"num_tokens": 27498816.0,
|
|
"reward": 1.0026633739471436,
|
|
"reward_std": 0.0034360941499471664,
|
|
"rewards/accuracy_reward_step": 0.25,
|
|
"rewards/final_brier_reward_step": 0.7553269863128662,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 131
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8268191730603576,
|
|
"aux_distill/mean_u": 0.3091863213388308,
|
|
"aux_distill/n_active_tok": 251.75,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.524292324442025,
|
|
"calib/avg_num_step_conf": 7.8671875,
|
|
"calib/ece": 0.3333686274509804,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0003264153511159506,
|
|
"calib/mean_conf": 0.01172941176470588,
|
|
"calib/mu_c": 0.011943181818181818,
|
|
"calib/mu_w": 0.011616766467065867,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007419721841349502,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.2572464788732394,
|
|
"calib/step_q_c_n": 568.0,
|
|
"calib/step_q_gap": -0.03707578945317824,
|
|
"calib/step_q_w": 0.29432226832641767,
|
|
"calib/step_q_w_n": 1446.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 808.0,
|
|
"completions/max_terminated_length": 808.0,
|
|
"completions/mean_length": 304.0625,
|
|
"completions/mean_terminated_length": 305.2549133300781,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 71.0,
|
|
"epoch": 0.1408,
|
|
"grad_norm": 0.010893847793340683,
|
|
"learning_rate": 1.888888888888889e-06,
|
|
"loss": 0.1174,
|
|
"num_tokens": 27682248.0,
|
|
"reward": 1.0001033544540405,
|
|
"reward_std": 0.015689212828874588,
|
|
"rewards/accuracy_reward_step": 0.34375,
|
|
"rewards/final_brier_reward_step": 0.6603628396987915,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 132
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8875812739133835,
|
|
"aux_distill/mean_u": 0.2872198548795962,
|
|
"aux_distill/n_active_tok": 269.375,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.540341812400636,
|
|
"calib/avg_num_step_conf": 8.46484375,
|
|
"calib/ece": 0.121640625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0010227874933757299,
|
|
"calib/mean_conf": 0.011171875,
|
|
"calib/mu_c": 0.012058823529411766,
|
|
"calib/mu_w": 0.011036036036036036,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006010549807161988,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.32817587548638133,
|
|
"calib/step_q_c_n": 257.0,
|
|
"calib/step_q_gap": 0.02241974983193107,
|
|
"calib/step_q_w": 0.30575612565445026,
|
|
"calib/step_q_w_n": 1910.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1065.0,
|
|
"completions/max_terminated_length": 1065.0,
|
|
"completions/mean_length": 330.49609375,
|
|
"completions/mean_terminated_length": 331.79217529296875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 85.0,
|
|
"epoch": 0.14186666666666667,
|
|
"grad_norm": 0.008630180731415749,
|
|
"learning_rate": 1.8611111111111113e-06,
|
|
"loss": 0.0921,
|
|
"num_tokens": 27873199.0,
|
|
"reward": 1.001521110534668,
|
|
"reward_std": 0.0029451206792145967,
|
|
"rewards/accuracy_reward_step": 0.1328125,
|
|
"rewards/final_brier_reward_step": 0.8702297210693359,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 133
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8462215401232243,
|
|
"aux_distill/mean_u": 0.26503370410630267,
|
|
"aux_distill/n_active_tok": 284.875,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.590686274509804,
|
|
"calib/avg_num_step_conf": 8.921875,
|
|
"calib/ece": 0.1923046875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0021078431372549035,
|
|
"calib/mean_conf": 0.0108203125,
|
|
"calib/mu_c": 0.012500000000000002,
|
|
"calib/mu_w": 0.010392156862745099,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.0070510788112418486,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.36191326530612244,
|
|
"calib/step_q_c_n": 392.0,
|
|
"calib/step_q_gap": 0.09844883613064676,
|
|
"calib/step_q_w": 0.2634644291754757,
|
|
"calib/step_q_w_n": 1892.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1043.0,
|
|
"completions/max_terminated_length": 1043.0,
|
|
"completions/mean_length": 352.70703125,
|
|
"completions/mean_terminated_length": 354.0902099609375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 80.0,
|
|
"epoch": 0.14293333333333333,
|
|
"grad_norm": 0.009571945294737816,
|
|
"learning_rate": 1.8333333333333333e-06,
|
|
"loss": 0.0882,
|
|
"num_tokens": 28072444.0,
|
|
"reward": 1.0024555921554565,
|
|
"reward_std": 0.0033246742095798254,
|
|
"rewards/accuracy_reward_step": 0.203125,
|
|
"rewards/final_brier_reward_step": 0.8017863631248474,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 134
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8613733882084489,
|
|
"aux_distill/mean_u": 0.2467592106820447,
|
|
"aux_distill/n_active_tok": 275.625,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5023255813953489,
|
|
"calib/avg_num_step_conf": 8.73828125,
|
|
"calib/ece": 0.1460392156862745,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -8.720930232558252e-05,
|
|
"calib/mean_conf": 0.010823529411764706,
|
|
"calib/mu_c": 0.01075,
|
|
"calib/mu_w": 0.010837209302325582,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006664820967111177,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.2273584905660377,
|
|
"calib/step_q_c_n": 318.0,
|
|
"calib/step_q_gap": -0.062234474519944544,
|
|
"calib/step_q_w": 0.28959296508598226,
|
|
"calib/step_q_w_n": 1919.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1903.0,
|
|
"completions/max_terminated_length": 1903.0,
|
|
"completions/mean_length": 348.3359375,
|
|
"completions/mean_terminated_length": 349.7019958496094,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 67.0,
|
|
"epoch": 0.144,
|
|
"grad_norm": 0.00889371708035469,
|
|
"learning_rate": 1.8055555555555557e-06,
|
|
"loss": 0.0565,
|
|
"num_tokens": 28267498.0,
|
|
"reward": 0.9917947053909302,
|
|
"reward_std": 0.0301792211830616,
|
|
"rewards/accuracy_reward_step": 0.15625,
|
|
"rewards/final_brier_reward_step": 0.8390581607818604,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 135
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8540416387841105,
|
|
"aux_distill/mean_u": 0.2625678658452069,
|
|
"aux_distill/n_active_tok": 268.25,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5024567055980669,
|
|
"calib/avg_num_step_conf": 8.4453125,
|
|
"calib/ece": 0.242890625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 8.215867901731691e-05,
|
|
"calib/mean_conf": 0.011015625000000001,
|
|
"calib/mu_c": 0.011076923076923076,
|
|
"calib/mu_w": 0.01099476439790576,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.00556774243831151,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24743778280542988,
|
|
"calib/step_q_c_n": 442.0,
|
|
"calib/step_q_gap": 0.0013622014100810598,
|
|
"calib/step_q_w": 0.24607558139534882,
|
|
"calib/step_q_w_n": 1720.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1151.0,
|
|
"completions/max_terminated_length": 1151.0,
|
|
"completions/mean_length": 331.91796875,
|
|
"completions/mean_terminated_length": 333.2196350097656,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 70.0,
|
|
"epoch": 0.14506666666666668,
|
|
"grad_norm": 0.009111719205975533,
|
|
"learning_rate": 1.777777777777778e-06,
|
|
"loss": 0.1004,
|
|
"num_tokens": 28460957.0,
|
|
"reward": 0.9988301992416382,
|
|
"reward_std": 0.014630033634603024,
|
|
"rewards/accuracy_reward_step": 0.25390625,
|
|
"rewards/final_brier_reward_step": 0.7476605176925659,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 136
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8544684145599604,
|
|
"aux_distill/mean_u": 0.27133787342407634,
|
|
"aux_distill/n_active_tok": 266.625,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5065359477124183,
|
|
"calib/avg_num_step_conf": 8.33203125,
|
|
"calib/ece": 0.18886274509803924,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 4.9019607843137775e-05,
|
|
"calib/mean_conf": 0.011137254901960783,
|
|
"calib/mu_c": 0.011176470588235295,
|
|
"calib/mu_w": 0.011127450980392157,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005931040340721035,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.31501305483028724,
|
|
"calib/step_q_c_n": 383.0,
|
|
"calib/step_q_gap": 0.06895819768743008,
|
|
"calib/step_q_w": 0.24605485714285716,
|
|
"calib/step_q_w_n": 1750.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2025.0,
|
|
"completions/max_terminated_length": 2025.0,
|
|
"completions/mean_length": 338.55859375,
|
|
"completions/mean_terminated_length": 338.55859375,
|
|
"completions/min_length": 72.0,
|
|
"completions/min_terminated_length": 72.0,
|
|
"epoch": 0.14613333333333334,
|
|
"grad_norm": 0.007216993719339371,
|
|
"learning_rate": 1.75e-06,
|
|
"loss": 0.1535,
|
|
"num_tokens": 28654612.0,
|
|
"reward": 0.9982409477233887,
|
|
"reward_std": 0.014324428513646126,
|
|
"rewards/accuracy_reward_step": 0.19921875,
|
|
"rewards/final_brier_reward_step": 0.8011695146560669,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 137
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.823159608989954,
|
|
"aux_distill/mean_u": 0.21866880539355454,
|
|
"aux_distill/n_active_tok": 266.125,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5339099816289039,
|
|
"calib/avg_num_step_conf": 8.36328125,
|
|
"calib/ece": 0.26776470588235296,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0006376301285976743,
|
|
"calib/mean_conf": 0.010666666666666668,
|
|
"calib/mu_c": 0.011126760563380283,
|
|
"calib/mu_w": 0.010489130434782609,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005525934103155619,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2628949317738791,
|
|
"calib/step_q_c_n": 513.0,
|
|
"calib/step_q_gap": 0.0070472659262132975,
|
|
"calib/step_q_w": 0.25584766584766583,
|
|
"calib/step_q_w_n": 1628.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1020.0,
|
|
"completions/max_terminated_length": 1020.0,
|
|
"completions/mean_length": 317.9453125,
|
|
"completions/mean_terminated_length": 319.1921691894531,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 45.0,
|
|
"epoch": 0.1472,
|
|
"grad_norm": 0.00808706134557724,
|
|
"learning_rate": 1.7222222222222224e-06,
|
|
"loss": 0.0917,
|
|
"num_tokens": 28840342.0,
|
|
"reward": 0.9952016472816467,
|
|
"reward_std": 0.025856774300336838,
|
|
"rewards/accuracy_reward_step": 0.27734375,
|
|
"rewards/final_brier_reward_step": 0.7208722829818726,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 138
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8427855093032122,
|
|
"aux_distill/mean_u": 0.22618297184378638,
|
|
"aux_distill/n_active_tok": 234.75,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5467355154909855,
|
|
"calib/avg_num_step_conf": 7.63671875,
|
|
"calib/ece": 0.22011764705882356,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0030071766147383144,
|
|
"calib/mean_conf": 0.014392156862745097,
|
|
"calib/mu_c": 0.01206896551724138,
|
|
"calib/mu_w": 0.015076142131979694,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0035294117647058825,
|
|
"calib/std_conf": 0.055838730824167446,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23856725146198832,
|
|
"calib/step_q_c_n": 342.0,
|
|
"calib/step_q_gap": -0.00879170700050394,
|
|
"calib/step_q_w": 0.24735895846249226,
|
|
"calib/step_q_w_n": 1613.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1879.0,
|
|
"completions/max_terminated_length": 1879.0,
|
|
"completions/mean_length": 297.8828125,
|
|
"completions/mean_terminated_length": 299.0509948730469,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 69.0,
|
|
"epoch": 0.14826666666666666,
|
|
"grad_norm": 0.0090831583365798,
|
|
"learning_rate": 1.6944444444444446e-06,
|
|
"loss": 0.0901,
|
|
"num_tokens": 29019696.0,
|
|
"reward": 0.9971719980239868,
|
|
"reward_std": 0.019042737782001495,
|
|
"rewards/accuracy_reward_step": 0.2265625,
|
|
"rewards/final_brier_reward_step": 0.7716878652572632,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 139
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8314081691205502,
|
|
"aux_distill/mean_u": 0.21630803803920687,
|
|
"aux_distill/n_active_tok": 245.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5574928977272727,
|
|
"calib/avg_num_step_conf": 7.69921875,
|
|
"calib/ece": 0.301953125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0012045454545454547,
|
|
"calib/mean_conf": 0.010546875,
|
|
"calib/mu_c": 0.011375,
|
|
"calib/mu_w": 0.010170454545454545,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005124785628138508,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2856230188679245,
|
|
"calib/step_q_c_n": 530.0,
|
|
"calib/step_q_gap": 0.05282301886792454,
|
|
"calib/step_q_w": 0.23279999999999998,
|
|
"calib/step_q_w_n": 1441.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 807.0,
|
|
"completions/max_terminated_length": 807.0,
|
|
"completions/mean_length": 304.4453125,
|
|
"completions/mean_terminated_length": 305.63922119140625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 72.0,
|
|
"epoch": 0.14933333333333335,
|
|
"grad_norm": 0.012820438481867313,
|
|
"learning_rate": 1.6666666666666667e-06,
|
|
"loss": 0.1162,
|
|
"num_tokens": 29202650.0,
|
|
"reward": 1.003485918045044,
|
|
"reward_std": 0.004167159553617239,
|
|
"rewards/accuracy_reward_step": 0.3125,
|
|
"rewards/final_brier_reward_step": 0.6944718956947327,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 140
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8130811918526888,
|
|
"aux_distill/mean_u": 0.24643047645682192,
|
|
"aux_distill/n_active_tok": 256.125,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5140083745717549,
|
|
"calib/avg_num_step_conf": 8.08203125,
|
|
"calib/ece": 0.2669921875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0002931100114198712,
|
|
"calib/mean_conf": 0.010351562500000001,
|
|
"calib/mu_c": 0.01056338028169014,
|
|
"calib/mu_w": 0.01027027027027027,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005103751444633031,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24803131991051455,
|
|
"calib/step_q_c_n": 447.0,
|
|
"calib/step_q_gap": 0.028412331007925146,
|
|
"calib/step_q_w": 0.2196189889025894,
|
|
"calib/step_q_w_n": 1622.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1053.0,
|
|
"completions/max_terminated_length": 1053.0,
|
|
"completions/mean_length": 345.2109375,
|
|
"completions/mean_terminated_length": 346.5647277832031,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 80.0,
|
|
"epoch": 0.1504,
|
|
"grad_norm": 0.022635893896222115,
|
|
"learning_rate": 1.638888888888889e-06,
|
|
"loss": 0.1239,
|
|
"num_tokens": 29398120.0,
|
|
"reward": 1.0028630495071411,
|
|
"reward_std": 0.0038094979245215654,
|
|
"rewards/accuracy_reward_step": 0.27734375,
|
|
"rewards/final_brier_reward_step": 0.728382408618927,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 141
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.834231823682785,
|
|
"aux_distill/mean_u": 0.23377564743541748,
|
|
"aux_distill/n_active_tok": 254.5,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5165050062578223,
|
|
"calib/avg_num_step_conf": 8.03125,
|
|
"calib/ece": 0.2549609375,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.000297246558197746,
|
|
"calib/mean_conf": 0.0106640625,
|
|
"calib/mu_c": 0.01088235294117647,
|
|
"calib/mu_w": 0.010585106382978724,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004671685562630874,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.21933649289099527,
|
|
"calib/step_q_c_n": 422.0,
|
|
"calib/step_q_gap": -0.0052993700221014295,
|
|
"calib/step_q_w": 0.2246358629130967,
|
|
"calib/step_q_w_n": 1634.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/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": 331.01953125,
|
|
"completions/mean_terminated_length": 332.3176574707031,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 99.0,
|
|
"epoch": 0.15146666666666667,
|
|
"grad_norm": 0.011043130420148373,
|
|
"learning_rate": 1.6111111111111113e-06,
|
|
"loss": 0.0779,
|
|
"num_tokens": 29588021.0,
|
|
"reward": 1.002822756767273,
|
|
"reward_std": 0.003341506700962782,
|
|
"rewards/accuracy_reward_step": 0.265625,
|
|
"rewards/final_brier_reward_step": 0.7400206327438354,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 142
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8353591244667768,
|
|
"aux_distill/mean_u": 0.21951536373226854,
|
|
"aux_distill/n_active_tok": 226.5,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5444696495619524,
|
|
"calib/avg_num_step_conf": 7.125,
|
|
"calib/ece": 0.254765625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0010325406758448057,
|
|
"calib/mean_conf": 0.010859375000000001,
|
|
"calib/mu_c": 0.01161764705882353,
|
|
"calib/mu_w": 0.010585106382978724,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004424248479614928,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22172015113350127,
|
|
"calib/step_q_c_n": 397.0,
|
|
"calib/step_q_gap": 0.0011195905168228937,
|
|
"calib/step_q_w": 0.22060056061667838,
|
|
"calib/step_q_w_n": 1427.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1855.0,
|
|
"completions/max_terminated_length": 1855.0,
|
|
"completions/mean_length": 303.89453125,
|
|
"completions/mean_terminated_length": 305.0863037109375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 85.0,
|
|
"epoch": 0.15253333333333333,
|
|
"grad_norm": 0.012597390450537205,
|
|
"learning_rate": 1.5833333333333333e-06,
|
|
"loss": 0.0566,
|
|
"num_tokens": 29773154.0,
|
|
"reward": 1.0030171871185303,
|
|
"reward_std": 0.003955106250941753,
|
|
"rewards/accuracy_reward_step": 0.265625,
|
|
"rewards/final_brier_reward_step": 0.7404093742370605,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 143
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7986927554011345,
|
|
"aux_distill/mean_u": 0.18862368095835338,
|
|
"aux_distill/n_active_tok": 228.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5245000000000001,
|
|
"calib/avg_num_step_conf": 7.16015625,
|
|
"calib/ece": 0.30298039215686273,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0007357142857142857,
|
|
"calib/mean_conf": 0.010745098039215688,
|
|
"calib/mu_c": 0.01125,
|
|
"calib/mu_w": 0.010514285714285714,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004833880174291448,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.21494625719769672,
|
|
"calib/step_q_c_n": 521.0,
|
|
"calib/step_q_gap": 0.02100746146598942,
|
|
"calib/step_q_w": 0.1939387957317073,
|
|
"calib/step_q_w_n": 1312.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 774.0,
|
|
"completions/max_terminated_length": 774.0,
|
|
"completions/mean_length": 297.7578125,
|
|
"completions/mean_terminated_length": 298.9255065917969,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 73.0,
|
|
"epoch": 0.1536,
|
|
"grad_norm": 0.011825519613921642,
|
|
"learning_rate": 1.5555555555555558e-06,
|
|
"loss": 0.0774,
|
|
"num_tokens": 29953508.0,
|
|
"reward": 0.9995401501655579,
|
|
"reward_std": 0.01496032252907753,
|
|
"rewards/accuracy_reward_step": 0.3125,
|
|
"rewards/final_brier_reward_step": 0.6904866695404053,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 144
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8074095947667956,
|
|
"aux_distill/mean_u": 0.22168011550755964,
|
|
"aux_distill/n_active_tok": 248.75,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5094478692039668,
|
|
"calib/avg_num_step_conf": 7.7734375,
|
|
"calib/ece": 0.346078431372549,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0003182792816939159,
|
|
"calib/mean_conf": 0.010784313725490196,
|
|
"calib/mu_c": 0.01098901098901099,
|
|
"calib/mu_w": 0.010670731707317074,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004534459457117572,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22369472789115646,
|
|
"calib/step_q_c_n": 588.0,
|
|
"calib/step_q_gap": 0.04426798038759011,
|
|
"calib/step_q_w": 0.17942674750356635,
|
|
"calib/step_q_w_n": 1402.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/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": 309.578125,
|
|
"completions/mean_terminated_length": 309.578125,
|
|
"completions/min_length": 91.0,
|
|
"completions/min_terminated_length": 91.0,
|
|
"epoch": 0.15466666666666667,
|
|
"grad_norm": 0.012043782509863377,
|
|
"learning_rate": 1.527777777777778e-06,
|
|
"loss": 0.116,
|
|
"num_tokens": 30135464.0,
|
|
"reward": 0.9999317526817322,
|
|
"reward_std": 0.015501899644732475,
|
|
"rewards/accuracy_reward_step": 0.35546875,
|
|
"rewards/final_brier_reward_step": 0.6483011245727539,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 145
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8472644705325365,
|
|
"aux_distill/mean_u": 0.2108220952642413,
|
|
"aux_distill/n_active_tok": 256.625,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5413558578323826,
|
|
"calib/avg_num_step_conf": 8.01953125,
|
|
"calib/ece": 0.15788235294117647,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.001061869240895132,
|
|
"calib/mean_conf": 0.010745098039215688,
|
|
"calib/mu_c": 0.01162790697674419,
|
|
"calib/mu_w": 0.010566037735849057,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004833880174291448,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.19057553956834533,
|
|
"calib/step_q_c_n": 278.0,
|
|
"calib/step_q_gap": -0.02633657310771101,
|
|
"calib/step_q_w": 0.21691211267605634,
|
|
"calib/step_q_w_n": 1775.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2752.0,
|
|
"completions/max_terminated_length": 2752.0,
|
|
"completions/mean_length": 345.4375,
|
|
"completions/mean_terminated_length": 345.4375,
|
|
"completions/min_length": 81.0,
|
|
"completions/min_terminated_length": 81.0,
|
|
"epoch": 0.15573333333333333,
|
|
"grad_norm": 0.008280741982161999,
|
|
"learning_rate": 1.5e-06,
|
|
"loss": 0.1301,
|
|
"num_tokens": 30331112.0,
|
|
"reward": 0.9979776740074158,
|
|
"reward_std": 0.01372268982231617,
|
|
"rewards/accuracy_reward_step": 0.16796875,
|
|
"rewards/final_brier_reward_step": 0.8318929672241211,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 146
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7779547851532698,
|
|
"aux_distill/mean_u": 0.19806066419509277,
|
|
"aux_distill/n_active_tok": 259.75,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5379230905546695,
|
|
"calib/avg_num_step_conf": 8.15625,
|
|
"calib/ece": 0.2119607843137255,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0006911217437533222,
|
|
"calib/mean_conf": 0.011568627450980393,
|
|
"calib/mu_c": 0.012105263157894735,
|
|
"calib/mu_w": 0.011414141414141413,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005229084957108481,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.19256997455470737,
|
|
"calib/step_q_c_n": 393.0,
|
|
"calib/step_q_gap": 0.0012975851741763855,
|
|
"calib/step_q_w": 0.19127238938053098,
|
|
"calib/step_q_w_n": 1695.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1285.0,
|
|
"completions/max_terminated_length": 1285.0,
|
|
"completions/mean_length": 323.95703125,
|
|
"completions/mean_terminated_length": 325.22747802734375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 70.0,
|
|
"epoch": 0.1568,
|
|
"grad_norm": 0.010307558812201023,
|
|
"learning_rate": 1.4722222222222225e-06,
|
|
"loss": 0.0717,
|
|
"num_tokens": 30517725.0,
|
|
"reward": 0.9987087249755859,
|
|
"reward_std": 0.014628560282289982,
|
|
"rewards/accuracy_reward_step": 0.22265625,
|
|
"rewards/final_brier_reward_step": 0.7786675691604614,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 147
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8126857522875071,
|
|
"aux_distill/mean_u": 0.2364917961944136,
|
|
"aux_distill/n_active_tok": 230.0,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.4886096833555429,
|
|
"calib/avg_num_step_conf": 7.25390625,
|
|
"calib/ece": 0.39140625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0002690526048607141,
|
|
"calib/mean_conf": 0.0109375,
|
|
"calib/mu_c": 0.010776699029126214,
|
|
"calib/mu_w": 0.011045751633986928,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004318835925339142,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.18042207792207793,
|
|
"calib/step_q_c_n": 616.0,
|
|
"calib/step_q_gap": -0.03018227340749502,
|
|
"calib/step_q_w": 0.21060435132957295,
|
|
"calib/step_q_w_n": 1241.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 829.0,
|
|
"completions/max_terminated_length": 829.0,
|
|
"completions/mean_length": 292.89453125,
|
|
"completions/mean_terminated_length": 294.04315185546875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 60.0,
|
|
"epoch": 0.15786666666666666,
|
|
"grad_norm": 0.011318879202008247,
|
|
"learning_rate": 1.4444444444444445e-06,
|
|
"loss": 0.0996,
|
|
"num_tokens": 30697818.0,
|
|
"reward": 1.0042667388916016,
|
|
"reward_std": 0.003717840416356921,
|
|
"rewards/accuracy_reward_step": 0.40234375,
|
|
"rewards/final_brier_reward_step": 0.6061898469924927,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 148
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8063416481018066,
|
|
"aux_distill/mean_u": 0.19125913072319098,
|
|
"aux_distill/n_active_tok": 256.5,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.6097099621689785,
|
|
"calib/avg_num_step_conf": 8.14453125,
|
|
"calib/ece": 0.2275,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0026330390920554884,
|
|
"calib/mean_conf": 0.01078125,
|
|
"calib/mu_c": 0.01278688524590164,
|
|
"calib/mu_w": 0.010153846153846152,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004525858861862574,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.3081656804733728,
|
|
"calib/step_q_c_n": 338.0,
|
|
"calib/step_q_gap": 0.07366081498968649,
|
|
"calib/step_q_w": 0.2345048654836863,
|
|
"calib/step_q_w_n": 1747.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1537.0,
|
|
"completions/max_terminated_length": 1537.0,
|
|
"completions/mean_length": 351.89453125,
|
|
"completions/mean_terminated_length": 353.2745361328125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 91.0,
|
|
"epoch": 0.15893333333333334,
|
|
"grad_norm": 0.009276711381971836,
|
|
"learning_rate": 1.4166666666666667e-06,
|
|
"loss": 0.1032,
|
|
"num_tokens": 30892359.0,
|
|
"reward": 1.0029785633087158,
|
|
"reward_std": 0.0035227080807089806,
|
|
"rewards/accuracy_reward_step": 0.23828125,
|
|
"rewards/final_brier_reward_step": 0.7676757574081421,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 149
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8243851810693741,
|
|
"aux_distill/mean_u": 0.22265294731953492,
|
|
"aux_distill/n_active_tok": 228.125,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5312522045855379,
|
|
"calib/avg_num_step_conf": 7.1640625,
|
|
"calib/ece": 0.305234375,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0008141093474426809,
|
|
"calib/mean_conf": 0.011171875,
|
|
"calib/mu_c": 0.011728395061728394,
|
|
"calib/mu_w": 0.010914285714285713,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.00501888523323407,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.21167130977130977,
|
|
"calib/step_q_c_n": 481.0,
|
|
"calib/step_q_gap": 0.03153073327463568,
|
|
"calib/step_q_w": 0.1801405764966741,
|
|
"calib/step_q_w_n": 1353.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 944.0,
|
|
"completions/max_terminated_length": 944.0,
|
|
"completions/mean_length": 288.16015625,
|
|
"completions/mean_terminated_length": 290.42913818359375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 63.0,
|
|
"epoch": 0.16,
|
|
"grad_norm": 0.009506311267614365,
|
|
"learning_rate": 1.3888888888888892e-06,
|
|
"loss": 0.0897,
|
|
"num_tokens": 31071088.0,
|
|
"reward": 1.0036358833312988,
|
|
"reward_std": 0.0035198626574128866,
|
|
"rewards/accuracy_reward_step": 0.31640625,
|
|
"rewards/final_brier_reward_step": 0.6908656358718872,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 150
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8354400210082531,
|
|
"aux_distill/mean_u": 0.21287575401324318,
|
|
"aux_distill/n_active_tok": 267.375,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.4718146718146718,
|
|
"calib/avg_num_step_conf": 8.55859375,
|
|
"calib/ece": 0.26301960784313727,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0004787644787644773,
|
|
"calib/mean_conf": 0.011490196078431372,
|
|
"calib/mu_c": 0.011142857142857144,
|
|
"calib/mu_w": 0.011621621621621621,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004944287847050819,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2201898734177215,
|
|
"calib/step_q_c_n": 474.0,
|
|
"calib/step_q_gap": 0.003247532124768665,
|
|
"calib/step_q_w": 0.21694234129295284,
|
|
"calib/step_q_w_n": 1717.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2159.0,
|
|
"completions/max_terminated_length": 2159.0,
|
|
"completions/mean_length": 345.140625,
|
|
"completions/mean_terminated_length": 346.494140625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 81.0,
|
|
"epoch": 0.16106666666666666,
|
|
"grad_norm": 0.009733528830111027,
|
|
"learning_rate": 1.3611111111111112e-06,
|
|
"loss": 0.1118,
|
|
"num_tokens": 31266468.0,
|
|
"reward": 0.9990626573562622,
|
|
"reward_std": 0.014477596618235111,
|
|
"rewards/accuracy_reward_step": 0.2734375,
|
|
"rewards/final_brier_reward_step": 0.7285941243171692,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 151
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8009604308754206,
|
|
"aux_distill/mean_u": 0.1808466434731913,
|
|
"aux_distill/n_active_tok": 222.5,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.44384057971014496,
|
|
"calib/avg_num_step_conf": 7.01953125,
|
|
"calib/ece": 0.26953125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0012318840579710125,
|
|
"calib/mean_conf": 0.01171875,
|
|
"calib/mu_c": 0.010833333333333334,
|
|
"calib/mu_w": 0.012065217391304346,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004611360801054283,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.27999999999999997,
|
|
"calib/step_q_c_n": 480.0,
|
|
"calib/step_q_gap": 0.06599848139711459,
|
|
"calib/step_q_w": 0.21400151860288538,
|
|
"calib/step_q_w_n": 1317.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1475.0,
|
|
"completions/max_terminated_length": 1475.0,
|
|
"completions/mean_length": 315.1171875,
|
|
"completions/mean_terminated_length": 316.35296630859375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 87.0,
|
|
"epoch": 0.16213333333333332,
|
|
"grad_norm": 0.011670260690152645,
|
|
"learning_rate": 1.3333333333333334e-06,
|
|
"loss": 0.0747,
|
|
"num_tokens": 31452530.0,
|
|
"reward": 0.9990614652633667,
|
|
"reward_std": 0.014696644619107246,
|
|
"rewards/accuracy_reward_step": 0.28125,
|
|
"rewards/final_brier_reward_step": 0.720779299736023,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 152
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8575300592929125,
|
|
"aux_distill/mean_u": 0.25102041297178856,
|
|
"aux_distill/n_active_tok": 272.5,
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/auroc": 0.5212359550561798,
|
|
"calib/avg_num_step_conf": 8.8203125,
|
|
"calib/ece": 0.28529644268774706,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.000455430711610489,
|
|
"calib/mean_conf": 0.011146245059288538,
|
|
"calib/mu_c": 0.011466666666666668,
|
|
"calib/mu_w": 0.01101123595505618,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004519430572410558,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.1944256120527307,
|
|
"calib/step_q_c_n": 531.0,
|
|
"calib/step_q_gap": -0.009506640408184164,
|
|
"calib/step_q_w": 0.20393225246091487,
|
|
"calib/step_q_w_n": 1727.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1845.0,
|
|
"completions/max_terminated_length": 1845.0,
|
|
"completions/mean_length": 385.03515625,
|
|
"completions/mean_terminated_length": 388.0669250488281,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 70.0,
|
|
"epoch": 0.1632,
|
|
"grad_norm": 0.009350521489977837,
|
|
"learning_rate": 1.3055555555555556e-06,
|
|
"loss": 0.1013,
|
|
"num_tokens": 31658419.0,
|
|
"reward": 0.9915691614151001,
|
|
"reward_std": 0.037060413509607315,
|
|
"rewards/accuracy_reward_step": 0.29296875,
|
|
"rewards/final_brier_reward_step": 0.7018882632255554,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 153
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7743056090548635,
|
|
"aux_distill/mean_u": 0.18933966077084272,
|
|
"aux_distill/n_active_tok": 226.0,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5237829932446549,
|
|
"calib/avg_num_step_conf": 7.09375,
|
|
"calib/ece": 0.3130078125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0007040880284142348,
|
|
"calib/mean_conf": 0.0112109375,
|
|
"calib/mu_c": 0.011686746987951807,
|
|
"calib/mu_w": 0.010982658959537572,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.00464548225387782,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.17774011299435027,
|
|
"calib/step_q_c_n": 531.0,
|
|
"calib/step_q_gap": -0.05550665743366526,
|
|
"calib/step_q_w": 0.23324677042801553,
|
|
"calib/step_q_w_n": 1285.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 698.0,
|
|
"completions/max_terminated_length": 698.0,
|
|
"completions/mean_length": 291.1015625,
|
|
"completions/mean_terminated_length": 292.2431640625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 101.0,
|
|
"epoch": 0.16426666666666667,
|
|
"grad_norm": 0.011979682371020317,
|
|
"learning_rate": 1.2777777777777779e-06,
|
|
"loss": 0.1075,
|
|
"num_tokens": 31837381.0,
|
|
"reward": 1.0037152767181396,
|
|
"reward_std": 0.004051705356687307,
|
|
"rewards/accuracy_reward_step": 0.32421875,
|
|
"rewards/final_brier_reward_step": 0.683212161064148,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 154
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7905787313356996,
|
|
"aux_distill/mean_u": 0.18707750770316992,
|
|
"aux_distill/n_active_tok": 227.625,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.4896075823079481,
|
|
"calib/avg_num_step_conf": 7.1484375,
|
|
"calib/ece": 0.23109375,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0001662786830728298,
|
|
"calib/mean_conf": 0.011093750000000001,
|
|
"calib/mu_c": 0.010967741935483872,
|
|
"calib/mu_w": 0.011134020618556702,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004798172666495027,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2575570776255708,
|
|
"calib/step_q_c_n": 438.0,
|
|
"calib/step_q_gap": 0.05267791095890412,
|
|
"calib/step_q_w": 0.20487916666666667,
|
|
"calib/step_q_w_n": 1392.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 763.0,
|
|
"completions/max_terminated_length": 763.0,
|
|
"completions/mean_length": 297.05078125,
|
|
"completions/mean_terminated_length": 298.2156982421875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 76.0,
|
|
"epoch": 0.16533333333333333,
|
|
"grad_norm": 0.01167488656938076,
|
|
"learning_rate": 1.25e-06,
|
|
"loss": 0.1079,
|
|
"num_tokens": 32020642.0,
|
|
"reward": 1.0025831460952759,
|
|
"reward_std": 0.003791379276663065,
|
|
"rewards/accuracy_reward_step": 0.2421875,
|
|
"rewards/final_brier_reward_step": 0.7629789113998413,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 155
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7945300079882145,
|
|
"aux_distill/mean_u": 0.17373346807153683,
|
|
"aux_distill/n_active_tok": 215.25,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.5321022284556646,
|
|
"calib/avg_num_step_conf": 6.7265625,
|
|
"calib/ece": 0.25862745098039214,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.000689574567554932,
|
|
"calib/mean_conf": 0.011960784313725489,
|
|
"calib/mu_c": 0.012463768115942029,
|
|
"calib/mu_w": 0.011774193548387097,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005095021731423283,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.18386792452830192,
|
|
"calib/step_q_c_n": 424.0,
|
|
"calib/step_q_gap": -0.04083931738232985,
|
|
"calib/step_q_w": 0.22470724191063177,
|
|
"calib/step_q_w_n": 1298.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2113.0,
|
|
"completions/max_terminated_length": 2113.0,
|
|
"completions/mean_length": 302.0390625,
|
|
"completions/mean_terminated_length": 302.0390625,
|
|
"completions/min_length": 79.0,
|
|
"completions/min_terminated_length": 79.0,
|
|
"epoch": 0.1664,
|
|
"grad_norm": 0.011816912330687046,
|
|
"learning_rate": 1.2222222222222223e-06,
|
|
"loss": 0.1044,
|
|
"num_tokens": 32202724.0,
|
|
"reward": 0.9954628944396973,
|
|
"reward_std": 0.0261252298951149,
|
|
"rewards/accuracy_reward_step": 0.26953125,
|
|
"rewards/final_brier_reward_step": 0.7292070388793945,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 156
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8267823327332735,
|
|
"aux_distill/mean_u": 0.22326160780332782,
|
|
"aux_distill/n_active_tok": 226.875,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5319767441860466,
|
|
"calib/avg_num_step_conf": 7.1640625,
|
|
"calib/ece": 0.32421875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0007387140902872773,
|
|
"calib/mean_conf": 0.01171875,
|
|
"calib/mu_c": 0.012209302325581395,
|
|
"calib/mu_w": 0.011470588235294118,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004938587696649721,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23325515021459226,
|
|
"calib/step_q_c_n": 466.0,
|
|
"calib/step_q_gap": 0.0031820507993875757,
|
|
"calib/step_q_w": 0.23007309941520468,
|
|
"calib/step_q_w_n": 1368.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 802.0,
|
|
"completions/max_terminated_length": 802.0,
|
|
"completions/mean_length": 295.01171875,
|
|
"completions/mean_terminated_length": 296.16864013671875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 69.0,
|
|
"epoch": 0.16746666666666668,
|
|
"grad_norm": 0.011997681111097336,
|
|
"learning_rate": 1.1944444444444446e-06,
|
|
"loss": 0.085,
|
|
"num_tokens": 32381975.0,
|
|
"reward": 1.0040206909179688,
|
|
"reward_std": 0.0045976778492331505,
|
|
"rewards/accuracy_reward_step": 0.3359375,
|
|
"rewards/final_brier_reward_step": 0.6721038818359375,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 157
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8071128753945231,
|
|
"aux_distill/mean_u": 0.1967308060436971,
|
|
"aux_distill/n_active_tok": 216.125,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5416777807415311,
|
|
"calib/avg_num_step_conf": 7.16015625,
|
|
"calib/ece": 0.3492549019607843,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0010082688716991223,
|
|
"calib/mean_conf": 0.011529411764705882,
|
|
"calib/mu_c": 0.012173913043478262,
|
|
"calib/mu_w": 0.01116564417177914,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005127216655688678,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22468283582089552,
|
|
"calib/step_q_c_n": 536.0,
|
|
"calib/step_q_gap": 0.014818919089977983,
|
|
"calib/step_q_w": 0.20986391673091753,
|
|
"calib/step_q_w_n": 1297.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1445.0,
|
|
"completions/max_terminated_length": 1445.0,
|
|
"completions/mean_length": 303.7578125,
|
|
"completions/mean_terminated_length": 304.94903564453125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 72.0,
|
|
"epoch": 0.16853333333333334,
|
|
"grad_norm": 0.012824364937841892,
|
|
"learning_rate": 1.1666666666666668e-06,
|
|
"loss": 0.1108,
|
|
"num_tokens": 32564977.0,
|
|
"reward": 1.0003893375396729,
|
|
"reward_std": 0.01565980166196823,
|
|
"rewards/accuracy_reward_step": 0.359375,
|
|
"rewards/final_brier_reward_step": 0.6453101634979248,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 158
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8039664840325713,
|
|
"aux_distill/mean_u": 0.19331534523575358,
|
|
"aux_distill/n_active_tok": 210.375,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5180052120350629,
|
|
"calib/avg_num_step_conf": 6.59375,
|
|
"calib/ece": 0.2500390625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0003529969201610979,
|
|
"calib/mean_conf": 0.0116796875,
|
|
"calib/mu_c": 0.011940298507462687,
|
|
"calib/mu_w": 0.011587301587301589,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004912410294584905,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.20959330143540672,
|
|
"calib/step_q_c_n": 418.0,
|
|
"calib/step_q_gap": -0.0004089820291602042,
|
|
"calib/step_q_w": 0.21000228346456692,
|
|
"calib/step_q_w_n": 1270.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 711.0,
|
|
"completions/max_terminated_length": 711.0,
|
|
"completions/mean_length": 276.69140625,
|
|
"completions/mean_terminated_length": 277.7764892578125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 72.0,
|
|
"epoch": 0.1696,
|
|
"grad_norm": 0.011785455979406834,
|
|
"learning_rate": 1.138888888888889e-06,
|
|
"loss": 0.0765,
|
|
"num_tokens": 32740594.0,
|
|
"reward": 1.003044605255127,
|
|
"reward_std": 0.0038546710275113583,
|
|
"rewards/accuracy_reward_step": 0.26171875,
|
|
"rewards/final_brier_reward_step": 0.744370698928833,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 159
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8207447240129113,
|
|
"aux_distill/mean_u": 0.19969249503579878,
|
|
"aux_distill/n_active_tok": 222.25,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5319074333800841,
|
|
"calib/avg_num_step_conf": 7.26953125,
|
|
"calib/ece": 0.25894117647058823,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0009209911173445537,
|
|
"calib/mean_conf": 0.011647058823529413,
|
|
"calib/mu_c": 0.012318840579710146,
|
|
"calib/mu_w": 0.011397849462365592,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005498033621434317,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.20069499999999998,
|
|
"calib/step_q_c_n": 400.0,
|
|
"calib/step_q_gap": -0.04821848391512665,
|
|
"calib/step_q_w": 0.24891348391512663,
|
|
"calib/step_q_w_n": 1461.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2247.0,
|
|
"completions/max_terminated_length": 2247.0,
|
|
"completions/mean_length": 292.94921875,
|
|
"completions/mean_terminated_length": 294.0980529785156,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 91.0,
|
|
"epoch": 0.17066666666666666,
|
|
"grad_norm": 0.016354886814951897,
|
|
"learning_rate": 1.111111111111111e-06,
|
|
"loss": 0.0906,
|
|
"num_tokens": 32920429.0,
|
|
"reward": 0.9993313550949097,
|
|
"reward_std": 0.015432949177920818,
|
|
"rewards/accuracy_reward_step": 0.26953125,
|
|
"rewards/final_brier_reward_step": 0.7330378293991089,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 160
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7666348684579134,
|
|
"aux_distill/mean_u": 0.19474998295763324,
|
|
"aux_distill/n_active_tok": 206.5,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5205498094719652,
|
|
"calib/avg_num_step_conf": 6.59375,
|
|
"calib/ece": 0.33294117647058824,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0003504354926510602,
|
|
"calib/mean_conf": 0.01215686274509804,
|
|
"calib/mu_c": 0.012386363636363635,
|
|
"calib/mu_w": 0.012035928143712574,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005428221638587695,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.18919921875,
|
|
"calib/step_q_c_n": 512.0,
|
|
"calib/step_q_gap": 0.00034998405612243766,
|
|
"calib/step_q_w": 0.18884923469387757,
|
|
"calib/step_q_w_n": 1176.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1132.0,
|
|
"completions/max_terminated_length": 1132.0,
|
|
"completions/mean_length": 284.984375,
|
|
"completions/mean_terminated_length": 286.10198974609375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 70.0,
|
|
"epoch": 0.17173333333333332,
|
|
"grad_norm": 0.014039784669876099,
|
|
"learning_rate": 1.0833333333333335e-06,
|
|
"loss": 0.0816,
|
|
"num_tokens": 33097305.0,
|
|
"reward": 1.0002632141113281,
|
|
"reward_std": 0.016159221529960632,
|
|
"rewards/accuracy_reward_step": 0.34375,
|
|
"rewards/final_brier_reward_step": 0.6606827974319458,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 161
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7794603575021029,
|
|
"aux_distill/mean_u": 0.17591780110662303,
|
|
"aux_distill/n_active_tok": 187.0,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5845205057045946,
|
|
"calib/avg_num_step_conf": 5.890625,
|
|
"calib/ece": 0.437109375,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0018538390379278461,
|
|
"calib/mean_conf": 0.012109375000000002,
|
|
"calib/mu_c": 0.013130434782608696,
|
|
"calib/mu_w": 0.01127659574468085,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005748959654526634,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.25271074380165287,
|
|
"calib/step_q_c_n": 605.0,
|
|
"calib/step_q_gap": 0.03650653560674702,
|
|
"calib/step_q_w": 0.21620420819490585,
|
|
"calib/step_q_w_n": 903.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 849.0,
|
|
"completions/max_terminated_length": 849.0,
|
|
"completions/mean_length": 254.84375,
|
|
"completions/mean_terminated_length": 255.84315490722656,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 93.0,
|
|
"epoch": 0.1728,
|
|
"grad_norm": 0.022141166031360626,
|
|
"learning_rate": 1.0555555555555557e-06,
|
|
"loss": 0.1011,
|
|
"num_tokens": 33266689.0,
|
|
"reward": 1.0058085918426514,
|
|
"reward_std": 0.00576754380017519,
|
|
"rewards/accuracy_reward_step": 0.44921875,
|
|
"rewards/final_brier_reward_step": 0.5623984336853027,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 162
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8130544321611524,
|
|
"aux_distill/mean_u": 0.23703817534060598,
|
|
"aux_distill/n_active_tok": 214.25,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.505476330491032,
|
|
"calib/avg_num_step_conf": 6.6953125,
|
|
"calib/ece": 0.28615686274509805,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0005049985298441629,
|
|
"calib/mean_conf": 0.011882352941176472,
|
|
"calib/mu_c": 0.012236842105263157,
|
|
"calib/mu_w": 0.011731843575418994,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005772303916365068,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.19712616822429907,
|
|
"calib/step_q_c_n": 428.0,
|
|
"calib/step_q_gap": -0.031108668478655815,
|
|
"calib/step_q_w": 0.22823483670295489,
|
|
"calib/step_q_w_n": 1286.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/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": 298.65625,
|
|
"completions/mean_terminated_length": 298.65625,
|
|
"completions/min_length": 68.0,
|
|
"completions/min_terminated_length": 68.0,
|
|
"epoch": 0.17386666666666667,
|
|
"grad_norm": 0.011381106451153755,
|
|
"learning_rate": 1.0277777777777777e-06,
|
|
"loss": 0.1025,
|
|
"num_tokens": 33447977.0,
|
|
"reward": 0.9996396899223328,
|
|
"reward_std": 0.014509235508739948,
|
|
"rewards/accuracy_reward_step": 0.296875,
|
|
"rewards/final_brier_reward_step": 0.706310510635376,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 163
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8056136062368751,
|
|
"aux_distill/mean_u": 0.17818100797914577,
|
|
"aux_distill/n_active_tok": 220.375,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5464912280701754,
|
|
"calib/avg_num_step_conf": 7.03515625,
|
|
"calib/ece": 0.2853125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.001146198830409357,
|
|
"calib/mean_conf": 0.0115625,
|
|
"calib/mu_c": 0.012368421052631579,
|
|
"calib/mu_w": 0.011222222222222222,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.004749588797990832,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.14924210526315787,
|
|
"calib/step_q_c_n": 475.0,
|
|
"calib/step_q_gap": -0.04428421449551484,
|
|
"calib/step_q_w": 0.1935263197586727,
|
|
"calib/step_q_w_n": 1326.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 951.0,
|
|
"completions/max_terminated_length": 951.0,
|
|
"completions/mean_length": 299.203125,
|
|
"completions/mean_terminated_length": 300.3764953613281,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 49.0,
|
|
"epoch": 0.17493333333333333,
|
|
"grad_norm": 0.012084920890629292,
|
|
"learning_rate": 1.0000000000000002e-06,
|
|
"loss": 0.0816,
|
|
"num_tokens": 33630709.0,
|
|
"reward": 0.999687671661377,
|
|
"reward_std": 0.014886870048940182,
|
|
"rewards/accuracy_reward_step": 0.296875,
|
|
"rewards/final_brier_reward_step": 0.7064065933227539,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 164
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8530535958707333,
|
|
"aux_distill/mean_u": 0.21941893875711735,
|
|
"aux_distill/n_active_tok": 209.625,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5641502746765905,
|
|
"calib/avg_num_step_conf": 6.94140625,
|
|
"calib/ece": 0.21129411764705883,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0016400850611376933,
|
|
"calib/mean_conf": 0.01223529411764706,
|
|
"calib/mu_c": 0.013508771929824562,
|
|
"calib/mu_w": 0.011868686868686869,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005679644545955035,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.20603399433427763,
|
|
"calib/step_q_c_n": 353.0,
|
|
"calib/step_q_gap": -0.004884334317407785,
|
|
"calib/step_q_w": 0.21091832865168542,
|
|
"calib/step_q_w_n": 1424.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/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": 300.19140625,
|
|
"completions/mean_terminated_length": 301.36865234375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 72.0,
|
|
"epoch": 0.176,
|
|
"grad_norm": 0.012888389639556408,
|
|
"learning_rate": 9.722222222222224e-07,
|
|
"loss": 0.0669,
|
|
"num_tokens": 33813134.0,
|
|
"reward": 0.9990109205245972,
|
|
"reward_std": 0.015108481980860233,
|
|
"rewards/accuracy_reward_step": 0.22265625,
|
|
"rewards/final_brier_reward_step": 0.7792718410491943,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 165
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8504758551716805,
|
|
"aux_distill/mean_u": 0.2541528017707982,
|
|
"aux_distill/n_active_tok": 225.625,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.4875205606808267,
|
|
"calib/avg_num_step_conf": 7.09765625,
|
|
"calib/ece": 0.2967578125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -9.22548809268401e-05,
|
|
"calib/mean_conf": 0.011835937500000001,
|
|
"calib/mu_c": 0.011772151898734177,
|
|
"calib/mu_w": 0.011864406779661017,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.005461909784690127,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.1830891304347826,
|
|
"calib/step_q_c_n": 506.0,
|
|
"calib/step_q_gap": -0.012006216628527883,
|
|
"calib/step_q_w": 0.19509534706331047,
|
|
"calib/step_q_w_n": 1311.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1163.0,
|
|
"completions/max_terminated_length": 1163.0,
|
|
"completions/mean_length": 320.6953125,
|
|
"completions/mean_terminated_length": 321.9529724121094,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 96.0,
|
|
"epoch": 0.17706666666666668,
|
|
"grad_norm": 0.01425196509808302,
|
|
"learning_rate": 9.444444444444445e-07,
|
|
"loss": 0.096,
|
|
"num_tokens": 34001416.0,
|
|
"reward": 1.0035477876663208,
|
|
"reward_std": 0.004095728974789381,
|
|
"rewards/accuracy_reward_step": 0.30859375,
|
|
"rewards/final_brier_reward_step": 0.6985019445419312,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 166
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8620406305417418,
|
|
"aux_distill/mean_u": 0.22939131249790382,
|
|
"aux_distill/n_active_tok": 208.0,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5423875432525952,
|
|
"calib/avg_num_step_conf": 6.5,
|
|
"calib/ece": 0.3211764705882353,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0010000000000000026,
|
|
"calib/mean_conf": 0.012156862745098041,
|
|
"calib/mu_c": 0.012823529411764708,
|
|
"calib/mu_w": 0.011823529411764705,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006043543140165663,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.22652014652014651,
|
|
"calib/step_q_c_n": 546.0,
|
|
"calib/step_q_gap": 0.0010997529602180534,
|
|
"calib/step_q_w": 0.22542039355992846,
|
|
"calib/step_q_w_n": 1118.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2255.0,
|
|
"completions/max_terminated_length": 2255.0,
|
|
"completions/mean_length": 297.59375,
|
|
"completions/mean_terminated_length": 297.59375,
|
|
"completions/min_length": 88.0,
|
|
"completions/min_terminated_length": 88.0,
|
|
"epoch": 0.17813333333333334,
|
|
"grad_norm": 0.012138527818024158,
|
|
"learning_rate": 9.166666666666666e-07,
|
|
"loss": 0.1287,
|
|
"num_tokens": 34183208.0,
|
|
"reward": 1.0002597570419312,
|
|
"reward_std": 0.01507105678319931,
|
|
"rewards/accuracy_reward_step": 0.33203125,
|
|
"rewards/final_brier_reward_step": 0.6723945140838623,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 167
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8371885055676103,
|
|
"aux_distill/mean_u": 0.25030672633272105,
|
|
"aux_distill/n_active_tok": 206.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5726744186046512,
|
|
"calib/avg_num_step_conf": 6.49609375,
|
|
"calib/ece": 0.31515625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.001960132890365448,
|
|
"calib/mean_conf": 0.012968750000000001,
|
|
"calib/mu_c": 0.014285714285714285,
|
|
"calib/mu_w": 0.012325581395348837,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.00653827373528365,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2508533472803347,
|
|
"calib/step_q_c_n": 478.0,
|
|
"calib/step_q_gap": 0.031394022385819925,
|
|
"calib/step_q_w": 0.21945932489451478,
|
|
"calib/step_q_w_n": 1185.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 740.0,
|
|
"completions/max_terminated_length": 740.0,
|
|
"completions/mean_length": 290.72265625,
|
|
"completions/mean_terminated_length": 291.8627624511719,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 62.0,
|
|
"epoch": 0.1792,
|
|
"grad_norm": 0.015005046501755714,
|
|
"learning_rate": 8.88888888888889e-07,
|
|
"loss": 0.0781,
|
|
"num_tokens": 34362305.0,
|
|
"reward": 1.0045819282531738,
|
|
"reward_std": 0.005785136483609676,
|
|
"rewards/accuracy_reward_step": 0.328125,
|
|
"rewards/final_brier_reward_step": 0.6810390949249268,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 168
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7752046966925263,
|
|
"aux_distill/mean_u": 0.17514847724133747,
|
|
"aux_distill/n_active_tok": 194.875,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5888518518518518,
|
|
"calib/avg_num_step_conf": 6.08984375,
|
|
"calib/ece": 0.28125490196078434,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.003122222222222224,
|
|
"calib/mean_conf": 0.012862745098039217,
|
|
"calib/mu_c": 0.015066666666666667,
|
|
"calib/mu_w": 0.011944444444444443,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006747605553063314,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23940054495912808,
|
|
"calib/step_q_c_n": 367.0,
|
|
"calib/step_q_gap": 0.023670679187316024,
|
|
"calib/step_q_w": 0.21572986577181205,
|
|
"calib/step_q_w_n": 1192.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 949.0,
|
|
"completions/max_terminated_length": 949.0,
|
|
"completions/mean_length": 279.140625,
|
|
"completions/mean_terminated_length": 280.2353210449219,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 80.0,
|
|
"epoch": 0.18026666666666666,
|
|
"grad_norm": 0.01291783805936575,
|
|
"learning_rate": 8.611111111111112e-07,
|
|
"loss": 0.0862,
|
|
"num_tokens": 34537949.0,
|
|
"reward": 1.0004026889801025,
|
|
"reward_std": 0.01507566962391138,
|
|
"rewards/accuracy_reward_step": 0.29296875,
|
|
"rewards/final_brier_reward_step": 0.7117429971694946,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 169
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7850721022114158,
|
|
"aux_distill/mean_u": 0.18490419322101734,
|
|
"aux_distill/n_active_tok": 212.125,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5295285186990738,
|
|
"calib/avg_num_step_conf": 6.65625,
|
|
"calib/ece": 0.311328125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.00017967825057455325,
|
|
"calib/mean_conf": 0.012890625000000001,
|
|
"calib/mu_c": 0.013012048192771086,
|
|
"calib/mu_w": 0.012832369942196532,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006392420285727073,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2407480496453901,
|
|
"calib/step_q_c_n": 564.0,
|
|
"calib/step_q_gap": 0.016790681224337467,
|
|
"calib/step_q_w": 0.22395736842105263,
|
|
"calib/step_q_w_n": 1140.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 813.0,
|
|
"completions/max_terminated_length": 813.0,
|
|
"completions/mean_length": 290.69140625,
|
|
"completions/mean_terminated_length": 291.8313903808594,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 86.0,
|
|
"epoch": 0.18133333333333335,
|
|
"grad_norm": 0.017970673739910126,
|
|
"learning_rate": 8.333333333333333e-07,
|
|
"loss": 0.087,
|
|
"num_tokens": 34716518.0,
|
|
"reward": 1.0041152238845825,
|
|
"reward_std": 0.004571193363517523,
|
|
"rewards/accuracy_reward_step": 0.32421875,
|
|
"rewards/final_brier_reward_step": 0.684011697769165,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 170
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7836430184543133,
|
|
"aux_distill/mean_u": 0.19953727431926524,
|
|
"aux_distill/n_active_tok": 214.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.593505386344261,
|
|
"calib/avg_num_step_conf": 6.796875,
|
|
"calib/ece": 0.2562890625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0032992327365728887,
|
|
"calib/mean_conf": 0.0132421875,
|
|
"calib/mu_c": 0.01565217391304348,
|
|
"calib/mu_w": 0.01235294117647059,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007291096982954193,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23558839779005522,
|
|
"calib/step_q_c_n": 362.0,
|
|
"calib/step_q_gap": 0.031814522608632906,
|
|
"calib/step_q_w": 0.20377387518142231,
|
|
"calib/step_q_w_n": 1378.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1211.0,
|
|
"completions/max_terminated_length": 1211.0,
|
|
"completions/mean_length": 295.1875,
|
|
"completions/mean_terminated_length": 296.3451232910156,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 89.0,
|
|
"epoch": 0.1824,
|
|
"grad_norm": 0.012783910147845745,
|
|
"learning_rate": 8.055555555555557e-07,
|
|
"loss": 0.0813,
|
|
"num_tokens": 34898982.0,
|
|
"reward": 1.004104495048523,
|
|
"reward_std": 0.005231703631579876,
|
|
"rewards/accuracy_reward_step": 0.26953125,
|
|
"rewards/final_brier_reward_step": 0.7386777400970459,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 171
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7730856984853745,
|
|
"aux_distill/mean_u": 0.16074557660169184,
|
|
"aux_distill/n_active_tok": 193.75,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5363810277722805,
|
|
"calib/avg_num_step_conf": 6.140625,
|
|
"calib/ece": 0.3488671875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0016801899861468433,
|
|
"calib/mean_conf": 0.014414062500000002,
|
|
"calib/mu_c": 0.015483870967741935,
|
|
"calib/mu_w": 0.013803680981595092,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008270832923357462,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.24181102362204723,
|
|
"calib/step_q_c_n": 508.0,
|
|
"calib/step_q_gap": 0.008519670238588573,
|
|
"calib/step_q_w": 0.23329135338345866,
|
|
"calib/step_q_w_n": 1064.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 950.0,
|
|
"completions/max_terminated_length": 950.0,
|
|
"completions/mean_length": 263.9140625,
|
|
"completions/mean_terminated_length": 264.94903564453125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 67.0,
|
|
"epoch": 0.18346666666666667,
|
|
"grad_norm": 0.01724671758711338,
|
|
"learning_rate": 7.777777777777779e-07,
|
|
"loss": 0.097,
|
|
"num_tokens": 35069896.0,
|
|
"reward": 1.0054869651794434,
|
|
"reward_std": 0.0069214459508657455,
|
|
"rewards/accuracy_reward_step": 0.36328125,
|
|
"rewards/final_brier_reward_step": 0.6476925611495972,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 172
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7590753780677915,
|
|
"aux_distill/mean_u": 0.18627376776313878,
|
|
"aux_distill/n_active_tok": 209.875,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.540965250965251,
|
|
"calib/avg_num_step_conf": 6.60546875,
|
|
"calib/ece": 0.2604705882352941,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0019150579150579161,
|
|
"calib/mean_conf": 0.01403921568627451,
|
|
"calib/mu_c": 0.01542857142857143,
|
|
"calib/mu_w": 0.013513513513513514,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.006899732260544964,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23605496828752645,
|
|
"calib/step_q_c_n": 473.0,
|
|
"calib/step_q_gap": 0.017361700635638105,
|
|
"calib/step_q_w": 0.21869326765188835,
|
|
"calib/step_q_w_n": 1218.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 695.0,
|
|
"completions/max_terminated_length": 695.0,
|
|
"completions/mean_length": 288.703125,
|
|
"completions/mean_terminated_length": 289.8352966308594,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 55.0,
|
|
"epoch": 0.18453333333333333,
|
|
"grad_norm": 0.014482683502137661,
|
|
"learning_rate": 7.5e-07,
|
|
"loss": 0.0799,
|
|
"num_tokens": 35246964.0,
|
|
"reward": 1.0001906156539917,
|
|
"reward_std": 0.016043417155742645,
|
|
"rewards/accuracy_reward_step": 0.2734375,
|
|
"rewards/final_brier_reward_step": 0.7308500409126282,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 173
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7893230896443129,
|
|
"aux_distill/mean_u": 0.20265565190850546,
|
|
"aux_distill/n_active_tok": 205.0,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.4901930501930501,
|
|
"calib/avg_num_step_conf": 6.77734375,
|
|
"calib/ece": 0.26050980392156864,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0003938223938223951,
|
|
"calib/mean_conf": 0.014,
|
|
"calib/mu_c": 0.014285714285714285,
|
|
"calib/mu_w": 0.01389189189189189,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.007545937746270388,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.14781926229508197,
|
|
"calib/step_q_c_n": 488.0,
|
|
"calib/step_q_gap": -0.08902917395191084,
|
|
"calib/step_q_w": 0.2368484362469928,
|
|
"calib/step_q_w_n": 1247.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1479.0,
|
|
"completions/max_terminated_length": 1479.0,
|
|
"completions/mean_length": 294.26953125,
|
|
"completions/mean_terminated_length": 295.4235534667969,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 81.0,
|
|
"epoch": 0.1856,
|
|
"grad_norm": 0.013904116116464138,
|
|
"learning_rate": 7.222222222222222e-07,
|
|
"loss": 0.0318,
|
|
"num_tokens": 35426529.0,
|
|
"reward": 0.9998739361763,
|
|
"reward_std": 0.01602732203900814,
|
|
"rewards/accuracy_reward_step": 0.2734375,
|
|
"rewards/final_brier_reward_step": 0.7302167415618896,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 174
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.81294996291399,
|
|
"aux_distill/mean_u": 0.19859697078451224,
|
|
"aux_distill/n_active_tok": 218.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5849780701754386,
|
|
"calib/avg_num_step_conf": 7.140625,
|
|
"calib/ece": 0.28203125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.00321637426900585,
|
|
"calib/mean_conf": 0.01484375,
|
|
"calib/mu_c": 0.017105263157894738,
|
|
"calib/mu_w": 0.013888888888888888,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008522431339559152,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.18974770642201833,
|
|
"calib/step_q_c_n": 436.0,
|
|
"calib/step_q_gap": -0.014610195876832266,
|
|
"calib/step_q_w": 0.2043579022988506,
|
|
"calib/step_q_w_n": 1392.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 957.0,
|
|
"completions/max_terminated_length": 957.0,
|
|
"completions/mean_length": 302.40625,
|
|
"completions/mean_terminated_length": 303.5921630859375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 72.0,
|
|
"epoch": 0.18666666666666668,
|
|
"grad_norm": 0.01465411577373743,
|
|
"learning_rate": 6.944444444444446e-07,
|
|
"loss": 0.0535,
|
|
"num_tokens": 35609769.0,
|
|
"reward": 1.0049315690994263,
|
|
"reward_std": 0.005668940953910351,
|
|
"rewards/accuracy_reward_step": 0.296875,
|
|
"rewards/final_brier_reward_step": 0.7129883170127869,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 175
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8950878940522671,
|
|
"aux_distill/mean_u": 0.26817866945524593,
|
|
"aux_distill/n_active_tok": 227.375,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.6145161290322582,
|
|
"calib/avg_num_step_conf": 7.13671875,
|
|
"calib/ece": 0.259140625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0024485407066052265,
|
|
"calib/mean_conf": 0.015078125000000001,
|
|
"calib/mu_c": 0.01685714285714286,
|
|
"calib/mu_w": 0.014408602150537632,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.000390625,
|
|
"calib/std_conf": 0.009099802002481977,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.18709090909090909,
|
|
"calib/step_q_c_n": 385.0,
|
|
"calib/step_q_gap": 0.020935361240701056,
|
|
"calib/step_q_w": 0.16615554785020803,
|
|
"calib/step_q_w_n": 1442.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1162.0,
|
|
"completions/max_terminated_length": 1162.0,
|
|
"completions/mean_length": 310.3046875,
|
|
"completions/mean_terminated_length": 311.5215759277344,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 84.0,
|
|
"epoch": 0.18773333333333334,
|
|
"grad_norm": 0.013478383421897888,
|
|
"learning_rate": 6.666666666666667e-07,
|
|
"loss": 0.0958,
|
|
"num_tokens": 35793271.0,
|
|
"reward": 0.9966421723365784,
|
|
"reward_std": 0.027028311043977737,
|
|
"rewards/accuracy_reward_step": 0.2734375,
|
|
"rewards/final_brier_reward_step": 0.7276594042778015,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 176
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7781459763646126,
|
|
"aux_distill/mean_u": 0.2034369505524209,
|
|
"aux_distill/n_active_tok": 222.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5171164772727271,
|
|
"calib/avg_num_step_conf": 7.0234375,
|
|
"calib/ece": 0.29917968749999996,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.003443181818181821,
|
|
"calib/mean_conf": 0.0188671875,
|
|
"calib/mu_c": 0.016499999999999997,
|
|
"calib/mu_w": 0.019943181818181818,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0027734375,
|
|
"calib/std_conf": 0.04424591631823036,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.16645539906103285,
|
|
"calib/step_q_c_n": 426.0,
|
|
"calib/step_q_gap": -0.020672880822349082,
|
|
"calib/step_q_w": 0.18712827988338193,
|
|
"calib/step_q_w_n": 1372.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1444.0,
|
|
"completions/max_terminated_length": 1444.0,
|
|
"completions/mean_length": 301.12109375,
|
|
"completions/mean_terminated_length": 302.3019714355469,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 91.0,
|
|
"epoch": 0.1888,
|
|
"grad_norm": 0.013484400697052479,
|
|
"learning_rate": 6.388888888888889e-07,
|
|
"loss": 0.068,
|
|
"num_tokens": 35974190.0,
|
|
"reward": 1.0039993524551392,
|
|
"reward_std": 0.007987409830093384,
|
|
"rewards/accuracy_reward_step": 0.3125,
|
|
"rewards/final_brier_reward_step": 0.6954988241195679,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 177
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8372670179232955,
|
|
"aux_distill/mean_u": 0.22859842521539914,
|
|
"aux_distill/n_active_tok": 221.375,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5706492637215529,
|
|
"calib/avg_num_step_conf": 6.98046875,
|
|
"calib/ece": 0.33625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.003116465863453816,
|
|
"calib/mean_conf": 0.015312500000000001,
|
|
"calib/mu_c": 0.017333333333333333,
|
|
"calib/mu_w": 0.014216867469879517,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008833308765689106,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.21936416184971097,
|
|
"calib/step_q_c_n": 519.0,
|
|
"calib/step_q_gap": -0.012766910705494072,
|
|
"calib/step_q_w": 0.23213107255520504,
|
|
"calib/step_q_w_n": 1268.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1080.0,
|
|
"completions/max_terminated_length": 1080.0,
|
|
"completions/mean_length": 300.6171875,
|
|
"completions/mean_terminated_length": 301.79608154296875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 70.0,
|
|
"epoch": 0.18986666666666666,
|
|
"grad_norm": 0.014162966050207615,
|
|
"learning_rate": 6.111111111111112e-07,
|
|
"loss": 0.065,
|
|
"num_tokens": 36157220.0,
|
|
"reward": 1.0059374570846558,
|
|
"reward_std": 0.007233859039843082,
|
|
"rewards/accuracy_reward_step": 0.3515625,
|
|
"rewards/final_brier_reward_step": 0.6603125333786011,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 178
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8423464596271515,
|
|
"aux_distill/mean_u": 0.21650959811507547,
|
|
"aux_distill/n_active_tok": 205.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5967660616537022,
|
|
"calib/avg_num_step_conf": 6.47265625,
|
|
"calib/ece": 0.289609375,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.002469029098242587,
|
|
"calib/mean_conf": 0.015078125000000001,
|
|
"calib/mu_c": 0.0167948717948718,
|
|
"calib/mu_w": 0.014325842696629213,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.008052182715535892,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.17574938574938576,
|
|
"calib/step_q_c_n": 407.0,
|
|
"calib/step_q_gap": -0.06633861425061424,
|
|
"calib/step_q_w": 0.242088,
|
|
"calib/step_q_w_n": 1250.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1040.0,
|
|
"completions/max_terminated_length": 1040.0,
|
|
"completions/mean_length": 285.65234375,
|
|
"completions/mean_terminated_length": 286.7725524902344,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 64.0,
|
|
"epoch": 0.19093333333333334,
|
|
"grad_norm": 0.014211541041731834,
|
|
"learning_rate": 5.833333333333334e-07,
|
|
"loss": 0.0594,
|
|
"num_tokens": 36336611.0,
|
|
"reward": 1.0049710273742676,
|
|
"reward_std": 0.004556154832243919,
|
|
"rewards/accuracy_reward_step": 0.3046875,
|
|
"rewards/final_brier_reward_step": 0.7052546739578247,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 179
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.801798083819449,
|
|
"aux_distill/mean_u": 0.22506361019896942,
|
|
"aux_distill/n_active_tok": 239.375,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.5592140921409214,
|
|
"calib/avg_num_step_conf": 7.6953125,
|
|
"calib/ece": 0.3385433070866142,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0018780487804878065,
|
|
"calib/mean_conf": 0.01578740157480315,
|
|
"calib/mu_c": 0.017,
|
|
"calib/mu_w": 0.015121951219512195,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.00878270483802163,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.21153817504655495,
|
|
"calib/step_q_c_n": 537.0,
|
|
"calib/step_q_gap": 0.00631151768437771,
|
|
"calib/step_q_w": 0.20522665736217724,
|
|
"calib/step_q_w_n": 1433.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2826.0,
|
|
"completions/max_terminated_length": 2826.0,
|
|
"completions/mean_length": 345.01953125,
|
|
"completions/mean_terminated_length": 346.37255859375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 69.0,
|
|
"epoch": 0.192,
|
|
"grad_norm": 0.015591487288475037,
|
|
"learning_rate": 5.555555555555555e-07,
|
|
"loss": 0.0698,
|
|
"num_tokens": 36528792.0,
|
|
"reward": 0.9980021119117737,
|
|
"reward_std": 0.028464868664741516,
|
|
"rewards/accuracy_reward_step": 0.3515625,
|
|
"rewards/final_brier_reward_step": 0.6522542834281921,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 180
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7981845401227474,
|
|
"aux_distill/mean_u": 0.1806097187679103,
|
|
"aux_distill/n_active_tok": 202.0,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5448807854137446,
|
|
"calib/avg_num_step_conf": 6.3125,
|
|
"calib/ece": 0.25356862745098036,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.001105656848994857,
|
|
"calib/mean_conf": 0.017019607843137254,
|
|
"calib/mu_c": 0.01782608695652174,
|
|
"calib/mu_w": 0.016720430107526885,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.011804768388619833,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.21485241730279897,
|
|
"calib/step_q_c_n": 393.0,
|
|
"calib/step_q_gap": -0.025214385640782377,
|
|
"calib/step_q_w": 0.24006680294358135,
|
|
"calib/step_q_w_n": 1223.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1254.0,
|
|
"completions/max_terminated_length": 1254.0,
|
|
"completions/mean_length": 264.6796875,
|
|
"completions/mean_terminated_length": 265.7176513671875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 82.0,
|
|
"epoch": 0.19306666666666666,
|
|
"grad_norm": 0.01583191752433777,
|
|
"learning_rate": 5.277777777777779e-07,
|
|
"loss": 0.0544,
|
|
"num_tokens": 36702814.0,
|
|
"reward": 1.0006847381591797,
|
|
"reward_std": 0.01699506677687168,
|
|
"rewards/accuracy_reward_step": 0.26953125,
|
|
"rewards/final_brier_reward_step": 0.7357445359230042,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 181
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7627274738624692,
|
|
"aux_distill/mean_u": 0.1822232042934428,
|
|
"aux_distill/n_active_tok": 227.75,
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/auroc": 0.5438858893915074,
|
|
"calib/avg_num_step_conf": 7.1171875,
|
|
"calib/ece": 0.28545098039215683,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.001837151612432513,
|
|
"calib/mean_conf": 0.016509803921568627,
|
|
"calib/mu_c": 0.017792207792207793,
|
|
"calib/mu_w": 0.01595505617977528,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.00965964426946115,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22081318681318682,
|
|
"calib/step_q_c_n": 455.0,
|
|
"calib/step_q_gap": -0.004984106529900223,
|
|
"calib/step_q_w": 0.22579729334308704,
|
|
"calib/step_q_w_n": 1367.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 3029.0,
|
|
"completions/max_terminated_length": 3029.0,
|
|
"completions/mean_length": 321.87109375,
|
|
"completions/mean_terminated_length": 321.87109375,
|
|
"completions/min_length": 75.0,
|
|
"completions/min_terminated_length": 75.0,
|
|
"epoch": 0.19413333333333332,
|
|
"grad_norm": 0.012693443335592747,
|
|
"learning_rate": 5.000000000000001e-07,
|
|
"loss": 0.1194,
|
|
"num_tokens": 36891373.0,
|
|
"reward": 0.9973569512367249,
|
|
"reward_std": 0.02882469817996025,
|
|
"rewards/accuracy_reward_step": 0.30078125,
|
|
"rewards/final_brier_reward_step": 0.701745331287384,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 182
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7574250753968954,
|
|
"aux_distill/mean_u": 0.1943297293494716,
|
|
"aux_distill/n_active_tok": 228.25,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.4795465249276728,
|
|
"calib/avg_num_step_conf": 7.21484375,
|
|
"calib/ece": 0.3306640625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0009008948395344135,
|
|
"calib/mean_conf": 0.016992187500000002,
|
|
"calib/mu_c": 0.01640449438202247,
|
|
"calib/mu_w": 0.017305389221556885,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.009799550447078873,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.18748726655348047,
|
|
"calib/step_q_c_n": 589.0,
|
|
"calib/step_q_gap": -0.06748069847036686,
|
|
"calib/step_q_w": 0.25496796502384733,
|
|
"calib/step_q_w_n": 1258.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 924.0,
|
|
"completions/max_terminated_length": 924.0,
|
|
"completions/mean_length": 306.62890625,
|
|
"completions/mean_terminated_length": 307.8313903808594,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 95.0,
|
|
"epoch": 0.1952,
|
|
"grad_norm": 0.015803849324584007,
|
|
"learning_rate": 4.7222222222222226e-07,
|
|
"loss": 0.0647,
|
|
"num_tokens": 37076550.0,
|
|
"reward": 1.005510687828064,
|
|
"reward_std": 0.006886619608849287,
|
|
"rewards/accuracy_reward_step": 0.34765625,
|
|
"rewards/final_brier_reward_step": 0.6633652448654175,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 183
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8180130543187261,
|
|
"aux_distill/mean_u": 0.21504098794632884,
|
|
"aux_distill/n_active_tok": 213.0,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.4544132254860223,
|
|
"calib/avg_num_step_conf": 6.69140625,
|
|
"calib/ece": 0.2993725490196078,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0023563218390804594,
|
|
"calib/mean_conf": 0.01827450980392157,
|
|
"calib/mu_c": 0.016666666666666666,
|
|
"calib/mu_w": 0.019022988505747126,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.01148584539194007,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.18374944852941177,
|
|
"calib/step_q_c_n": 544.0,
|
|
"calib/step_q_gap": -0.015600423155789245,
|
|
"calib/step_q_w": 0.19934987168520102,
|
|
"calib/step_q_w_n": 1169.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 697.0,
|
|
"completions/max_terminated_length": 697.0,
|
|
"completions/mean_length": 295.5859375,
|
|
"completions/mean_terminated_length": 296.7451171875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 79.0,
|
|
"epoch": 0.19626666666666667,
|
|
"grad_norm": 0.015823913738131523,
|
|
"learning_rate": 4.444444444444445e-07,
|
|
"loss": 0.0709,
|
|
"num_tokens": 37257500.0,
|
|
"reward": 1.0011351108551025,
|
|
"reward_std": 0.01635909266769886,
|
|
"rewards/accuracy_reward_step": 0.31640625,
|
|
"rewards/final_brier_reward_step": 0.6897703409194946,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 184
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7792685413733125,
|
|
"aux_distill/mean_u": 0.19762557337851924,
|
|
"aux_distill/n_active_tok": 250.0,
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/auroc": 0.5308478038815118,
|
|
"calib/avg_num_step_conf": 7.85546875,
|
|
"calib/ece": 0.33318897637795275,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0010166836908409964,
|
|
"calib/mean_conf": 0.01720472440944882,
|
|
"calib/mu_c": 0.017865168539325845,
|
|
"calib/mu_w": 0.01684848484848485,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.010408081795236494,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.20600725952813068,
|
|
"calib/step_q_c_n": 551.0,
|
|
"calib/step_q_gap": 0.009205889665116956,
|
|
"calib/step_q_w": 0.19680136986301372,
|
|
"calib/step_q_w_n": 1460.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2516.0,
|
|
"completions/max_terminated_length": 2516.0,
|
|
"completions/mean_length": 326.39453125,
|
|
"completions/mean_terminated_length": 327.6745300292969,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 94.0,
|
|
"epoch": 0.19733333333333333,
|
|
"grad_norm": 0.014444109983742237,
|
|
"learning_rate": 4.1666666666666667e-07,
|
|
"loss": 0.103,
|
|
"num_tokens": 37447977.0,
|
|
"reward": 0.9942923188209534,
|
|
"reward_std": 0.039944350719451904,
|
|
"rewards/accuracy_reward_step": 0.34765625,
|
|
"rewards/final_brier_reward_step": 0.6526472568511963,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 185
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8481473177671432,
|
|
"aux_distill/mean_u": 0.2418076824490043,
|
|
"aux_distill/n_active_tok": 233.875,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.6300784533482768,
|
|
"calib/avg_num_step_conf": 7.30859375,
|
|
"calib/ece": 0.30854901960784314,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.005070748108713923,
|
|
"calib/mean_conf": 0.016941176470588237,
|
|
"calib/mu_c": 0.02036144578313253,
|
|
"calib/mu_w": 0.015290697674418607,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.01029260493087031,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2827586206896551,
|
|
"calib/step_q_c_n": 464.0,
|
|
"calib/step_q_gap": 0.10086096610543338,
|
|
"calib/step_q_w": 0.18189765458422175,
|
|
"calib/step_q_w_n": 1407.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1993.0,
|
|
"completions/max_terminated_length": 1993.0,
|
|
"completions/mean_length": 324.77734375,
|
|
"completions/mean_terminated_length": 326.0509948730469,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 86.0,
|
|
"epoch": 0.1984,
|
|
"grad_norm": 0.014936697669327259,
|
|
"learning_rate": 3.8888888888888895e-07,
|
|
"loss": 0.0918,
|
|
"num_tokens": 37636160.0,
|
|
"reward": 1.0024995803833008,
|
|
"reward_std": 0.018394682556390762,
|
|
"rewards/accuracy_reward_step": 0.32421875,
|
|
"rewards/final_brier_reward_step": 0.6846867799758911,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 186
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7872933838516474,
|
|
"aux_distill/mean_u": 0.18320702169394124,
|
|
"aux_distill/n_active_tok": 254.375,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5531104697771364,
|
|
"calib/avg_num_step_conf": 7.94921875,
|
|
"calib/ece": 0.24219607843137259,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0018927368927368939,
|
|
"calib/mean_conf": 0.01662745098039216,
|
|
"calib/mu_c": 0.018030303030303032,
|
|
"calib/mu_w": 0.016137566137566138,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.009761791832292309,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22751269035532995,
|
|
"calib/step_q_c_n": 394.0,
|
|
"calib/step_q_gap": 0.0014950182042025861,
|
|
"calib/step_q_w": 0.22601767215112736,
|
|
"calib/step_q_w_n": 1641.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/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": 343.58984375,
|
|
"completions/mean_terminated_length": 343.58984375,
|
|
"completions/min_length": 69.0,
|
|
"completions/min_terminated_length": 69.0,
|
|
"epoch": 0.19946666666666665,
|
|
"grad_norm": 0.012962001375854015,
|
|
"learning_rate": 3.611111111111111e-07,
|
|
"loss": 0.1133,
|
|
"num_tokens": 37825663.0,
|
|
"reward": 1.0005570650100708,
|
|
"reward_std": 0.016831982880830765,
|
|
"rewards/accuracy_reward_step": 0.2578125,
|
|
"rewards/final_brier_reward_step": 0.7472078800201416,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 187
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8266328135505319,
|
|
"aux_distill/mean_u": 0.23075207178941026,
|
|
"aux_distill/n_active_tok": 220.125,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5012258335668254,
|
|
"calib/avg_num_step_conf": 7.19140625,
|
|
"calib/ece": 0.30701960784313725,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0015529560100868596,
|
|
"calib/mean_conf": 0.01847058823529412,
|
|
"calib/mu_c": 0.019518072289156627,
|
|
"calib/mu_w": 0.017965116279069767,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.012971088425729619,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.15902970297029703,
|
|
"calib/step_q_c_n": 505.0,
|
|
"calib/step_q_gap": -0.0739418539159305,
|
|
"calib/step_q_w": 0.23297155688622753,
|
|
"calib/step_q_w_n": 1336.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1360.0,
|
|
"completions/max_terminated_length": 1360.0,
|
|
"completions/mean_length": 315.6171875,
|
|
"completions/mean_terminated_length": 316.85491943359375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 89.0,
|
|
"epoch": 0.20053333333333334,
|
|
"grad_norm": 0.01490770187228918,
|
|
"learning_rate": 3.3333333333333335e-07,
|
|
"loss": 0.0597,
|
|
"num_tokens": 38010533.0,
|
|
"reward": 1.0021681785583496,
|
|
"reward_std": 0.019167276099324226,
|
|
"rewards/accuracy_reward_step": 0.32421875,
|
|
"rewards/final_brier_reward_step": 0.6840238571166992,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 188
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.785682198125869,
|
|
"aux_distill/mean_u": 0.20517065482001212,
|
|
"aux_distill/n_active_tok": 218.125,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5821377840909091,
|
|
"calib/avg_num_step_conf": 6.87890625,
|
|
"calib/ece": 0.2951171875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0034431818181818175,
|
|
"calib/mean_conf": 0.0173828125,
|
|
"calib/mu_c": 0.01975,
|
|
"calib/mu_w": 0.016306818181818183,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.011849723819137885,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22650107991360688,
|
|
"calib/step_q_c_n": 463.0,
|
|
"calib/step_q_gap": 0.03564668854226635,
|
|
"calib/step_q_w": 0.19085439137134053,
|
|
"calib/step_q_w_n": 1298.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 853.0,
|
|
"completions/max_terminated_length": 853.0,
|
|
"completions/mean_length": 294.375,
|
|
"completions/mean_terminated_length": 295.5294189453125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 76.0,
|
|
"epoch": 0.2016,
|
|
"grad_norm": 0.014363201335072517,
|
|
"learning_rate": 3.055555555555556e-07,
|
|
"loss": 0.0937,
|
|
"num_tokens": 38193661.0,
|
|
"reward": 1.0059505701065063,
|
|
"reward_std": 0.00714213028550148,
|
|
"rewards/accuracy_reward_step": 0.3125,
|
|
"rewards/final_brier_reward_step": 0.6994011402130127,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 189
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8067413792014122,
|
|
"aux_distill/mean_u": 0.1998867498130261,
|
|
"aux_distill/n_active_tok": 258.125,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5696138996138996,
|
|
"calib/avg_num_step_conf": 8.06640625,
|
|
"calib/ece": 0.25670588235294123,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.002830115830115835,
|
|
"calib/mean_conf": 0.01780392156862745,
|
|
"calib/mu_c": 0.01985714285714286,
|
|
"calib/mu_w": 0.017027027027027027,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.010246031480133552,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.21214574898785424,
|
|
"calib/step_q_c_n": 494.0,
|
|
"calib/step_q_gap": -0.022446230642954146,
|
|
"calib/step_q_w": 0.23459197963080838,
|
|
"calib/step_q_w_n": 1571.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1483.0,
|
|
"completions/max_terminated_length": 1483.0,
|
|
"completions/mean_length": 345.1640625,
|
|
"completions/mean_terminated_length": 346.5176696777344,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 59.0,
|
|
"epoch": 0.20266666666666666,
|
|
"grad_norm": 0.012308570556342602,
|
|
"learning_rate": 2.7777777777777776e-07,
|
|
"loss": 0.0772,
|
|
"num_tokens": 38387631.0,
|
|
"reward": 0.9974077939987183,
|
|
"reward_std": 0.028176523745059967,
|
|
"rewards/accuracy_reward_step": 0.2734375,
|
|
"rewards/final_brier_reward_step": 0.7291905879974365,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 190
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7780831595882773,
|
|
"aux_distill/mean_u": 0.18509156854576636,
|
|
"aux_distill/n_active_tok": 242.625,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5230202578268878,
|
|
"calib/avg_num_step_conf": 7.640625,
|
|
"calib/ece": 0.275546875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0004405156537753245,
|
|
"calib/mean_conf": 0.017421875,
|
|
"calib/mu_c": 0.017733333333333334,
|
|
"calib/mu_w": 0.01729281767955801,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.010513837143706146,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23987234042553193,
|
|
"calib/step_q_c_n": 470.0,
|
|
"calib/step_q_gap": 0.03117785859511471,
|
|
"calib/step_q_w": 0.20869448183041722,
|
|
"calib/step_q_w_n": 1486.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1188.0,
|
|
"completions/max_terminated_length": 1188.0,
|
|
"completions/mean_length": 322.1484375,
|
|
"completions/mean_terminated_length": 323.4117736816406,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 90.0,
|
|
"epoch": 0.20373333333333332,
|
|
"grad_norm": 0.014701644890010357,
|
|
"learning_rate": 2.5000000000000004e-07,
|
|
"loss": 0.0938,
|
|
"num_tokens": 38574269.0,
|
|
"reward": 1.004988193511963,
|
|
"reward_std": 0.005673405714333057,
|
|
"rewards/accuracy_reward_step": 0.29296875,
|
|
"rewards/final_brier_reward_step": 0.7170077562332153,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 191
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.7313472367823124,
|
|
"aux_distill/mean_u": 0.17417473060254704,
|
|
"aux_distill/n_active_tok": 232.875,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.612087912087912,
|
|
"calib/avg_num_step_conf": 7.46484375,
|
|
"calib/ece": 0.336953125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.006053946053946047,
|
|
"calib/mean_conf": 0.018515625000000004,
|
|
"calib/mu_c": 0.02241758241758241,
|
|
"calib/mu_w": 0.016363636363636365,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.01349709712713719,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.17961389961389962,
|
|
"calib/step_q_c_n": 518.0,
|
|
"calib/step_q_gap": -0.03224044353039329,
|
|
"calib/step_q_w": 0.2118543431442929,
|
|
"calib/step_q_w_n": 1393.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1295.0,
|
|
"completions/max_terminated_length": 1295.0,
|
|
"completions/mean_length": 321.51171875,
|
|
"completions/mean_terminated_length": 322.7725830078125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 93.0,
|
|
"epoch": 0.2048,
|
|
"grad_norm": 0.016461456194519997,
|
|
"learning_rate": 2.2222222222222224e-07,
|
|
"loss": 0.0852,
|
|
"num_tokens": 38761552.0,
|
|
"reward": 1.0077061653137207,
|
|
"reward_std": 0.008474325761198997,
|
|
"rewards/accuracy_reward_step": 0.35546875,
|
|
"rewards/final_brier_reward_step": 0.659943699836731,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 192
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8442441988736391,
|
|
"aux_distill/mean_u": 0.21636011236901698,
|
|
"aux_distill/n_active_tok": 231.125,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.49377264492753636,
|
|
"calib/avg_num_step_conf": 7.33203125,
|
|
"calib/ece": 0.2634375,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0008212560386473382,
|
|
"calib/mean_conf": 0.017812500000000002,
|
|
"calib/mu_c": 0.017222222222222226,
|
|
"calib/mu_w": 0.018043478260869564,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.01185442296149416,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.24503211991434687,
|
|
"calib/step_q_c_n": 467.0,
|
|
"calib/step_q_gap": 0.016645027715765315,
|
|
"calib/step_q_w": 0.22838709219858155,
|
|
"calib/step_q_w_n": 1410.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1001.0,
|
|
"completions/max_terminated_length": 1001.0,
|
|
"completions/mean_length": 314.51953125,
|
|
"completions/mean_terminated_length": 315.7529602050781,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 69.0,
|
|
"epoch": 0.20586666666666667,
|
|
"grad_norm": 0.018526718020439148,
|
|
"learning_rate": 1.9444444444444447e-07,
|
|
"loss": 0.0655,
|
|
"num_tokens": 38947781.0,
|
|
"reward": 1.000710368156433,
|
|
"reward_std": 0.017061732709407806,
|
|
"rewards/accuracy_reward_step": 0.28125,
|
|
"rewards/final_brier_reward_step": 0.7240769863128662,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 193
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8148846421390772,
|
|
"aux_distill/mean_u": 0.21376291768715036,
|
|
"aux_distill/n_active_tok": 208.75,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5002201257861636,
|
|
"calib/avg_num_step_conf": 6.5390625,
|
|
"calib/ece": 0.3948046875,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -2.138364779874155e-05,
|
|
"calib/mean_conf": 0.019257812500000006,
|
|
"calib/mu_c": 0.019245283018867926,
|
|
"calib/mu_w": 0.019266666666666668,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.011483511340824449,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.23845024469820558,
|
|
"calib/step_q_c_n": 613.0,
|
|
"calib/step_q_gap": 0.0560185764606938,
|
|
"calib/step_q_w": 0.18243166823751178,
|
|
"calib/step_q_w_n": 1061.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 824.0,
|
|
"completions/max_terminated_length": 824.0,
|
|
"completions/mean_length": 281.90625,
|
|
"completions/mean_terminated_length": 283.01177978515625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 81.0,
|
|
"epoch": 0.20693333333333333,
|
|
"grad_norm": 0.018510188907384872,
|
|
"learning_rate": 1.6666666666666668e-07,
|
|
"loss": 0.0671,
|
|
"num_tokens": 39125893.0,
|
|
"reward": 1.003811240196228,
|
|
"reward_std": 0.01898595690727234,
|
|
"rewards/accuracy_reward_step": 0.4140625,
|
|
"rewards/final_brier_reward_step": 0.5974664092063904,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 194
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8549988893792033,
|
|
"aux_distill/mean_u": 0.248499681289213,
|
|
"aux_distill/n_active_tok": 221.25,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5184979358526516,
|
|
"calib/avg_num_step_conf": 7.125,
|
|
"calib/ece": 0.24529411764705883,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0006232137186408387,
|
|
"calib/mean_conf": 0.017450980392156864,
|
|
"calib/mu_c": 0.01791044776119403,
|
|
"calib/mu_w": 0.017287234042553192,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.010854318243400978,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.18694690265486727,
|
|
"calib/step_q_c_n": 452.0,
|
|
"calib/step_q_gap": -0.028434511339301805,
|
|
"calib/step_q_w": 0.21538141399416907,
|
|
"calib/step_q_w_n": 1372.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1769.0,
|
|
"completions/max_terminated_length": 1769.0,
|
|
"completions/mean_length": 318.50390625,
|
|
"completions/mean_terminated_length": 319.7529602050781,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 81.0,
|
|
"epoch": 0.208,
|
|
"grad_norm": 0.01629587449133396,
|
|
"learning_rate": 1.3888888888888888e-07,
|
|
"loss": 0.0841,
|
|
"num_tokens": 39313414.0,
|
|
"reward": 1.0005707740783691,
|
|
"reward_std": 0.016977444291114807,
|
|
"rewards/accuracy_reward_step": 0.26171875,
|
|
"rewards/final_brier_reward_step": 0.7433292865753174,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 195
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.832436092197895,
|
|
"aux_distill/mean_u": 0.22138925953920585,
|
|
"aux_distill/n_active_tok": 210.125,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5842329545454547,
|
|
"calib/avg_num_step_conf": 6.578125,
|
|
"calib/ece": 0.2933203125,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0028295454545454526,
|
|
"calib/mean_conf": 0.0191796875,
|
|
"calib/mu_c": 0.021124999999999998,
|
|
"calib/mu_w": 0.018295454545454545,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.011062276547001695,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22217864923747274,
|
|
"calib/step_q_c_n": 459.0,
|
|
"calib/step_q_gap": 0.014548690053799274,
|
|
"calib/step_q_w": 0.20762995918367347,
|
|
"calib/step_q_w_n": 1225.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 810.0,
|
|
"completions/max_terminated_length": 810.0,
|
|
"completions/mean_length": 274.28125,
|
|
"completions/mean_terminated_length": 275.35687255859375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 56.0,
|
|
"epoch": 0.20906666666666668,
|
|
"grad_norm": 0.015408625826239586,
|
|
"learning_rate": 1.1111111111111112e-07,
|
|
"loss": 0.0717,
|
|
"num_tokens": 39486174.0,
|
|
"reward": 1.0063564777374268,
|
|
"reward_std": 0.00686432421207428,
|
|
"rewards/accuracy_reward_step": 0.3125,
|
|
"rewards/final_brier_reward_step": 0.7002128958702087,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 196
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8513198988512158,
|
|
"aux_distill/mean_u": 0.2619026741844905,
|
|
"aux_distill/n_active_tok": 218.375,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.5108518518518519,
|
|
"calib/avg_num_step_conf": 6.82421875,
|
|
"calib/ece": 0.27505882352941174,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0003666666666666714,
|
|
"calib/mean_conf": 0.01905882352941177,
|
|
"calib/mu_c": 0.018799999999999997,
|
|
"calib/mu_w": 0.01916666666666667,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.012394514543651142,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.18097387173396673,
|
|
"calib/step_q_c_n": 421.0,
|
|
"calib/step_q_gap": -0.052125675777345504,
|
|
"calib/step_q_w": 0.23309954751131223,
|
|
"calib/step_q_w_n": 1326.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2771.0,
|
|
"completions/max_terminated_length": 2771.0,
|
|
"completions/mean_length": 292.44921875,
|
|
"completions/mean_terminated_length": 292.44921875,
|
|
"completions/min_length": 72.0,
|
|
"completions/min_terminated_length": 72.0,
|
|
"epoch": 0.21013333333333334,
|
|
"grad_norm": 0.016117248684167862,
|
|
"learning_rate": 8.333333333333334e-08,
|
|
"loss": 0.1364,
|
|
"num_tokens": 39666097.0,
|
|
"reward": 1.001344084739685,
|
|
"reward_std": 0.017887791618704796,
|
|
"rewards/accuracy_reward_step": 0.29296875,
|
|
"rewards/final_brier_reward_step": 0.7136257886886597,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 197
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8347745090723038,
|
|
"aux_distill/mean_u": 0.18832998660862468,
|
|
"aux_distill/n_active_tok": 217.125,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5501277139208173,
|
|
"calib/avg_num_step_conf": 6.80859375,
|
|
"calib/ece": 0.2989019607843137,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0016581524052788413,
|
|
"calib/mean_conf": 0.01874509803921569,
|
|
"calib/mu_c": 0.019876543209876543,
|
|
"calib/mu_w": 0.018218390804597702,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.01151272635510995,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.20193293885601576,
|
|
"calib/step_q_c_n": 507.0,
|
|
"calib/step_q_gap": -0.0051042779724632115,
|
|
"calib/step_q_w": 0.20703721682847898,
|
|
"calib/step_q_w_n": 1236.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 815.0,
|
|
"completions/max_terminated_length": 815.0,
|
|
"completions/mean_length": 284.28125,
|
|
"completions/mean_terminated_length": 285.3960876464844,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 69.0,
|
|
"epoch": 0.2112,
|
|
"grad_norm": 0.014969132840633392,
|
|
"learning_rate": 5.555555555555556e-08,
|
|
"loss": 0.0785,
|
|
"num_tokens": 39844257.0,
|
|
"reward": 1.0040948390960693,
|
|
"reward_std": 0.012611201032996178,
|
|
"rewards/accuracy_reward_step": 0.3203125,
|
|
"rewards/final_brier_reward_step": 0.691783607006073,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 198
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8050779681652784,
|
|
"aux_distill/mean_u": 0.22997851558929855,
|
|
"aux_distill/n_active_tok": 214.625,
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/auroc": 0.5454513596860108,
|
|
"calib/avg_num_step_conf": 6.7578125,
|
|
"calib/ece": 0.301015625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0015727502102607198,
|
|
"calib/mean_conf": 0.019296875,
|
|
"calib/mu_c": 0.020365853658536586,
|
|
"calib/mu_w": 0.018793103448275866,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.012448870038456301,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.22131524008350728,
|
|
"calib/step_q_c_n": 479.0,
|
|
"calib/step_q_gap": 0.0123304279332275,
|
|
"calib/step_q_w": 0.20898481215027978,
|
|
"calib/step_q_w_n": 1251.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 943.0,
|
|
"completions/max_terminated_length": 943.0,
|
|
"completions/mean_length": 299.14453125,
|
|
"completions/mean_terminated_length": 300.3176574707031,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 61.0,
|
|
"epoch": 0.21226666666666666,
|
|
"grad_norm": 0.016602490097284317,
|
|
"learning_rate": 2.777777777777778e-08,
|
|
"loss": 0.1056,
|
|
"num_tokens": 40025038.0,
|
|
"reward": 1.0062596797943115,
|
|
"reward_std": 0.008099588565528393,
|
|
"rewards/accuracy_reward_step": 0.3203125,
|
|
"rewards/final_brier_reward_step": 0.6922070384025574,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 199
|
|
},
|
|
{
|
|
"aux_distill/lambda": 0.10000000000000005,
|
|
"aux_distill/loss": 0.8087783623486757,
|
|
"aux_distill/mean_u": 0.21598880375647106,
|
|
"aux_distill/n_active_tok": 220.375,
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/auroc": 0.6062042705689114,
|
|
"calib/avg_num_step_conf": 6.88671875,
|
|
"calib/ece": 0.27082352941176474,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.006024339256383453,
|
|
"calib/mean_conf": 0.019372549019607846,
|
|
"calib/mu_c": 0.02364864864864865,
|
|
"calib/mu_w": 0.017624309392265197,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.013988736062981313,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.23423076923076921,
|
|
"calib/step_q_c_n": 416.0,
|
|
"calib/step_q_gap": 0.03253158586031635,
|
|
"calib/step_q_w": 0.20169918337045287,
|
|
"calib/step_q_w_n": 1347.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 1928.0,
|
|
"completions/max_terminated_length": 1928.0,
|
|
"completions/mean_length": 316.23828125,
|
|
"completions/mean_terminated_length": 316.23828125,
|
|
"completions/min_length": 54.0,
|
|
"completions/min_terminated_length": 54.0,
|
|
"epoch": 0.21333333333333335,
|
|
"grad_norm": 0.013604212552309036,
|
|
"learning_rate": 0.0,
|
|
"loss": 0.0891,
|
|
"num_tokens": 40214043.0,
|
|
"reward": 1.0026452541351318,
|
|
"reward_std": 0.01684332825243473,
|
|
"rewards/accuracy_reward_step": 0.2890625,
|
|
"rewards/final_brier_reward_step": 0.7201343774795532,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 200
|
|
},
|
|
{
|
|
"epoch": 0.21333333333333335,
|
|
"step": 200,
|
|
"total_flos": 0.0,
|
|
"train_loss": 0.09831852782517672,
|
|
"train_runtime": 18030.7973,
|
|
"train_samples_per_second": 2.84,
|
|
"train_steps_per_second": 0.011
|
|
}
|
|
],
|
|
"logging_steps": 1,
|
|
"max_steps": 200,
|
|
"num_input_tokens_seen": 40214043,
|
|
"num_train_epochs": 1,
|
|
"save_steps": 20,
|
|
"stateful_callbacks": {
|
|
"TrainerControl": {
|
|
"args": {
|
|
"should_epoch_stop": false,
|
|
"should_evaluate": false,
|
|
"should_log": false,
|
|
"should_save": true,
|
|
"should_training_stop": true
|
|
},
|
|
"attributes": {}
|
|
}
|
|
},
|
|
"total_flos": 0.0,
|
|
"train_batch_size": 8,
|
|
"trial_name": null,
|
|
"trial_params": null
|
|
}
|