9040 lines
344 KiB
JSON
9040 lines
344 KiB
JSON
{
|
|
"best_global_step": null,
|
|
"best_metric": null,
|
|
"best_model_checkpoint": null,
|
|
"epoch": 0.21333333333333335,
|
|
"eval_steps": 500,
|
|
"global_step": 200,
|
|
"is_hyper_param_search": false,
|
|
"is_local_process_zero": true,
|
|
"is_world_process_zero": true,
|
|
"log_history": [
|
|
{
|
|
"calib/answer_extract_rate": 0.0546875,
|
|
"calib/avg_num_step_conf": 0.44921875,
|
|
"calib/ece": 0.16135714285714264,
|
|
"calib/final_conf_rate": 0.0546875,
|
|
"calib/format_rate": 0.05078125,
|
|
"calib/frac_conf_gt_0.9": 0.8571428571428571,
|
|
"calib/gap": 0.12215151515151501,
|
|
"calib/mean_conf": 0.9256428571428571,
|
|
"calib/mu_c": 0.9518181818181817,
|
|
"calib/mu_w": 0.8296666666666667,
|
|
"calib/nonempty_final_conf_rate": 0.0546875,
|
|
"calib/nonempty_reasoning_rate": 0.08203125,
|
|
"calib/nonempty_step_conf_rate": 0.078125,
|
|
"calib/pce": 0.15064285714285694,
|
|
"calib/std_conf": 0.1237495207164497,
|
|
"calib/step_conf_rate": 0.078125,
|
|
"calib/step_q_c": 0.8832727272727273,
|
|
"calib/step_q_c_n": 55.0,
|
|
"calib/step_q_gap": 0.042872727272727285,
|
|
"calib/step_q_w": 0.8404,
|
|
"calib/step_q_w_n": 60.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2909.0,
|
|
"completions/max_terminated_length": 2909.0,
|
|
"completions/mean_length": 579.68359375,
|
|
"completions/mean_terminated_length": 648.0305786132812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 1.0,
|
|
"epoch": 0.0010666666666666667,
|
|
"grad_norm": 0.6144071817398071,
|
|
"learning_rate": 0.0,
|
|
"loss": -0.0112,
|
|
"num_tokens": 255983.0,
|
|
"reward": 0.068359375,
|
|
"reward_std": 0.14980050921440125,
|
|
"rewards/accuracy_reward_step": 0.04296875,
|
|
"rewards/format_reward_step": 0.05078125,
|
|
"step": 1
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.0546875,
|
|
"calib/avg_num_step_conf": 0.359375,
|
|
"calib/ece": 0.6484615384615385,
|
|
"calib/final_conf_rate": 0.05078125,
|
|
"calib/format_rate": 0.046875,
|
|
"calib/frac_conf_gt_0.9": 0.9230769230769231,
|
|
"calib/gap": 0.016388888888888786,
|
|
"calib/mean_conf": 0.9561538461538462,
|
|
"calib/mu_c": 0.9674999999999999,
|
|
"calib/mu_w": 0.9511111111111111,
|
|
"calib/nonempty_final_conf_rate": 0.05078125,
|
|
"calib/nonempty_reasoning_rate": 0.07421875,
|
|
"calib/nonempty_step_conf_rate": 0.06640625,
|
|
"calib/pce": 0.6484615384615385,
|
|
"calib/std_conf": 0.024663414679817527,
|
|
"calib/step_conf_rate": 0.06640625,
|
|
"calib/step_q_c": 0.8756521739130435,
|
|
"calib/step_q_c_n": 23.0,
|
|
"calib/step_q_gap": 0.015072463768115885,
|
|
"calib/step_q_w": 0.8605797101449276,
|
|
"calib/step_q_w_n": 69.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.05078125,
|
|
"completions/max_length": 2949.0,
|
|
"completions/max_terminated_length": 2949.0,
|
|
"completions/mean_length": 785.5390625,
|
|
"completions/mean_terminated_length": 827.5637817382812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0021333333333333334,
|
|
"grad_norm": 0.46494486927986145,
|
|
"learning_rate": 2.5000000000000004e-07,
|
|
"loss": -0.0007,
|
|
"num_tokens": 560369.0,
|
|
"reward": 0.0390625,
|
|
"reward_std": 0.0974610224366188,
|
|
"rewards/accuracy_reward_step": 0.015625,
|
|
"rewards/format_reward_step": 0.046875,
|
|
"step": 2
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.03515625,
|
|
"calib/avg_num_step_conf": 0.16796875,
|
|
"calib/ece": 0.52625,
|
|
"calib/final_conf_rate": 0.03125,
|
|
"calib/format_rate": 0.01953125,
|
|
"calib/frac_conf_gt_0.9": 0.75,
|
|
"calib/gap": -0.12750000000000017,
|
|
"calib/mean_conf": 0.90125,
|
|
"calib/mu_c": 0.8374999999999999,
|
|
"calib/mu_w": 0.9650000000000001,
|
|
"calib/nonempty_final_conf_rate": 0.03125,
|
|
"calib/nonempty_reasoning_rate": 0.04296875,
|
|
"calib/nonempty_step_conf_rate": 0.02734375,
|
|
"calib/pce": 0.46375,
|
|
"calib/std_conf": 0.1540647185438639,
|
|
"calib/step_conf_rate": 0.02734375,
|
|
"calib/step_q_c": 0.8846666666666667,
|
|
"calib/step_q_c_n": 15.0,
|
|
"calib/step_q_gap": 0.11323809523809536,
|
|
"calib/step_q_w": 0.7714285714285714,
|
|
"calib/step_q_w_n": 28.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3048.0,
|
|
"completions/max_terminated_length": 3048.0,
|
|
"completions/mean_length": 700.59765625,
|
|
"completions/mean_terminated_length": 769.75537109375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0032,
|
|
"grad_norm": 0.32013899087905884,
|
|
"learning_rate": 5.000000000000001e-07,
|
|
"loss": 0.0059,
|
|
"num_tokens": 844978.0,
|
|
"reward": 0.025390625,
|
|
"reward_std": 0.07181552797555923,
|
|
"rewards/accuracy_reward_step": 0.015625,
|
|
"rewards/format_reward_step": 0.01953125,
|
|
"step": 3
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.02734375,
|
|
"calib/avg_num_step_conf": 0.0234375,
|
|
"calib/ece": 0.9333333333333333,
|
|
"calib/final_conf_rate": 0.0234375,
|
|
"calib/format_rate": 0.00390625,
|
|
"calib/frac_conf_gt_0.9": 0.8333333333333334,
|
|
"calib/mean_conf": 0.9333333333333332,
|
|
"calib/mu_c": NaN,
|
|
"calib/mu_w": 0.9333333333333332,
|
|
"calib/nonempty_final_conf_rate": 0.0234375,
|
|
"calib/nonempty_reasoning_rate": 0.02734375,
|
|
"calib/nonempty_step_conf_rate": 0.00390625,
|
|
"calib/pce": 0.9333333333333333,
|
|
"calib/std_conf": 0.06523461930260308,
|
|
"calib/step_conf_rate": 0.00390625,
|
|
"calib/step_q_w": 0.8500000000000001,
|
|
"calib/step_q_w_n": 6.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.109375,
|
|
"completions/max_length": 2940.0,
|
|
"completions/max_terminated_length": 2940.0,
|
|
"completions/mean_length": 701.765625,
|
|
"completions/mean_terminated_length": 787.9473876953125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.004266666666666667,
|
|
"grad_norm": 0.044668663293123245,
|
|
"learning_rate": 7.5e-07,
|
|
"loss": 0.0007,
|
|
"num_tokens": 1130798.0,
|
|
"reward": 0.001953125,
|
|
"reward_std": 0.005524271633476019,
|
|
"rewards/accuracy_reward_step": 0.0,
|
|
"rewards/format_reward_step": 0.00390625,
|
|
"step": 4
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.0546875,
|
|
"calib/avg_num_step_conf": 0.26953125,
|
|
"calib/ece": 0.5896428571428574,
|
|
"calib/final_conf_rate": 0.0546875,
|
|
"calib/format_rate": 0.046875,
|
|
"calib/frac_conf_gt_0.9": 0.9285714285714286,
|
|
"calib/gap": 0.0454444444444444,
|
|
"calib/mean_conf": 0.9467857142857143,
|
|
"calib/mu_c": 0.976,
|
|
"calib/mu_w": 0.9305555555555556,
|
|
"calib/nonempty_final_conf_rate": 0.0546875,
|
|
"calib/nonempty_reasoning_rate": 0.05859375,
|
|
"calib/nonempty_step_conf_rate": 0.05078125,
|
|
"calib/pce": 0.5896428571428574,
|
|
"calib/std_conf": 0.08433790756866821,
|
|
"calib/step_conf_rate": 0.05078125,
|
|
"calib/step_q_c": 0.8807142857142859,
|
|
"calib/step_q_c_n": 14.0,
|
|
"calib/step_q_gap": -0.013376623376623153,
|
|
"calib/step_q_w": 0.894090909090909,
|
|
"calib/step_q_w_n": 55.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.08203125,
|
|
"completions/max_length": 2926.0,
|
|
"completions/max_terminated_length": 2926.0,
|
|
"completions/mean_length": 699.68359375,
|
|
"completions/mean_terminated_length": 762.20849609375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.005333333333333333,
|
|
"grad_norm": 0.7100959420204163,
|
|
"learning_rate": 1.0000000000000002e-06,
|
|
"loss": 0.0069,
|
|
"num_tokens": 1416605.0,
|
|
"reward": 0.04296875,
|
|
"reward_std": 0.0989597737789154,
|
|
"rewards/accuracy_reward_step": 0.01953125,
|
|
"rewards/format_reward_step": 0.046875,
|
|
"step": 5
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.0546875,
|
|
"calib/avg_num_step_conf": 0.25390625,
|
|
"calib/ece": 0.6266666666666667,
|
|
"calib/final_conf_rate": 0.046875,
|
|
"calib/format_rate": 0.04296875,
|
|
"calib/frac_conf_gt_0.9": 0.9166666666666666,
|
|
"calib/gap": 0.00374999999999992,
|
|
"calib/mean_conf": 0.9600000000000001,
|
|
"calib/mu_c": 0.9624999999999999,
|
|
"calib/mu_w": 0.95875,
|
|
"calib/nonempty_final_conf_rate": 0.046875,
|
|
"calib/nonempty_reasoning_rate": 0.05859375,
|
|
"calib/nonempty_step_conf_rate": 0.05078125,
|
|
"calib/pce": 0.6266666666666667,
|
|
"calib/std_conf": 0.03415650255319866,
|
|
"calib/step_conf_rate": 0.05078125,
|
|
"calib/step_q_c": 0.7994444444444445,
|
|
"calib/step_q_c_n": 18.0,
|
|
"calib/step_q_gap": -0.07119385342789597,
|
|
"calib/step_q_w": 0.8706382978723405,
|
|
"calib/step_q_w_n": 47.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3042.0,
|
|
"completions/max_terminated_length": 3042.0,
|
|
"completions/mean_length": 611.203125,
|
|
"completions/mean_terminated_length": 674.4310302734375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0064,
|
|
"grad_norm": 0.41114652156829834,
|
|
"learning_rate": 1.25e-06,
|
|
"loss": -0.0202,
|
|
"num_tokens": 1679025.0,
|
|
"reward": 0.037109375,
|
|
"reward_std": 0.09193675220012665,
|
|
"rewards/accuracy_reward_step": 0.015625,
|
|
"rewards/format_reward_step": 0.04296875,
|
|
"step": 6
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.046875,
|
|
"calib/avg_num_step_conf": 0.28515625,
|
|
"calib/ece": 0.6388888888888888,
|
|
"calib/final_conf_rate": 0.03515625,
|
|
"calib/format_rate": 0.02734375,
|
|
"calib/frac_conf_gt_0.9": 0.8888888888888888,
|
|
"calib/gap": 0.01666666666666672,
|
|
"calib/mean_conf": 0.9722222222222222,
|
|
"calib/mu_c": 0.9833333333333334,
|
|
"calib/mu_w": 0.9666666666666667,
|
|
"calib/nonempty_final_conf_rate": 0.03515625,
|
|
"calib/nonempty_reasoning_rate": 0.0703125,
|
|
"calib/nonempty_step_conf_rate": 0.0546875,
|
|
"calib/pce": 0.6388888888888888,
|
|
"calib/std_conf": 0.029355210696939787,
|
|
"calib/step_conf_rate": 0.0546875,
|
|
"calib/step_q_c": 0.9376666666666664,
|
|
"calib/step_q_c_n": 30.0,
|
|
"calib/step_q_gap": 0.04092248062015491,
|
|
"calib/step_q_w": 0.8967441860465115,
|
|
"calib/step_q_w_n": 43.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3066.0,
|
|
"completions/max_terminated_length": 3066.0,
|
|
"completions/mean_length": 690.18359375,
|
|
"completions/mean_terminated_length": 778.3568115234375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.007466666666666667,
|
|
"grad_norm": 0.3305240869522095,
|
|
"learning_rate": 1.5e-06,
|
|
"loss": -0.0039,
|
|
"num_tokens": 1963136.0,
|
|
"reward": 0.025390625,
|
|
"reward_std": 0.0679999589920044,
|
|
"rewards/accuracy_reward_step": 0.01171875,
|
|
"rewards/format_reward_step": 0.02734375,
|
|
"step": 7
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.046875,
|
|
"calib/avg_num_step_conf": 0.21484375,
|
|
"calib/ece": 0.4277777777777778,
|
|
"calib/final_conf_rate": 0.03515625,
|
|
"calib/format_rate": 0.01953125,
|
|
"calib/frac_conf_gt_0.9": 0.8888888888888888,
|
|
"calib/gap": -0.010000000000000009,
|
|
"calib/mean_conf": 0.9544444444444444,
|
|
"calib/mu_c": 0.95,
|
|
"calib/mu_w": 0.96,
|
|
"calib/nonempty_final_conf_rate": 0.03515625,
|
|
"calib/nonempty_reasoning_rate": 0.05078125,
|
|
"calib/nonempty_step_conf_rate": 0.0234375,
|
|
"calib/pce": 0.4133333333333334,
|
|
"calib/std_conf": 0.037151674438445526,
|
|
"calib/step_conf_rate": 0.0234375,
|
|
"calib/step_q_c": 0.8305555555555555,
|
|
"calib/step_q_c_n": 18.0,
|
|
"calib/step_q_gap": 0.07939879879879874,
|
|
"calib/step_q_w": 0.7511567567567567,
|
|
"calib/step_q_w_n": 37.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2928.0,
|
|
"completions/max_terminated_length": 2928.0,
|
|
"completions/mean_length": 646.625,
|
|
"completions/mean_terminated_length": 719.7216796875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.008533333333333334,
|
|
"grad_norm": 0.33451029658317566,
|
|
"learning_rate": 1.75e-06,
|
|
"loss": -0.0008,
|
|
"num_tokens": 2235184.0,
|
|
"reward": 0.029296875,
|
|
"reward_std": 0.0774708092212677,
|
|
"rewards/accuracy_reward_step": 0.01953125,
|
|
"rewards/format_reward_step": 0.01953125,
|
|
"step": 8
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.05859375,
|
|
"calib/avg_num_step_conf": 0.34375,
|
|
"calib/ece": 0.4984615384615386,
|
|
"calib/final_conf_rate": 0.05078125,
|
|
"calib/format_rate": 0.0390625,
|
|
"calib/frac_conf_gt_0.9": 0.9230769230769231,
|
|
"calib/gap": 0.040238095238095295,
|
|
"calib/mean_conf": 0.9600000000000002,
|
|
"calib/mu_c": 0.9816666666666668,
|
|
"calib/mu_w": 0.9414285714285715,
|
|
"calib/nonempty_final_conf_rate": 0.05078125,
|
|
"calib/nonempty_reasoning_rate": 0.09375,
|
|
"calib/nonempty_step_conf_rate": 0.078125,
|
|
"calib/pce": 0.4984615384615386,
|
|
"calib/std_conf": 0.025720389995846898,
|
|
"calib/step_conf_rate": 0.078125,
|
|
"calib/step_q_c": 0.9084375,
|
|
"calib/step_q_c_n": 32.0,
|
|
"calib/step_q_gap": 0.04022321428571429,
|
|
"calib/step_q_w": 0.8682142857142857,
|
|
"calib/step_q_w_n": 56.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0625,
|
|
"completions/max_length": 2904.0,
|
|
"completions/max_terminated_length": 2904.0,
|
|
"completions/mean_length": 652.80078125,
|
|
"completions/mean_terminated_length": 696.3208618164062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0096,
|
|
"grad_norm": 0.47666192054748535,
|
|
"learning_rate": 2.0000000000000003e-06,
|
|
"loss": 0.0229,
|
|
"num_tokens": 2509837.0,
|
|
"reward": 0.046875,
|
|
"reward_std": 0.1082572191953659,
|
|
"rewards/accuracy_reward_step": 0.02734375,
|
|
"rewards/format_reward_step": 0.0390625,
|
|
"step": 9
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.07421875,
|
|
"calib/avg_num_step_conf": 0.27734375,
|
|
"calib/ece": 0.4323529411764707,
|
|
"calib/final_conf_rate": 0.06640625,
|
|
"calib/format_rate": 0.0546875,
|
|
"calib/frac_conf_gt_0.9": 1.0,
|
|
"calib/gap": 0.0033333333333334103,
|
|
"calib/mean_conf": 0.9617647058823531,
|
|
"calib/mu_c": 0.9633333333333334,
|
|
"calib/mu_w": 0.96,
|
|
"calib/nonempty_final_conf_rate": 0.06640625,
|
|
"calib/nonempty_reasoning_rate": 0.08203125,
|
|
"calib/nonempty_step_conf_rate": 0.0625,
|
|
"calib/pce": 0.4323529411764707,
|
|
"calib/std_conf": 0.01854574443580525,
|
|
"calib/step_conf_rate": 0.0625,
|
|
"calib/step_q_c": 0.8643589743589746,
|
|
"calib/step_q_c_n": 39.0,
|
|
"calib/step_q_gap": 0.015296474358974521,
|
|
"calib/step_q_w": 0.8490625,
|
|
"calib/step_q_w_n": 32.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2553.0,
|
|
"completions/max_terminated_length": 2553.0,
|
|
"completions/mean_length": 610.796875,
|
|
"completions/mean_terminated_length": 673.9827270507812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.010666666666666666,
|
|
"grad_norm": 0.35018742084503174,
|
|
"learning_rate": 2.25e-06,
|
|
"loss": -0.0116,
|
|
"num_tokens": 2773001.0,
|
|
"reward": 0.0625,
|
|
"reward_std": 0.09454704821109772,
|
|
"rewards/accuracy_reward_step": 0.03515625,
|
|
"rewards/format_reward_step": 0.0546875,
|
|
"step": 10
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.1015625,
|
|
"calib/avg_num_step_conf": 0.62109375,
|
|
"calib/ece": 0.52276,
|
|
"calib/final_conf_rate": 0.09765625,
|
|
"calib/format_rate": 0.078125,
|
|
"calib/frac_conf_gt_0.9": 0.96,
|
|
"calib/gap": -0.0033051948051948665,
|
|
"calib/mean_conf": 0.96276,
|
|
"calib/mu_c": 0.9609090909090909,
|
|
"calib/mu_w": 0.9642142857142858,
|
|
"calib/nonempty_final_conf_rate": 0.09765625,
|
|
"calib/nonempty_reasoning_rate": 0.12109375,
|
|
"calib/nonempty_step_conf_rate": 0.10546875,
|
|
"calib/pce": 0.52276,
|
|
"calib/std_conf": 0.027734858932397695,
|
|
"calib/step_conf_rate": 0.10546875,
|
|
"calib/step_q_c": 0.8749122807017544,
|
|
"calib/step_q_c_n": 57.0,
|
|
"calib/step_q_gap": 0.014226006191950247,
|
|
"calib/step_q_w": 0.8606862745098042,
|
|
"calib/step_q_w_n": 102.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.06640625,
|
|
"completions/max_length": 3000.0,
|
|
"completions/max_terminated_length": 3000.0,
|
|
"completions/mean_length": 690.19140625,
|
|
"completions/mean_terminated_length": 739.2844848632812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.011733333333333333,
|
|
"grad_norm": 0.3576546311378479,
|
|
"learning_rate": 2.5e-06,
|
|
"loss": -0.0108,
|
|
"num_tokens": 3054170.0,
|
|
"reward": 0.08203125,
|
|
"reward_std": 0.18240919709205627,
|
|
"rewards/accuracy_reward_step": 0.04296875,
|
|
"rewards/format_reward_step": 0.078125,
|
|
"step": 11
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.17578125,
|
|
"calib/avg_num_step_conf": 1.23828125,
|
|
"calib/ece": 0.5485106382978725,
|
|
"calib/final_conf_rate": 0.18359375,
|
|
"calib/format_rate": 0.1484375,
|
|
"calib/frac_conf_gt_0.9": 0.8723404255319149,
|
|
"calib/gap": -0.03720370370370363,
|
|
"calib/mean_conf": 0.937872340425532,
|
|
"calib/mu_c": 0.9165000000000001,
|
|
"calib/mu_w": 0.9537037037037037,
|
|
"calib/nonempty_final_conf_rate": 0.18359375,
|
|
"calib/nonempty_reasoning_rate": 0.20703125,
|
|
"calib/nonempty_step_conf_rate": 0.1875,
|
|
"calib/pce": 0.5304255319148937,
|
|
"calib/std_conf": 0.12163664359367686,
|
|
"calib/step_conf_rate": 0.1875,
|
|
"calib/step_q_c": 0.8323376623376624,
|
|
"calib/step_q_c_n": 154.0,
|
|
"calib/step_q_gap": -0.035331049318779395,
|
|
"calib/step_q_w": 0.8676687116564418,
|
|
"calib/step_q_w_n": 163.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2946.0,
|
|
"completions/max_terminated_length": 2946.0,
|
|
"completions/mean_length": 557.2578125,
|
|
"completions/mean_terminated_length": 582.2775268554688,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 1.0,
|
|
"epoch": 0.0128,
|
|
"grad_norm": 0.4705829620361328,
|
|
"learning_rate": 2.7500000000000004e-06,
|
|
"loss": -0.0058,
|
|
"num_tokens": 3301004.0,
|
|
"reward": 0.15234375,
|
|
"reward_std": 0.30028218030929565,
|
|
"rewards/accuracy_reward_step": 0.078125,
|
|
"rewards/format_reward_step": 0.1484375,
|
|
"step": 12
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.2109375,
|
|
"calib/avg_num_step_conf": 1.44140625,
|
|
"calib/ece": 0.4812499999999999,
|
|
"calib/final_conf_rate": 0.203125,
|
|
"calib/format_rate": 0.1640625,
|
|
"calib/frac_conf_gt_0.9": 0.8461538461538461,
|
|
"calib/gap": -0.03523703703703718,
|
|
"calib/mean_conf": 0.9350961538461535,
|
|
"calib/mu_c": 0.9168,
|
|
"calib/mu_w": 0.9520370370370371,
|
|
"calib/nonempty_final_conf_rate": 0.203125,
|
|
"calib/nonempty_reasoning_rate": 0.2578125,
|
|
"calib/nonempty_step_conf_rate": 0.23828125,
|
|
"calib/pce": 0.46778846153846143,
|
|
"calib/std_conf": 0.09902160984338888,
|
|
"calib/step_conf_rate": 0.23828125,
|
|
"calib/step_q_c": 0.8649668874172186,
|
|
"calib/step_q_c_n": 151.0,
|
|
"calib/step_q_gap": -0.019712011665350082,
|
|
"calib/step_q_w": 0.8846788990825687,
|
|
"calib/step_q_w_n": 218.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3041.0,
|
|
"completions/max_terminated_length": 3041.0,
|
|
"completions/mean_length": 631.0546875,
|
|
"completions/mean_terminated_length": 684.5338745117188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 1.0,
|
|
"epoch": 0.013866666666666666,
|
|
"grad_norm": 0.508955180644989,
|
|
"learning_rate": 3e-06,
|
|
"loss": -0.0062,
|
|
"num_tokens": 3567146.0,
|
|
"reward": 0.18359375,
|
|
"reward_std": 0.2880544066429138,
|
|
"rewards/accuracy_reward_step": 0.1015625,
|
|
"rewards/format_reward_step": 0.1640625,
|
|
"step": 13
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.37109375,
|
|
"calib/avg_num_step_conf": 1.9453125,
|
|
"calib/ece": 0.5424275362318839,
|
|
"calib/final_conf_rate": 0.359375,
|
|
"calib/format_rate": 0.28125,
|
|
"calib/frac_conf_gt_0.9": 0.9130434782608695,
|
|
"calib/gap": 0.023949221949222,
|
|
"calib/mean_conf": 0.9446014492753624,
|
|
"calib/mu_c": 0.9589189189189188,
|
|
"calib/mu_w": 0.9349696969696968,
|
|
"calib/nonempty_final_conf_rate": 0.359375,
|
|
"calib/nonempty_reasoning_rate": 0.3984375,
|
|
"calib/nonempty_step_conf_rate": 0.33203125,
|
|
"calib/pce": 0.5424275362318839,
|
|
"calib/std_conf": 0.08500616969837242,
|
|
"calib/step_conf_rate": 0.33203125,
|
|
"calib/step_q_c": 0.8756976744186047,
|
|
"calib/step_q_c_n": 172.0,
|
|
"calib/step_q_gap": 0.018601559899177222,
|
|
"calib/step_q_w": 0.8570961145194275,
|
|
"calib/step_q_w_n": 326.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2916.0,
|
|
"completions/max_terminated_length": 2916.0,
|
|
"completions/mean_length": 531.89453125,
|
|
"completions/mean_terminated_length": 553.5162353515625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 1.0,
|
|
"epoch": 0.014933333333333333,
|
|
"grad_norm": 0.6900785565376282,
|
|
"learning_rate": 3.2500000000000002e-06,
|
|
"loss": -0.0625,
|
|
"num_tokens": 3808711.0,
|
|
"reward": 0.28515625,
|
|
"reward_std": 0.34519943594932556,
|
|
"rewards/accuracy_reward_step": 0.14453125,
|
|
"rewards/format_reward_step": 0.28125,
|
|
"step": 14
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.56640625,
|
|
"calib/avg_num_step_conf": 3.875,
|
|
"calib/ece": 0.5345314685314685,
|
|
"calib/final_conf_rate": 0.55859375,
|
|
"calib/format_rate": 0.4921875,
|
|
"calib/frac_conf_gt_0.9": 0.8741258741258742,
|
|
"calib/gap": 0.011973970944310075,
|
|
"calib/mean_conf": 0.947118881118881,
|
|
"calib/mu_c": 0.9541525423728814,
|
|
"calib/mu_w": 0.9421785714285713,
|
|
"calib/nonempty_final_conf_rate": 0.55859375,
|
|
"calib/nonempty_reasoning_rate": 0.63671875,
|
|
"calib/nonempty_step_conf_rate": 0.5859375,
|
|
"calib/pce": 0.5345314685314685,
|
|
"calib/std_conf": 0.0888806451600279,
|
|
"calib/step_conf_rate": 0.5859375,
|
|
"calib/step_q_c": 0.8749010695187166,
|
|
"calib/step_q_c_n": 374.0,
|
|
"calib/step_q_gap": 0.0007360209750272295,
|
|
"calib/step_q_w": 0.8741650485436894,
|
|
"calib/step_q_w_n": 618.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.01953125,
|
|
"completions/max_length": 3020.0,
|
|
"completions/max_terminated_length": 3020.0,
|
|
"completions/mean_length": 522.0625,
|
|
"completions/mean_terminated_length": 532.462158203125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 1.0,
|
|
"epoch": 0.016,
|
|
"grad_norm": 0.9986044764518738,
|
|
"learning_rate": 3.5e-06,
|
|
"loss": -0.0452,
|
|
"num_tokens": 4050239.0,
|
|
"reward": 0.48046875,
|
|
"reward_std": 0.4949692487716675,
|
|
"rewards/accuracy_reward_step": 0.234375,
|
|
"rewards/format_reward_step": 0.4921875,
|
|
"step": 15
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.68359375,
|
|
"calib/avg_num_step_conf": 5.328125,
|
|
"calib/ece": 0.5159766081871344,
|
|
"calib/final_conf_rate": 0.66796875,
|
|
"calib/format_rate": 0.609375,
|
|
"calib/frac_conf_gt_0.9": 0.9239766081871345,
|
|
"calib/gap": 0.00936842105263147,
|
|
"calib/mean_conf": 0.9557426900584794,
|
|
"calib/mu_c": 0.9609473684210526,
|
|
"calib/mu_w": 0.9515789473684211,
|
|
"calib/nonempty_final_conf_rate": 0.66796875,
|
|
"calib/nonempty_reasoning_rate": 0.76953125,
|
|
"calib/nonempty_step_conf_rate": 0.734375,
|
|
"calib/pce": 0.5136374269005847,
|
|
"calib/std_conf": 0.0471109601210145,
|
|
"calib/step_conf_rate": 0.734375,
|
|
"calib/step_q_c": 0.8763611615245009,
|
|
"calib/step_q_c_n": 551.0,
|
|
"calib/step_q_gap": -0.0035773378604930794,
|
|
"calib/step_q_w": 0.879938499384994,
|
|
"calib/step_q_w_n": 813.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2926.0,
|
|
"completions/max_terminated_length": 2926.0,
|
|
"completions/mean_length": 500.1796875,
|
|
"completions/mean_terminated_length": 510.1434326171875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 1.0,
|
|
"epoch": 0.017066666666666667,
|
|
"grad_norm": 0.8285671472549438,
|
|
"learning_rate": 3.7500000000000005e-06,
|
|
"loss": -0.0197,
|
|
"num_tokens": 4287133.0,
|
|
"reward": 0.6171875,
|
|
"reward_std": 0.49173909425735474,
|
|
"rewards/accuracy_reward_step": 0.3125,
|
|
"rewards/format_reward_step": 0.609375,
|
|
"step": 16
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.90234375,
|
|
"calib/avg_num_step_conf": 6.94140625,
|
|
"calib/ece": 0.4604086956521738,
|
|
"calib/final_conf_rate": 0.8984375,
|
|
"calib/format_rate": 0.83203125,
|
|
"calib/frac_conf_gt_0.9": 0.9130434782608695,
|
|
"calib/gap": 0.006691167574107593,
|
|
"calib/mean_conf": 0.9517130434782608,
|
|
"calib/mu_c": 0.9550877192982455,
|
|
"calib/mu_w": 0.948396551724138,
|
|
"calib/nonempty_final_conf_rate": 0.8984375,
|
|
"calib/nonempty_reasoning_rate": 0.96875,
|
|
"calib/nonempty_step_conf_rate": 0.9453125,
|
|
"calib/pce": 0.45823478260869555,
|
|
"calib/std_conf": 0.0747794165142755,
|
|
"calib/step_conf_rate": 0.9453125,
|
|
"calib/step_q_c": 0.879920993227991,
|
|
"calib/step_q_c_n": 886.0,
|
|
"calib/step_q_gap": 0.01322290119656555,
|
|
"calib/step_q_w": 0.8666980920314254,
|
|
"calib/step_q_w_n": 891.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2139.0,
|
|
"completions/max_terminated_length": 2139.0,
|
|
"completions/mean_length": 411.65625,
|
|
"completions/mean_terminated_length": 414.89764404296875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.018133333333333335,
|
|
"grad_norm": 0.8043549656867981,
|
|
"learning_rate": 4.000000000000001e-06,
|
|
"loss": -0.0669,
|
|
"num_tokens": 4496045.0,
|
|
"reward": 0.873046875,
|
|
"reward_std": 0.5186734199523926,
|
|
"rewards/accuracy_reward_step": 0.45703125,
|
|
"rewards/format_reward_step": 0.83203125,
|
|
"step": 17
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9296875,
|
|
"calib/avg_num_step_conf": 6.6328125,
|
|
"calib/ece": 0.5076639004149379,
|
|
"calib/final_conf_rate": 0.94140625,
|
|
"calib/format_rate": 0.87109375,
|
|
"calib/frac_conf_gt_0.9": 0.91701244813278,
|
|
"calib/gap": 0.004045808966861464,
|
|
"calib/mean_conf": 0.9539709543568464,
|
|
"calib/mu_c": 0.9562037037037037,
|
|
"calib/mu_w": 0.9521578947368422,
|
|
"calib/nonempty_final_conf_rate": 0.94140625,
|
|
"calib/nonempty_reasoning_rate": 0.97265625,
|
|
"calib/nonempty_step_conf_rate": 0.9296875,
|
|
"calib/pce": 0.5067510373443985,
|
|
"calib/std_conf": 0.047515395918551195,
|
|
"calib/step_conf_rate": 0.9296875,
|
|
"calib/step_q_c": 0.8654918032786885,
|
|
"calib/step_q_c_n": 732.0,
|
|
"calib/step_q_gap": 0.02443279706750834,
|
|
"calib/step_q_w": 0.8410590062111801,
|
|
"calib/step_q_w_n": 966.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 1605.0,
|
|
"completions/max_terminated_length": 1605.0,
|
|
"completions/mean_length": 394.25,
|
|
"completions/mean_terminated_length": 395.7961120605469,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 2.0,
|
|
"epoch": 0.0192,
|
|
"grad_norm": 0.47119131684303284,
|
|
"learning_rate": 4.25e-06,
|
|
"loss": -0.0478,
|
|
"num_tokens": 4707693.0,
|
|
"reward": 0.857421875,
|
|
"reward_std": 0.4101829528808594,
|
|
"rewards/accuracy_reward_step": 0.421875,
|
|
"rewards/format_reward_step": 0.87109375,
|
|
"step": 18
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 7.15625,
|
|
"calib/ece": 0.512344,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9453125,
|
|
"calib/frac_conf_gt_0.9": 0.92,
|
|
"calib/gap": 0.009028571428571275,
|
|
"calib/mean_conf": 0.9495439999999998,
|
|
"calib/mu_c": 0.9545999999999998,
|
|
"calib/mu_w": 0.9455714285714285,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
|
"calib/pce": 0.510944,
|
|
"calib/std_conf": 0.047297527038947816,
|
|
"calib/step_conf_rate": 0.984375,
|
|
"calib/step_q_c": 0.8617517814726842,
|
|
"calib/step_q_c_n": 842.0,
|
|
"calib/step_q_gap": 0.01605481177571444,
|
|
"calib/step_q_w": 0.8456969696969697,
|
|
"calib/step_q_w_n": 990.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 801.0,
|
|
"completions/max_terminated_length": 801.0,
|
|
"completions/mean_length": 355.06640625,
|
|
"completions/mean_terminated_length": 357.8622131347656,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 62.0,
|
|
"epoch": 0.020266666666666665,
|
|
"grad_norm": 0.7457324862480164,
|
|
"learning_rate": 4.5e-06,
|
|
"loss": -0.0078,
|
|
"num_tokens": 4903350.0,
|
|
"reward": 0.90234375,
|
|
"reward_std": 0.35737013816833496,
|
|
"rewards/accuracy_reward_step": 0.4296875,
|
|
"rewards/format_reward_step": 0.9453125,
|
|
"step": 19
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 7.24609375,
|
|
"calib/ece": 0.48309236947791157,
|
|
"calib/final_conf_rate": 0.97265625,
|
|
"calib/format_rate": 0.94921875,
|
|
"calib/frac_conf_gt_0.9": 0.9236947791164659,
|
|
"calib/gap": -0.0017365967365967627,
|
|
"calib/mean_conf": 0.9529718875502008,
|
|
"calib/mu_c": 0.952051282051282,
|
|
"calib/mu_w": 0.9537878787878787,
|
|
"calib/nonempty_final_conf_rate": 0.97265625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
|
"calib/pce": 0.48309236947791157,
|
|
"calib/std_conf": 0.03786473511743426,
|
|
"calib/step_conf_rate": 0.98828125,
|
|
"calib/step_q_c": 0.8567741935483871,
|
|
"calib/step_q_c_n": 837.0,
|
|
"calib/step_q_gap": 0.000634704353888127,
|
|
"calib/step_q_w": 0.856139489194499,
|
|
"calib/step_q_w_n": 1018.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1159.0,
|
|
"completions/max_terminated_length": 1159.0,
|
|
"completions/mean_length": 364.70703125,
|
|
"completions/mean_terminated_length": 367.5787353515625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 25.0,
|
|
"epoch": 0.021333333333333333,
|
|
"grad_norm": 0.5896674394607544,
|
|
"learning_rate": 4.75e-06,
|
|
"loss": -0.0232,
|
|
"num_tokens": 5101587.0,
|
|
"reward": 0.931640625,
|
|
"reward_std": 0.35057777166366577,
|
|
"rewards/accuracy_reward_step": 0.45703125,
|
|
"rewards/format_reward_step": 0.94921875,
|
|
"step": 20
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 7.8984375,
|
|
"calib/ece": 0.37701960784313726,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.9176470588235294,
|
|
"calib/gap": 0.006309538770893464,
|
|
"calib/mean_conf": 0.9468235294117647,
|
|
"calib/mu_c": 0.9495205479452055,
|
|
"calib/mu_w": 0.943211009174312,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.3756470588235294,
|
|
"calib/std_conf": 0.07272637568913073,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.8523810810810811,
|
|
"calib/step_q_c_n": 1110.0,
|
|
"calib/step_q_gap": -0.0064018136557609795,
|
|
"calib/step_q_w": 0.8587828947368421,
|
|
"calib/step_q_w_n": 912.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 874.0,
|
|
"completions/max_terminated_length": 874.0,
|
|
"completions/mean_length": 372.984375,
|
|
"completions/mean_terminated_length": 375.9212646484375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 137.0,
|
|
"epoch": 0.0224,
|
|
"grad_norm": 0.3262301981449127,
|
|
"learning_rate": 5e-06,
|
|
"loss": -0.0064,
|
|
"num_tokens": 5300031.0,
|
|
"reward": 1.068359375,
|
|
"reward_std": 0.39928555488586426,
|
|
"rewards/accuracy_reward_step": 0.5703125,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 21
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 8.52734375,
|
|
"calib/ece": 0.3571126482213438,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.9604743083003953,
|
|
"calib/gap": 0.012833246482542915,
|
|
"calib/mean_conf": 0.9579031620553359,
|
|
"calib/mu_c": 0.9630263157894736,
|
|
"calib/mu_w": 0.9501930693069307,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.3571126482213438,
|
|
"calib/std_conf": 0.04993861360051285,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.869948876839659,
|
|
"calib/step_q_c_n": 1291.0,
|
|
"calib/step_q_gap": 0.000882733341901254,
|
|
"calib/step_q_w": 0.8690661434977578,
|
|
"calib/step_q_w_n": 892.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 833.0,
|
|
"completions/max_terminated_length": 833.0,
|
|
"completions/mean_length": 390.2734375,
|
|
"completions/mean_terminated_length": 393.3464660644531,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 145.0,
|
|
"epoch": 0.023466666666666667,
|
|
"grad_norm": 0.8765632510185242,
|
|
"learning_rate": 4.9722222222222224e-06,
|
|
"loss": -0.0115,
|
|
"num_tokens": 5501757.0,
|
|
"reward": 1.08984375,
|
|
"reward_std": 0.3492136001586914,
|
|
"rewards/accuracy_reward_step": 0.59765625,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 22
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 9.171875,
|
|
"calib/ece": 0.43206640625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.8828125,
|
|
"calib/gap": -0.017310294117647174,
|
|
"calib/mean_conf": 0.94362890625,
|
|
"calib/mu_c": 0.9355147058823529,
|
|
"calib/mu_w": 0.952825,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.42222265625,
|
|
"calib/std_conf": 0.07139813947631066,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.8419904076738609,
|
|
"calib/step_q_c_n": 1251.0,
|
|
"calib/step_q_gap": -0.021208316118299653,
|
|
"calib/step_q_w": 0.8631987237921606,
|
|
"calib/step_q_w_n": 1097.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1567.0,
|
|
"completions/max_terminated_length": 1567.0,
|
|
"completions/mean_length": 414.875,
|
|
"completions/mean_terminated_length": 418.1417236328125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 122.0,
|
|
"epoch": 0.024533333333333334,
|
|
"grad_norm": 0.8596509099006653,
|
|
"learning_rate": 4.944444444444445e-06,
|
|
"loss": -0.0007,
|
|
"num_tokens": 5711901.0,
|
|
"reward": 1.03125,
|
|
"reward_std": 0.38387420773506165,
|
|
"rewards/accuracy_reward_step": 0.53125,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 23
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 9.671875,
|
|
"calib/ece": 0.47250000000000003,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.876984126984127,
|
|
"calib/gap": 0.005223352498894274,
|
|
"calib/mean_conf": 0.9447222222222222,
|
|
"calib/mu_c": 0.9474789915966386,
|
|
"calib/mu_w": 0.9422556390977443,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.47250000000000003,
|
|
"calib/std_conf": 0.05814870631282675,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.8465031503150314,
|
|
"calib/step_q_c_n": 1111.0,
|
|
"calib/step_q_gap": 0.0033529671648482307,
|
|
"calib/step_q_w": 0.8431501831501832,
|
|
"calib/step_q_w_n": 1365.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2444.0,
|
|
"completions/max_terminated_length": 2444.0,
|
|
"completions/mean_length": 465.26171875,
|
|
"completions/mean_terminated_length": 465.26171875,
|
|
"completions/min_length": 132.0,
|
|
"completions/min_terminated_length": 132.0,
|
|
"epoch": 0.0256,
|
|
"grad_norm": 0.7665349841117859,
|
|
"learning_rate": 4.9166666666666665e-06,
|
|
"loss": 0.0118,
|
|
"num_tokens": 5935520.0,
|
|
"reward": 0.95703125,
|
|
"reward_std": 0.39235660433769226,
|
|
"rewards/accuracy_reward_step": 0.46484375,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 24
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 9.9140625,
|
|
"calib/ece": 0.3672117647058823,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.8235294117647058,
|
|
"calib/gap": -0.0018439110217415289,
|
|
"calib/mean_conf": 0.9397607843137256,
|
|
"calib/mu_c": 0.9389726027397263,
|
|
"calib/mu_w": 0.9408165137614678,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.3672117647058823,
|
|
"calib/std_conf": 0.047098828420678765,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.8269407407407406,
|
|
"calib/step_q_c_n": 1350.0,
|
|
"calib/step_q_gap": -0.0189683501683503,
|
|
"calib/step_q_w": 0.8459090909090909,
|
|
"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": 0.0078125,
|
|
"completions/max_length": 1246.0,
|
|
"completions/max_terminated_length": 1246.0,
|
|
"completions/mean_length": 437.8515625,
|
|
"completions/mean_terminated_length": 441.2992248535156,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 131.0,
|
|
"epoch": 0.02666666666666667,
|
|
"grad_norm": 0.8922926783561707,
|
|
"learning_rate": 4.888888888888889e-06,
|
|
"loss": 0.0266,
|
|
"num_tokens": 6150834.0,
|
|
"reward": 1.068359375,
|
|
"reward_std": 0.30944645404815674,
|
|
"rewards/accuracy_reward_step": 0.5703125,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 25
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 9.3046875,
|
|
"calib/ece": 0.3494488188976378,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.6889763779527559,
|
|
"calib/gap": 0.005464231354642468,
|
|
"calib/mean_conf": 0.924251968503937,
|
|
"calib/mu_c": 0.9265753424657535,
|
|
"calib/mu_w": 0.921111111111111,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.3494488188976378,
|
|
"calib/std_conf": 0.05110618381061187,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.8115927873779114,
|
|
"calib/step_q_c_n": 1331.0,
|
|
"calib/step_q_gap": -0.00734156086186033,
|
|
"calib/step_q_w": 0.8189343482397717,
|
|
"calib/step_q_w_n": 1051.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2200.0,
|
|
"completions/max_terminated_length": 2200.0,
|
|
"completions/mean_length": 437.19921875,
|
|
"completions/mean_terminated_length": 438.91375732421875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 225.0,
|
|
"epoch": 0.027733333333333332,
|
|
"grad_norm": 0.5688765048980713,
|
|
"learning_rate": 4.861111111111111e-06,
|
|
"loss": 0.0257,
|
|
"num_tokens": 6367997.0,
|
|
"reward": 1.0625,
|
|
"reward_std": 0.3060663640499115,
|
|
"rewards/accuracy_reward_step": 0.5703125,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 26
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 9.58984375,
|
|
"calib/ece": 0.3891764705882352,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.5215686274509804,
|
|
"calib/gap": 0.0022369230769230253,
|
|
"calib/mean_conf": 0.8989803921568628,
|
|
"calib/mu_c": 0.900076923076923,
|
|
"calib/mu_w": 0.89784,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.3891764705882352,
|
|
"calib/std_conf": 0.06398942626993945,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.7787020905923345,
|
|
"calib/step_q_c_n": 1148.0,
|
|
"calib/step_q_gap": 0.005649297937399433,
|
|
"calib/step_q_w": 0.7730527926549351,
|
|
"calib/step_q_w_n": 1307.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1641.0,
|
|
"completions/max_terminated_length": 1641.0,
|
|
"completions/mean_length": 456.8359375,
|
|
"completions/mean_terminated_length": 460.4330749511719,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 178.0,
|
|
"epoch": 0.0288,
|
|
"grad_norm": 1.1152070760726929,
|
|
"learning_rate": 4.833333333333333e-06,
|
|
"loss": 0.0124,
|
|
"num_tokens": 6590163.0,
|
|
"reward": 1.005859375,
|
|
"reward_std": 0.30036985874176025,
|
|
"rewards/accuracy_reward_step": 0.5078125,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 27
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 10.1484375,
|
|
"calib/ece": 0.21065625000000002,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.33203125,
|
|
"calib/gap": -0.01341971207087489,
|
|
"calib/mean_conf": 0.86253125,
|
|
"calib/mu_c": 0.8581279069767442,
|
|
"calib/mu_w": 0.8715476190476191,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.20065624999999998,
|
|
"calib/std_conf": 0.08976757640394163,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.734737144585602,
|
|
"calib/step_q_c_n": 1653.0,
|
|
"calib/step_q_gap": -0.015035342186884848,
|
|
"calib/step_q_w": 0.7497724867724869,
|
|
"calib/step_q_w_n": 945.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1459.0,
|
|
"completions/max_terminated_length": 1459.0,
|
|
"completions/mean_length": 477.328125,
|
|
"completions/mean_terminated_length": 481.08660888671875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 155.0,
|
|
"epoch": 0.029866666666666666,
|
|
"grad_norm": 0.5947628617286682,
|
|
"learning_rate": 4.805555555555556e-06,
|
|
"loss": -0.0196,
|
|
"num_tokens": 6819303.0,
|
|
"reward": 1.171875,
|
|
"reward_std": 0.27487409114837646,
|
|
"rewards/accuracy_reward_step": 0.671875,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 28
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 11.08203125,
|
|
"calib/ece": 0.31074803149606295,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.24803149606299213,
|
|
"calib/gap": -0.0075980148883375564,
|
|
"calib/mean_conf": 0.8224015748031496,
|
|
"calib/mu_c": 0.8186923076923077,
|
|
"calib/mu_w": 0.8262903225806453,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.3106692913385826,
|
|
"calib/std_conf": 0.11648789331453538,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6888409090909091,
|
|
"calib/step_q_c_n": 1320.0,
|
|
"calib/step_q_gap": 0.014767079163420571,
|
|
"calib/step_q_w": 0.6740738299274885,
|
|
"calib/step_q_w_n": 1517.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2265.0,
|
|
"completions/max_terminated_length": 2265.0,
|
|
"completions/mean_length": 568.7421875,
|
|
"completions/mean_terminated_length": 570.9725952148438,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 163.0,
|
|
"epoch": 0.030933333333333334,
|
|
"grad_norm": 0.6354207992553711,
|
|
"learning_rate": 4.777777777777778e-06,
|
|
"loss": 0.0026,
|
|
"num_tokens": 7072029.0,
|
|
"reward": 1.00390625,
|
|
"reward_std": 0.307236909866333,
|
|
"rewards/accuracy_reward_step": 0.5078125,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 29
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 10.52734375,
|
|
"calib/ece": 0.1993307086614174,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.16535433070866143,
|
|
"calib/gap": 0.021545130035696136,
|
|
"calib/mean_conf": 0.7692519685039371,
|
|
"calib/mu_c": 0.7782432432432433,
|
|
"calib/mu_w": 0.7566981132075472,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.19295275590551186,
|
|
"calib/std_conf": 0.14552319232341973,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6220078226857888,
|
|
"calib/step_q_c_n": 1534.0,
|
|
"calib/step_q_gap": 0.021112043185358198,
|
|
"calib/step_q_w": 0.6008957795004306,
|
|
"calib/step_q_w_n": 1161.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2180.0,
|
|
"completions/max_terminated_length": 2180.0,
|
|
"completions/mean_length": 571.359375,
|
|
"completions/mean_terminated_length": 573.6000366210938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 47.0,
|
|
"epoch": 0.032,
|
|
"grad_norm": 0.6959826350212097,
|
|
"learning_rate": 4.75e-06,
|
|
"loss": 0.022,
|
|
"num_tokens": 7325281.0,
|
|
"reward": 1.07421875,
|
|
"reward_std": 0.31208619475364685,
|
|
"rewards/accuracy_reward_step": 0.578125,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 30
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 11.01953125,
|
|
"calib/ece": 0.21984313725490198,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.07058823529411765,
|
|
"calib/gap": -0.01857487922705303,
|
|
"calib/mean_conf": 0.6356078431372548,
|
|
"calib/mu_c": 0.6255555555555555,
|
|
"calib/mu_w": 0.6441304347826086,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.19831372549019613,
|
|
"calib/std_conf": 0.18196833148648528,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.4764291845493562,
|
|
"calib/step_q_c_n": 1165.0,
|
|
"calib/step_q_gap": 0.001211793245008308,
|
|
"calib/step_q_w": 0.47521739130434787,
|
|
"calib/step_q_w_n": 1656.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2254.0,
|
|
"completions/max_terminated_length": 2254.0,
|
|
"completions/mean_length": 572.30859375,
|
|
"completions/mean_terminated_length": 574.552978515625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 197.0,
|
|
"epoch": 0.03306666666666667,
|
|
"grad_norm": 0.4393335282802582,
|
|
"learning_rate": 4.722222222222222e-06,
|
|
"loss": 0.0167,
|
|
"num_tokens": 7577704.0,
|
|
"reward": 0.958984375,
|
|
"reward_std": 0.28661811351776123,
|
|
"rewards/accuracy_reward_step": 0.4609375,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 31
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 10.31640625,
|
|
"calib/ece": 0.18234126984126986,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.04365079365079365,
|
|
"calib/gap": -0.02955836381775856,
|
|
"calib/mean_conf": 0.5640873015873015,
|
|
"calib/mu_c": 0.5527096774193548,
|
|
"calib/mu_w": 0.5822680412371134,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06567460317460315,
|
|
"calib/std_conf": 0.17159950234143945,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.40218533886583674,
|
|
"calib/step_q_c_n": 1446.0,
|
|
"calib/step_q_gap": -0.01822470297516743,
|
|
"calib/step_q_w": 0.42041004184100417,
|
|
"calib/step_q_w_n": 1195.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2316.0,
|
|
"completions/max_terminated_length": 2316.0,
|
|
"completions/mean_length": 548.33984375,
|
|
"completions/mean_terminated_length": 550.490234375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 198.0,
|
|
"epoch": 0.034133333333333335,
|
|
"grad_norm": 5.75811243057251,
|
|
"learning_rate": 4.694444444444445e-06,
|
|
"loss": 0.042,
|
|
"num_tokens": 7824783.0,
|
|
"reward": 1.09765625,
|
|
"reward_std": 0.21498265862464905,
|
|
"rewards/accuracy_reward_step": 0.60546875,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 32
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 10.64453125,
|
|
"calib/ece": 0.11507936507936511,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.023809523809523808,
|
|
"calib/gap": 0.02414760914760894,
|
|
"calib/mean_conf": 0.5634126984126985,
|
|
"calib/mu_c": 0.5733783783783782,
|
|
"calib/mu_w": 0.5492307692307693,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.045595238095238105,
|
|
"calib/std_conf": 0.16762184716878084,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.42119708994709,
|
|
"calib/step_q_c_n": 1512.0,
|
|
"calib/step_q_gap": 0.013291071810239197,
|
|
"calib/step_q_w": 0.4079060181368508,
|
|
"calib/step_q_w_n": 1213.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2509.0,
|
|
"completions/max_terminated_length": 2509.0,
|
|
"completions/mean_length": 577.9453125,
|
|
"completions/mean_terminated_length": 577.9453125,
|
|
"completions/min_length": 203.0,
|
|
"completions/min_terminated_length": 203.0,
|
|
"epoch": 0.0352,
|
|
"grad_norm": 5.477323532104492,
|
|
"learning_rate": 4.666666666666667e-06,
|
|
"loss": 0.0567,
|
|
"num_tokens": 8079609.0,
|
|
"reward": 1.0703125,
|
|
"reward_std": 0.2605133056640625,
|
|
"rewards/accuracy_reward_step": 0.578125,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 33
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 9.59765625,
|
|
"calib/ece": 0.1350197628458498,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.011857707509881422,
|
|
"calib/gap": 0.01922571656050953,
|
|
"calib/mean_conf": 0.5447430830039526,
|
|
"calib/mu_c": 0.5520382165605096,
|
|
"calib/mu_w": 0.5328125,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.029604743083003947,
|
|
"calib/std_conf": 0.15592777668864602,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.40732686980609417,
|
|
"calib/step_q_c_n": 1444.0,
|
|
"calib/step_q_gap": 0.043407817486252165,
|
|
"calib/step_q_w": 0.363919052319842,
|
|
"calib/step_q_w_n": 1013.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2499.0,
|
|
"completions/max_terminated_length": 2499.0,
|
|
"completions/mean_length": 497.96484375,
|
|
"completions/mean_terminated_length": 499.91766357421875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 158.0,
|
|
"epoch": 0.03626666666666667,
|
|
"grad_norm": 1.898915410041809,
|
|
"learning_rate": 4.638888888888889e-06,
|
|
"loss": 0.0402,
|
|
"num_tokens": 8312200.0,
|
|
"reward": 1.10546875,
|
|
"reward_std": 0.35471436381340027,
|
|
"rewards/accuracy_reward_step": 0.61328125,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 34
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 10.8359375,
|
|
"calib/ece": 0.11332015810276687,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.003952569169960474,
|
|
"calib/gap": 0.03323832463535381,
|
|
"calib/mean_conf": 0.534505928853755,
|
|
"calib/mu_c": 0.5468553459119496,
|
|
"calib/mu_w": 0.5136170212765958,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.009683794466403172,
|
|
"calib/std_conf": 0.13347089275891036,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.3990954773869347,
|
|
"calib/step_q_c_n": 1592.0,
|
|
"calib/step_q_gap": 0.009763836100978696,
|
|
"calib/step_q_w": 0.389331641285956,
|
|
"calib/step_q_w_n": 1182.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2322.0,
|
|
"completions/max_terminated_length": 2322.0,
|
|
"completions/mean_length": 615.69140625,
|
|
"completions/mean_terminated_length": 615.69140625,
|
|
"completions/min_length": 223.0,
|
|
"completions/min_terminated_length": 223.0,
|
|
"epoch": 0.037333333333333336,
|
|
"grad_norm": 1.230494737625122,
|
|
"learning_rate": 4.611111111111112e-06,
|
|
"loss": 0.0197,
|
|
"num_tokens": 8579073.0,
|
|
"reward": 1.115234375,
|
|
"reward_std": 0.22809851169586182,
|
|
"rewards/accuracy_reward_step": 0.62109375,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 35
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 9.29296875,
|
|
"calib/ece": 0.19019607843137257,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0196078431372549,
|
|
"calib/gap": 0.006778074866310102,
|
|
"calib/mean_conf": 0.5654117647058824,
|
|
"calib/mu_c": 0.5672192513368984,
|
|
"calib/mu_w": 0.5604411764705883,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.011137254901960788,
|
|
"calib/std_conf": 0.13830247845887,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.43395945140131187,
|
|
"calib/step_q_c_n": 1677.0,
|
|
"calib/step_q_gap": -0.015314052872192396,
|
|
"calib/step_q_w": 0.44927350427350426,
|
|
"calib/step_q_w_n": 702.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2503.0,
|
|
"completions/max_terminated_length": 2503.0,
|
|
"completions/mean_length": 497.62890625,
|
|
"completions/mean_terminated_length": 499.5804138183594,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 171.0,
|
|
"epoch": 0.0384,
|
|
"grad_norm": 0.30834102630615234,
|
|
"learning_rate": 4.583333333333333e-06,
|
|
"loss": 0.0048,
|
|
"num_tokens": 8809178.0,
|
|
"reward": 1.228515625,
|
|
"reward_std": 0.2734318673610687,
|
|
"rewards/accuracy_reward_step": 0.73046875,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 36
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9609375,
|
|
"calib/avg_num_step_conf": 10.1796875,
|
|
"calib/ece": 0.17707317073170734,
|
|
"calib/final_conf_rate": 0.9609375,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.0040650406504065045,
|
|
"calib/gap": -0.00423618634886247,
|
|
"calib/mean_conf": 0.5060162601626017,
|
|
"calib/mu_c": 0.504225352112676,
|
|
"calib/mu_w": 0.5084615384615385,
|
|
"calib/nonempty_final_conf_rate": 0.9609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.05292682926829267,
|
|
"calib/std_conf": 0.13435325259767472,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.3797255574614065,
|
|
"calib/step_q_c_n": 1166.0,
|
|
"calib/step_q_gap": -0.03260777587192687,
|
|
"calib/step_q_w": 0.4123333333333334,
|
|
"calib/step_q_w_n": 1440.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2672.0,
|
|
"completions/max_terminated_length": 2672.0,
|
|
"completions/mean_length": 598.4296875,
|
|
"completions/mean_terminated_length": 600.7764892578125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 150.0,
|
|
"epoch": 0.039466666666666664,
|
|
"grad_norm": 2.395369052886963,
|
|
"learning_rate": 4.555555555555556e-06,
|
|
"loss": 0.0397,
|
|
"num_tokens": 9069472.0,
|
|
"reward": 1.03515625,
|
|
"reward_std": 0.3008233904838562,
|
|
"rewards/accuracy_reward_step": 0.5546875,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 37
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 10.0625,
|
|
"calib/ece": 0.13287449392712547,
|
|
"calib/final_conf_rate": 0.96484375,
|
|
"calib/format_rate": 0.96484375,
|
|
"calib/frac_conf_gt_0.9": 0.004048582995951417,
|
|
"calib/gap": 0.006826923076923008,
|
|
"calib/mean_conf": 0.4902024291497976,
|
|
"calib/mu_c": 0.493076923076923,
|
|
"calib/mu_w": 0.48625,
|
|
"calib/nonempty_final_conf_rate": 0.96484375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.02206477732793522,
|
|
"calib/std_conf": 0.1246385848612078,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.3904625199362042,
|
|
"calib/step_q_c_n": 1254.0,
|
|
"calib/step_q_gap": -0.022290883997230004,
|
|
"calib/step_q_w": 0.4127534039334342,
|
|
"calib/step_q_w_n": 1322.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2529.0,
|
|
"completions/max_terminated_length": 2529.0,
|
|
"completions/mean_length": 605.62890625,
|
|
"completions/mean_terminated_length": 608.0039672851562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 196.0,
|
|
"epoch": 0.04053333333333333,
|
|
"grad_norm": 3.511319875717163,
|
|
"learning_rate": 4.527777777777778e-06,
|
|
"loss": 0.0789,
|
|
"num_tokens": 9331401.0,
|
|
"reward": 1.041015625,
|
|
"reward_std": 0.2475888729095459,
|
|
"rewards/accuracy_reward_step": 0.55859375,
|
|
"rewards/format_reward_step": 0.96484375,
|
|
"step": 38
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 9.88671875,
|
|
"calib/ece": 0.19262948207171324,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.00398406374501992,
|
|
"calib/gap": -0.033144950287807395,
|
|
"calib/mean_conf": 0.4936653386454184,
|
|
"calib/mu_c": 0.4799319727891157,
|
|
"calib/mu_w": 0.5130769230769231,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.05031872509960161,
|
|
"calib/std_conf": 0.1339386178083945,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.3815764705882353,
|
|
"calib/step_q_c_n": 1275.0,
|
|
"calib/step_q_gap": -0.0407005994754589,
|
|
"calib/step_q_w": 0.4222770700636942,
|
|
"calib/step_q_w_n": 1256.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2517.0,
|
|
"completions/max_terminated_length": 2517.0,
|
|
"completions/mean_length": 606.2890625,
|
|
"completions/mean_terminated_length": 608.6666870117188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 196.0,
|
|
"epoch": 0.0416,
|
|
"grad_norm": 30.61140251159668,
|
|
"learning_rate": 4.5e-06,
|
|
"loss": 0.0292,
|
|
"num_tokens": 9592699.0,
|
|
"reward": 1.064453125,
|
|
"reward_std": 0.2312648594379425,
|
|
"rewards/accuracy_reward_step": 0.57421875,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 39
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 10.171875,
|
|
"calib/ece": 0.12492125984251967,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.003937007874015748,
|
|
"calib/gap": -0.005425373134328337,
|
|
"calib/mean_conf": 0.5126377952755905,
|
|
"calib/mu_c": 0.5100746268656716,
|
|
"calib/mu_w": 0.5155,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.055,
|
|
"calib/std_conf": 0.11670227070987084,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.4015594405594406,
|
|
"calib/step_q_c_n": 1430.0,
|
|
"calib/step_q_gap": -0.003057254500184625,
|
|
"calib/step_q_w": 0.4046166950596252,
|
|
"calib/step_q_w_n": 1174.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2514.0,
|
|
"completions/max_terminated_length": 2514.0,
|
|
"completions/mean_length": 622.18359375,
|
|
"completions/mean_terminated_length": 622.18359375,
|
|
"completions/min_length": 188.0,
|
|
"completions/min_terminated_length": 188.0,
|
|
"epoch": 0.042666666666666665,
|
|
"grad_norm": 3.3288333415985107,
|
|
"learning_rate": 4.472222222222223e-06,
|
|
"loss": 0.0055,
|
|
"num_tokens": 9858738.0,
|
|
"reward": 1.01953125,
|
|
"reward_std": 0.3156605362892151,
|
|
"rewards/accuracy_reward_step": 0.5234375,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 40
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 9.5390625,
|
|
"calib/ece": 0.299251968503937,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.003937007874015748,
|
|
"calib/gap": -0.015975111996017755,
|
|
"calib/mean_conf": 0.520984251968504,
|
|
"calib/mu_c": 0.5179024390243904,
|
|
"calib/mu_w": 0.5338775510204081,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.006574803149606293,
|
|
"calib/std_conf": 0.10996874890736674,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.4183542234332425,
|
|
"calib/step_q_c_n": 1835.0,
|
|
"calib/step_q_gap": -0.0306737831565434,
|
|
"calib/step_q_w": 0.4490280065897859,
|
|
"calib/step_q_w_n": 607.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2500.0,
|
|
"completions/max_terminated_length": 2500.0,
|
|
"completions/mean_length": 538.73828125,
|
|
"completions/mean_terminated_length": 540.8510131835938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 213.0,
|
|
"epoch": 0.04373333333333333,
|
|
"grad_norm": 1.267921805381775,
|
|
"learning_rate": 4.444444444444444e-06,
|
|
"loss": 0.0312,
|
|
"num_tokens": 10103903.0,
|
|
"reward": 1.296875,
|
|
"reward_std": 0.20018117129802704,
|
|
"rewards/accuracy_reward_step": 0.80078125,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 41
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 8.76953125,
|
|
"calib/ece": 0.13515748031496067,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.012499834491890072,
|
|
"calib/mean_conf": 0.5383858267716536,
|
|
"calib/mu_c": 0.5337106918238993,
|
|
"calib/mu_w": 0.5462105263157894,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.023779527559055134,
|
|
"calib/std_conf": 0.09652705934061376,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.44076308139534887,
|
|
"calib/step_q_c_n": 1376.0,
|
|
"calib/step_q_gap": -0.036532660837102215,
|
|
"calib/step_q_w": 0.4772957422324511,
|
|
"calib/step_q_w_n": 869.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2249.0,
|
|
"completions/max_terminated_length": 2249.0,
|
|
"completions/mean_length": 483.89453125,
|
|
"completions/mean_terminated_length": 485.79217529296875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 191.0,
|
|
"epoch": 0.0448,
|
|
"grad_norm": 1.4127931594848633,
|
|
"learning_rate": 4.416666666666667e-06,
|
|
"loss": 0.0278,
|
|
"num_tokens": 10332148.0,
|
|
"reward": 1.1171875,
|
|
"reward_std": 0.22029992938041687,
|
|
"rewards/accuracy_reward_step": 0.62109375,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 42
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 9.0703125,
|
|
"calib/ece": 0.19654901960784313,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.00392156862745098,
|
|
"calib/gap": 0.007513745946708117,
|
|
"calib/mean_conf": 0.5039999999999999,
|
|
"calib/mu_c": 0.5064161849710983,
|
|
"calib/mu_w": 0.4989024390243902,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.011058823529411763,
|
|
"calib/std_conf": 0.11128694583699156,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.4124197860962567,
|
|
"calib/step_q_c_n": 1496.0,
|
|
"calib/step_q_gap": 0.003594120236692422,
|
|
"calib/step_q_w": 0.40882566585956426,
|
|
"calib/step_q_w_n": 826.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2421.0,
|
|
"completions/max_terminated_length": 2421.0,
|
|
"completions/mean_length": 546.68359375,
|
|
"completions/mean_terminated_length": 550.9881591796875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 190.0,
|
|
"epoch": 0.04586666666666667,
|
|
"grad_norm": 10.022756576538086,
|
|
"learning_rate": 4.388888888888889e-06,
|
|
"loss": 0.0455,
|
|
"num_tokens": 10577323.0,
|
|
"reward": 1.173828125,
|
|
"reward_std": 0.2660510838031769,
|
|
"rewards/accuracy_reward_step": 0.67578125,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 43
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 10.40625,
|
|
"calib/ece": 0.14528,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.015833333333333366,
|
|
"calib/mean_conf": 0.4828,
|
|
"calib/mu_c": 0.47646666666666665,
|
|
"calib/mu_w": 0.4923,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.014040000000000009,
|
|
"calib/std_conf": 0.1054900943216945,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.36952823920265787,
|
|
"calib/step_q_c_n": 1505.0,
|
|
"calib/step_q_gap": -0.03165381429173386,
|
|
"calib/step_q_w": 0.40118205349439173,
|
|
"calib/step_q_w_n": 1159.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2506.0,
|
|
"completions/max_terminated_length": 2506.0,
|
|
"completions/mean_length": 639.3671875,
|
|
"completions/mean_terminated_length": 641.8745727539062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 201.0,
|
|
"epoch": 0.046933333333333334,
|
|
"grad_norm": 0.5156874656677246,
|
|
"learning_rate": 4.361111111111112e-06,
|
|
"loss": 0.0377,
|
|
"num_tokens": 10847321.0,
|
|
"reward": 1.07421875,
|
|
"reward_std": 0.21972496807575226,
|
|
"rewards/accuracy_reward_step": 0.5859375,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 44
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 9.48046875,
|
|
"calib/ece": 0.20632812500000003,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.017586580086579984,
|
|
"calib/mean_conf": 0.454609375,
|
|
"calib/mu_c": 0.4606547619047619,
|
|
"calib/mu_w": 0.4430681818181819,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0023437500000000003,
|
|
"calib/std_conf": 0.10552993017201033,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.3628175797712221,
|
|
"calib/step_q_c_n": 1661.0,
|
|
"calib/step_q_gap": -0.0009291565212061204,
|
|
"calib/step_q_w": 0.3637467362924282,
|
|
"calib/step_q_w_n": 766.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1306.0,
|
|
"completions/max_terminated_length": 1306.0,
|
|
"completions/mean_length": 534.0234375,
|
|
"completions/mean_terminated_length": 538.2283325195312,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 167.0,
|
|
"epoch": 0.048,
|
|
"grad_norm": 0.6220444440841675,
|
|
"learning_rate": 4.333333333333334e-06,
|
|
"loss": 0.0041,
|
|
"num_tokens": 11089079.0,
|
|
"reward": 1.15625,
|
|
"reward_std": 0.25460314750671387,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 45
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 9.71484375,
|
|
"calib/ece": 0.21824000000000007,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.004,
|
|
"calib/gap": 0.0035470795401687516,
|
|
"calib/mean_conf": 0.40152,
|
|
"calib/mu_c": 0.4029530201342282,
|
|
"calib/mu_w": 0.3994059405940594,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.011880000000000002,
|
|
"calib/std_conf": 0.10389268309173653,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.31864343958487773,
|
|
"calib/step_q_c_n": 1349.0,
|
|
"calib/step_q_gap": -0.04035480294587801,
|
|
"calib/step_q_w": 0.35899824253075574,
|
|
"calib/step_q_w_n": 1138.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2542.0,
|
|
"completions/max_terminated_length": 2542.0,
|
|
"completions/mean_length": 616.13671875,
|
|
"completions/mean_terminated_length": 616.13671875,
|
|
"completions/min_length": 173.0,
|
|
"completions/min_terminated_length": 173.0,
|
|
"epoch": 0.04906666666666667,
|
|
"grad_norm": 0.6727369427680969,
|
|
"learning_rate": 4.305555555555556e-06,
|
|
"loss": 0.0382,
|
|
"num_tokens": 11351578.0,
|
|
"reward": 1.0703125,
|
|
"reward_std": 0.2612079977989197,
|
|
"rewards/accuracy_reward_step": 0.58203125,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 46
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96484375,
|
|
"calib/avg_num_step_conf": 9.99609375,
|
|
"calib/ece": 0.3606882591093118,
|
|
"calib/final_conf_rate": 0.96484375,
|
|
"calib/format_rate": 0.96484375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0037929717341482205,
|
|
"calib/mean_conf": 0.3420647773279353,
|
|
"calib/mu_c": 0.34088235294117647,
|
|
"calib/mu_w": 0.3446753246753247,
|
|
"calib/nonempty_final_conf_rate": 0.96484375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.007246963562753035,
|
|
"calib/std_conf": 0.11379445500608808,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.25505913272010516,
|
|
"calib/step_q_c_n": 1522.0,
|
|
"calib/step_q_gap": -0.018113480587512942,
|
|
"calib/step_q_w": 0.2731726133076181,
|
|
"calib/step_q_w_n": 1037.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2524.0,
|
|
"completions/max_terminated_length": 2524.0,
|
|
"completions/mean_length": 641.0078125,
|
|
"completions/mean_terminated_length": 646.0551147460938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 210.0,
|
|
"epoch": 0.050133333333333335,
|
|
"grad_norm": 0.40471020340919495,
|
|
"learning_rate": 4.277777777777778e-06,
|
|
"loss": 0.0425,
|
|
"num_tokens": 11621652.0,
|
|
"reward": 1.146484375,
|
|
"reward_std": 0.12448950111865997,
|
|
"rewards/accuracy_reward_step": 0.6640625,
|
|
"rewards/format_reward_step": 0.96484375,
|
|
"step": 47
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 8.5859375,
|
|
"calib/ece": 0.29771653543307086,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.003937007874015748,
|
|
"calib/gap": -0.040253488842445895,
|
|
"calib/mean_conf": 0.3725196850393701,
|
|
"calib/mu_c": 0.35809815950920243,
|
|
"calib/mu_w": 0.3983516483516483,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.014251968503937012,
|
|
"calib/std_conf": 0.12516426060255484,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.26928518242740135,
|
|
"calib/step_q_c_n": 1343.0,
|
|
"calib/step_q_gap": -0.053159262017043185,
|
|
"calib/step_q_w": 0.32244444444444453,
|
|
"calib/step_q_w_n": 855.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2501.0,
|
|
"completions/max_terminated_length": 2501.0,
|
|
"completions/mean_length": 560.6015625,
|
|
"completions/mean_terminated_length": 560.6015625,
|
|
"completions/min_length": 185.0,
|
|
"completions/min_terminated_length": 185.0,
|
|
"epoch": 0.0512,
|
|
"grad_norm": 0.576022744178772,
|
|
"learning_rate": 4.25e-06,
|
|
"loss": -0.0073,
|
|
"num_tokens": 11868854.0,
|
|
"reward": 1.1328125,
|
|
"reward_std": 0.27789777517318726,
|
|
"rewards/accuracy_reward_step": 0.63671875,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 48
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 9.2109375,
|
|
"calib/ece": 0.3026086956521739,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.003952569169960474,
|
|
"calib/gap": 0.0026007326007325693,
|
|
"calib/mean_conf": 0.40205533596837945,
|
|
"calib/mu_c": 0.40285714285714286,
|
|
"calib/mu_w": 0.4002564102564103,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.006482213438735177,
|
|
"calib/std_conf": 0.1381774540062094,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.33195707070707065,
|
|
"calib/step_q_c_n": 1584.0,
|
|
"calib/step_q_gap": 0.018249060371153314,
|
|
"calib/step_q_w": 0.31370801033591733,
|
|
"calib/step_q_w_n": 774.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2518.0,
|
|
"completions/max_terminated_length": 2518.0,
|
|
"completions/mean_length": 557.3828125,
|
|
"completions/mean_terminated_length": 559.5686645507812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 189.0,
|
|
"epoch": 0.05226666666666667,
|
|
"grad_norm": 0.3916730582714081,
|
|
"learning_rate": 4.222222222222223e-06,
|
|
"loss": 0.0563,
|
|
"num_tokens": 12116080.0,
|
|
"reward": 1.177734375,
|
|
"reward_std": 0.2767038941383362,
|
|
"rewards/accuracy_reward_step": 0.68359375,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 49
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 10.34765625,
|
|
"calib/ece": 0.3345454545454546,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.007905138339920948,
|
|
"calib/gap": 0.01421337897589492,
|
|
"calib/mean_conf": 0.35667984189723323,
|
|
"calib/mu_c": 0.36128654970760227,
|
|
"calib/mu_w": 0.34707317073170735,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.007667984189723321,
|
|
"calib/std_conf": 0.15815651260389624,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.27458479532163743,
|
|
"calib/step_q_c_n": 1710.0,
|
|
"calib/step_q_gap": -0.02663991181361286,
|
|
"calib/step_q_w": 0.3012247071352503,
|
|
"calib/step_q_w_n": 939.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2393.0,
|
|
"completions/max_terminated_length": 2393.0,
|
|
"completions/mean_length": 591.53515625,
|
|
"completions/mean_terminated_length": 593.8549194335938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 214.0,
|
|
"epoch": 0.05333333333333334,
|
|
"grad_norm": 0.35859617590904236,
|
|
"learning_rate": 4.194444444444445e-06,
|
|
"loss": 0.0395,
|
|
"num_tokens": 12372873.0,
|
|
"reward": 1.162109375,
|
|
"reward_std": 0.17702001333236694,
|
|
"rewards/accuracy_reward_step": 0.66796875,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 50
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 10.19921875,
|
|
"calib/ece": 0.4156299212598425,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.011811023622047244,
|
|
"calib/gap": -0.022006750311835077,
|
|
"calib/mean_conf": 0.30271653543307087,
|
|
"calib/mu_c": 0.296045197740113,
|
|
"calib/mu_w": 0.3180519480519481,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.010748031496062994,
|
|
"calib/std_conf": 0.1506184833340684,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.21695022054190297,
|
|
"calib/step_q_c_n": 1587.0,
|
|
"calib/step_q_gap": -0.02467087320809705,
|
|
"calib/step_q_w": 0.24162109375000002,
|
|
"calib/step_q_w_n": 1024.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2531.0,
|
|
"completions/max_terminated_length": 2531.0,
|
|
"completions/mean_length": 612.70703125,
|
|
"completions/mean_terminated_length": 615.10986328125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 170.0,
|
|
"epoch": 0.0544,
|
|
"grad_norm": 0.27484986186027527,
|
|
"learning_rate": 4.166666666666667e-06,
|
|
"loss": 0.03,
|
|
"num_tokens": 12639022.0,
|
|
"reward": 1.185546875,
|
|
"reward_std": 0.27396029233932495,
|
|
"rewards/accuracy_reward_step": 0.69140625,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 51
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 9.3515625,
|
|
"calib/ece": 0.5295238095238094,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.003968253968253968,
|
|
"calib/gap": 0.014234817813765177,
|
|
"calib/mean_conf": 0.25785714285714284,
|
|
"calib/mu_c": 0.26107692307692304,
|
|
"calib/mu_w": 0.24684210526315786,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0067857142857142855,
|
|
"calib/std_conf": 0.14199881031726735,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.17987912087912086,
|
|
"calib/step_q_c_n": 1820.0,
|
|
"calib/step_q_gap": -0.00870972929509517,
|
|
"calib/step_q_w": 0.18858885017421603,
|
|
"calib/step_q_w_n": 574.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2515.0,
|
|
"completions/max_terminated_length": 2515.0,
|
|
"completions/mean_length": 552.58203125,
|
|
"completions/mean_terminated_length": 554.7490234375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 181.0,
|
|
"epoch": 0.055466666666666664,
|
|
"grad_norm": 0.31955885887145996,
|
|
"learning_rate": 4.138888888888889e-06,
|
|
"loss": 0.0232,
|
|
"num_tokens": 12888435.0,
|
|
"reward": 1.25390625,
|
|
"reward_std": 0.18764840066432953,
|
|
"rewards/accuracy_reward_step": 0.76171875,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 52
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 10.02734375,
|
|
"calib/ece": 0.4562890625,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.00390625,
|
|
"calib/gap": 0.02520731612803223,
|
|
"calib/mean_conf": 0.2816015625,
|
|
"calib/mu_c": 0.2883957219251337,
|
|
"calib/mu_w": 0.26318840579710145,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0037109375,
|
|
"calib/std_conf": 0.13509897065691728,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.2138771071234367,
|
|
"calib/step_q_c_n": 1839.0,
|
|
"calib/step_q_gap": 0.017352381848711435,
|
|
"calib/step_q_w": 0.19652472527472525,
|
|
"calib/step_q_w_n": 728.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1384.0,
|
|
"completions/max_terminated_length": 1384.0,
|
|
"completions/mean_length": 582.15234375,
|
|
"completions/mean_terminated_length": 586.7362060546875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 182.0,
|
|
"epoch": 0.05653333333333333,
|
|
"grad_norm": 0.3678324520587921,
|
|
"learning_rate": 4.111111111111111e-06,
|
|
"loss": 0.0045,
|
|
"num_tokens": 13143290.0,
|
|
"reward": 1.228515625,
|
|
"reward_std": 0.23388169705867767,
|
|
"rewards/accuracy_reward_step": 0.73046875,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 53
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 8.15234375,
|
|
"calib/ece": 0.4665748031496063,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.007874015748031496,
|
|
"calib/gap": 0.0041525810872544144,
|
|
"calib/mean_conf": 0.3168897637795276,
|
|
"calib/mu_c": 0.31778894472361807,
|
|
"calib/mu_w": 0.31363636363636366,
|
|
"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": 0.1755962050217898,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.240630068621335,
|
|
"calib/step_q_c_n": 1603.0,
|
|
"calib/step_q_gap": -0.00860546856874761,
|
|
"calib/step_q_w": 0.24923553719008262,
|
|
"calib/step_q_w_n": 484.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2513.0,
|
|
"completions/max_terminated_length": 2513.0,
|
|
"completions/mean_length": 522.5,
|
|
"completions/mean_terminated_length": 522.5,
|
|
"completions/min_length": 131.0,
|
|
"completions/min_terminated_length": 131.0,
|
|
"epoch": 0.0576,
|
|
"grad_norm": 0.469553679227829,
|
|
"learning_rate": 4.083333333333334e-06,
|
|
"loss": 0.0049,
|
|
"num_tokens": 13383282.0,
|
|
"reward": 1.271484375,
|
|
"reward_std": 0.18993809819221497,
|
|
"rewards/accuracy_reward_step": 0.77734375,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 54
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 9.60546875,
|
|
"calib/ece": 0.3109523809523809,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.007936507936507936,
|
|
"calib/gap": 0.030600598603259088,
|
|
"calib/mean_conf": 0.3242857142857143,
|
|
"calib/mu_c": 0.33606451612903226,
|
|
"calib/mu_w": 0.3054639175257732,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.010079365079365077,
|
|
"calib/std_conf": 0.1646951419924655,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.27128185907046476,
|
|
"calib/step_q_c_n": 1334.0,
|
|
"calib/step_q_gap": 0.014295192403798063,
|
|
"calib/step_q_w": 0.2569866666666667,
|
|
"calib/step_q_w_n": 1125.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2531.0,
|
|
"completions/max_terminated_length": 2531.0,
|
|
"completions/mean_length": 572.7109375,
|
|
"completions/mean_terminated_length": 572.7109375,
|
|
"completions/min_length": 145.0,
|
|
"completions/min_terminated_length": 145.0,
|
|
"epoch": 0.058666666666666666,
|
|
"grad_norm": 0.22412878274917603,
|
|
"learning_rate": 4.055555555555556e-06,
|
|
"loss": 0.0107,
|
|
"num_tokens": 13637720.0,
|
|
"reward": 1.09765625,
|
|
"reward_std": 0.27271413803100586,
|
|
"rewards/accuracy_reward_step": 0.60546875,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 55
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 9.125,
|
|
"calib/ece": 0.27885826771653544,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.00040517534899559005,
|
|
"calib/mean_conf": 0.37074803149606295,
|
|
"calib/mu_c": 0.37060606060606055,
|
|
"calib/mu_w": 0.37101123595505614,
|
|
"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": 0.15235626857751916,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2981626928471248,
|
|
"calib/step_q_c_n": 1426.0,
|
|
"calib/step_q_gap": -0.017123021438589514,
|
|
"calib/step_q_w": 0.31528571428571434,
|
|
"calib/step_q_w_n": 910.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/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": 555.58984375,
|
|
"completions/mean_terminated_length": 557.7686767578125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 172.0,
|
|
"epoch": 0.05973333333333333,
|
|
"grad_norm": 0.5936625003814697,
|
|
"learning_rate": 4.027777777777779e-06,
|
|
"loss": 0.0123,
|
|
"num_tokens": 13886791.0,
|
|
"reward": 1.140625,
|
|
"reward_std": 0.2601584792137146,
|
|
"rewards/accuracy_reward_step": 0.64453125,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 56
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 9.5703125,
|
|
"calib/ece": 0.43661417322834645,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.003937007874015748,
|
|
"calib/gap": 0.020655241935483903,
|
|
"calib/mean_conf": 0.3246456692913386,
|
|
"calib/mu_c": 0.32968749999999997,
|
|
"calib/mu_w": 0.30903225806451606,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0026771653543307085,
|
|
"calib/std_conf": 0.1510295130755166,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.2760599208592425,
|
|
"calib/step_q_c_n": 1769.0,
|
|
"calib/step_q_gap": 0.03532570646864047,
|
|
"calib/step_q_w": 0.24073421439060205,
|
|
"calib/step_q_w_n": 681.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2506.0,
|
|
"completions/max_terminated_length": 2506.0,
|
|
"completions/mean_length": 571.30859375,
|
|
"completions/mean_terminated_length": 571.30859375,
|
|
"completions/min_length": 201.0,
|
|
"completions/min_terminated_length": 201.0,
|
|
"epoch": 0.0608,
|
|
"grad_norm": 0.3216277062892914,
|
|
"learning_rate": 4.000000000000001e-06,
|
|
"loss": 0.0214,
|
|
"num_tokens": 14139838.0,
|
|
"reward": 1.24609375,
|
|
"reward_std": 0.2393544316291809,
|
|
"rewards/accuracy_reward_step": 0.75,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 57
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/avg_num_step_conf": 9.72265625,
|
|
"calib/ece": 0.39502008032128516,
|
|
"calib/final_conf_rate": 0.97265625,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.004016064257028112,
|
|
"calib/gap": -0.017761857707509876,
|
|
"calib/mean_conf": 0.2745381526104418,
|
|
"calib/mu_c": 0.26826086956521744,
|
|
"calib/mu_w": 0.2860227272727273,
|
|
"calib/nonempty_final_conf_rate": 0.97265625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.011485943775100402,
|
|
"calib/std_conf": 0.1462183530795982,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.1948092105263158,
|
|
"calib/step_q_c_n": 1520.0,
|
|
"calib/step_q_gap": -0.02595446336429308,
|
|
"calib/step_q_w": 0.22076367389060889,
|
|
"calib/step_q_w_n": 969.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 3047.0,
|
|
"completions/max_terminated_length": 3047.0,
|
|
"completions/mean_length": 662.15234375,
|
|
"completions/mean_terminated_length": 662.15234375,
|
|
"completions/min_length": 165.0,
|
|
"completions/min_terminated_length": 165.0,
|
|
"epoch": 0.06186666666666667,
|
|
"grad_norm": 0.30427154898643494,
|
|
"learning_rate": 3.972222222222223e-06,
|
|
"loss": 0.03,
|
|
"num_tokens": 14415669.0,
|
|
"reward": 1.11328125,
|
|
"reward_std": 0.3100544214248657,
|
|
"rewards/accuracy_reward_step": 0.62890625,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 58
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 8.84375,
|
|
"calib/ece": 0.4424007936507937,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.022777144079275558,
|
|
"calib/mean_conf": 0.25617063492063497,
|
|
"calib/mu_c": 0.24866863905325448,
|
|
"calib/mu_w": 0.27144578313253004,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.01396825396825397,
|
|
"calib/std_conf": 0.1438725966205067,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.18537381916329285,
|
|
"calib/step_q_c_n": 1482.0,
|
|
"calib/step_q_gap": -0.03044459515895781,
|
|
"calib/step_q_w": 0.21581841432225066,
|
|
"calib/step_q_w_n": 782.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0,
|
|
"completions/max_length": 2516.0,
|
|
"completions/max_terminated_length": 2516.0,
|
|
"completions/mean_length": 588.91015625,
|
|
"completions/mean_terminated_length": 588.91015625,
|
|
"completions/min_length": 150.0,
|
|
"completions/min_terminated_length": 150.0,
|
|
"epoch": 0.06293333333333333,
|
|
"grad_norm": 0.2894952595233917,
|
|
"learning_rate": 3.944444444444445e-06,
|
|
"loss": 0.0276,
|
|
"num_tokens": 14672678.0,
|
|
"reward": 1.150390625,
|
|
"reward_std": 0.2913172245025635,
|
|
"rewards/accuracy_reward_step": 0.66015625,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 59
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 8.9296875,
|
|
"calib/ece": 0.4264,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.03491341256366723,
|
|
"calib/mean_conf": 0.22688,
|
|
"calib/mu_c": 0.21361290322580645,
|
|
"calib/mu_w": 0.24852631578947368,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.016640000000000002,
|
|
"calib/std_conf": 0.13409200423589768,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.14738325281803544,
|
|
"calib/step_q_c_n": 1242.0,
|
|
"calib/step_q_gap": -0.033248931090010564,
|
|
"calib/step_q_w": 0.180632183908046,
|
|
"calib/step_q_w_n": 1044.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2520.0,
|
|
"completions/max_terminated_length": 2520.0,
|
|
"completions/mean_length": 568.3046875,
|
|
"completions/mean_terminated_length": 572.779541015625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 154.0,
|
|
"epoch": 0.064,
|
|
"grad_norm": 0.6230308413505554,
|
|
"learning_rate": 3.916666666666667e-06,
|
|
"loss": 0.0147,
|
|
"num_tokens": 14927020.0,
|
|
"reward": 1.09375,
|
|
"reward_std": 0.26803910732269287,
|
|
"rewards/accuracy_reward_step": 0.60546875,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 60
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 9.0859375,
|
|
"calib/ece": 0.5171259842519687,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.00935286935286933,
|
|
"calib/mean_conf": 0.19940944881889766,
|
|
"calib/mu_c": 0.19675824175824178,
|
|
"calib/mu_w": 0.2061111111111111,
|
|
"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": 0.08557842600406948,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.14423172242874843,
|
|
"calib/step_q_c_n": 1614.0,
|
|
"calib/step_q_gap": -0.007060412402712241,
|
|
"calib/step_q_w": 0.15129213483146067,
|
|
"calib/step_q_w_n": 712.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0078125,
|
|
"completions/max_length": 2509.0,
|
|
"completions/max_terminated_length": 2509.0,
|
|
"completions/mean_length": 518.34375,
|
|
"completions/mean_terminated_length": 522.4251708984375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 158.0,
|
|
"epoch": 0.06506666666666666,
|
|
"grad_norm": 0.5418161749839783,
|
|
"learning_rate": 3.88888888888889e-06,
|
|
"loss": -0.0156,
|
|
"num_tokens": 15163780.0,
|
|
"reward": 1.20703125,
|
|
"reward_std": 0.16414855420589447,
|
|
"rewards/accuracy_reward_step": 0.7109375,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 61
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 10.8828125,
|
|
"calib/ece": 0.4805952380952381,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.01922619047619048,
|
|
"calib/mean_conf": 0.19043650793650793,
|
|
"calib/mu_c": 0.1968452380952381,
|
|
"calib/mu_w": 0.17761904761904762,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0021825396825396826,
|
|
"calib/std_conf": 0.09542411561955262,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.12830283353010627,
|
|
"calib/step_q_c_n": 1694.0,
|
|
"calib/step_q_gap": -0.028620243392970673,
|
|
"calib/step_q_w": 0.15692307692307694,
|
|
"calib/step_q_w_n": 1092.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2529.0,
|
|
"completions/max_terminated_length": 2529.0,
|
|
"completions/mean_length": 628.1875,
|
|
"completions/mean_terminated_length": 633.1338500976562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 118.0,
|
|
"epoch": 0.06613333333333334,
|
|
"grad_norm": 0.34490272402763367,
|
|
"learning_rate": 3.861111111111112e-06,
|
|
"loss": 0.0182,
|
|
"num_tokens": 15431676.0,
|
|
"reward": 1.1484375,
|
|
"reward_std": 0.3012402057647705,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 62
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 10.375,
|
|
"calib/ece": 0.5203571428571427,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0023104002448917083,
|
|
"calib/mean_conf": 0.1987698412698413,
|
|
"calib/mu_c": 0.19810055865921788,
|
|
"calib/mu_w": 0.2004109589041096,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.004404761904761904,
|
|
"calib/std_conf": 0.06918226016976482,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.14131530424093425,
|
|
"calib/step_q_c_n": 1627.0,
|
|
"calib/step_q_gap": -0.03347575503992095,
|
|
"calib/step_q_w": 0.1747910592808552,
|
|
"calib/step_q_w_n": 1029.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2505.0,
|
|
"completions/max_terminated_length": 2505.0,
|
|
"completions/mean_length": 668.8046875,
|
|
"completions/mean_terminated_length": 674.0708618164062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 150.0,
|
|
"epoch": 0.0672,
|
|
"grad_norm": 0.457501083612442,
|
|
"learning_rate": 3.833333333333334e-06,
|
|
"loss": 0.0236,
|
|
"num_tokens": 15711530.0,
|
|
"reward": 1.1875,
|
|
"reward_std": 0.19891968369483948,
|
|
"rewards/accuracy_reward_step": 0.69921875,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 63
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 9.96875,
|
|
"calib/ece": 0.5169322709163348,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.00398406374501992,
|
|
"calib/gap": -0.011759080681452844,
|
|
"calib/mean_conf": 0.22657370517928288,
|
|
"calib/mu_c": 0.22338797814207656,
|
|
"calib/mu_w": 0.2351470588235294,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.007211155378486055,
|
|
"calib/std_conf": 0.09112080739138619,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.17443543717429066,
|
|
"calib/step_q_c_n": 1727.0,
|
|
"calib/step_q_gap": -0.05732213858328508,
|
|
"calib/step_q_w": 0.23175757575757575,
|
|
"calib/step_q_w_n": 825.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2508.0,
|
|
"completions/max_terminated_length": 2508.0,
|
|
"completions/mean_length": 547.78515625,
|
|
"completions/mean_terminated_length": 554.2806396484375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 183.0,
|
|
"epoch": 0.06826666666666667,
|
|
"grad_norm": 0.4879423975944519,
|
|
"learning_rate": 3.8055555555555556e-06,
|
|
"loss": -0.0166,
|
|
"num_tokens": 15955539.0,
|
|
"reward": 1.205078125,
|
|
"reward_std": 0.19858914613723755,
|
|
"rewards/accuracy_reward_step": 0.71484375,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 64
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 8.99609375,
|
|
"calib/ece": 0.41180392156862755,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.0021226602947032602,
|
|
"calib/mean_conf": 0.22349019607843137,
|
|
"calib/mu_c": 0.22271604938271605,
|
|
"calib/mu_w": 0.2248387096774193,
|
|
"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": 0.062490586757486026,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.1703058994901675,
|
|
"calib/step_q_c_n": 1373.0,
|
|
"calib/step_q_gap": -0.005059691907681962,
|
|
"calib/step_q_w": 0.17536559139784946,
|
|
"calib/step_q_w_n": 930.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1615.0,
|
|
"completions/max_terminated_length": 1615.0,
|
|
"completions/mean_length": 524.0390625,
|
|
"completions/mean_terminated_length": 528.1653442382812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 200.0,
|
|
"epoch": 0.06933333333333333,
|
|
"grad_norm": 0.32161787152290344,
|
|
"learning_rate": 3.777777777777778e-06,
|
|
"loss": -0.0061,
|
|
"num_tokens": 16194717.0,
|
|
"reward": 1.130859375,
|
|
"reward_std": 0.13571283221244812,
|
|
"rewards/accuracy_reward_step": 0.6328125,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 65
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.953125,
|
|
"calib/avg_num_step_conf": 12.6796875,
|
|
"calib/ece": 0.2913524590163934,
|
|
"calib/final_conf_rate": 0.953125,
|
|
"calib/format_rate": 0.953125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.016152597402597335,
|
|
"calib/mean_conf": 0.2496311475409836,
|
|
"calib/mu_c": 0.2570454545454545,
|
|
"calib/mu_w": 0.2408928571428572,
|
|
"calib/nonempty_final_conf_rate": 0.953125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.06874956939592984,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.19214477211796246,
|
|
"calib/step_q_c_n": 1492.0,
|
|
"calib/step_q_gap": -0.03424063267109115,
|
|
"calib/step_q_w": 0.2263854047890536,
|
|
"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": 0.015625,
|
|
"completions/max_length": 2557.0,
|
|
"completions/max_terminated_length": 2557.0,
|
|
"completions/mean_length": 726.1171875,
|
|
"completions/mean_terminated_length": 737.6428833007812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 154.0,
|
|
"epoch": 0.0704,
|
|
"grad_norm": 2.085747480392456,
|
|
"learning_rate": 3.7500000000000005e-06,
|
|
"loss": 0.0175,
|
|
"num_tokens": 16486955.0,
|
|
"reward": 0.9921875,
|
|
"reward_std": 0.2367658019065857,
|
|
"rewards/accuracy_reward_step": 0.515625,
|
|
"rewards/format_reward_step": 0.953125,
|
|
"step": 66
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 11.07421875,
|
|
"calib/ece": 0.3971259842519685,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.0179152823920265,
|
|
"calib/mean_conf": 0.2642913385826772,
|
|
"calib/mu_c": 0.27035714285714285,
|
|
"calib/mu_w": 0.25244186046511635,
|
|
"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": 0.08813068412326645,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.19979757085020244,
|
|
"calib/step_q_c_n": 1729.0,
|
|
"calib/step_q_gap": 0.009191784231757594,
|
|
"calib/step_q_w": 0.19060578661844485,
|
|
"calib/step_q_w_n": 1106.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2467.0,
|
|
"completions/max_terminated_length": 2467.0,
|
|
"completions/mean_length": 672.140625,
|
|
"completions/mean_terminated_length": 677.4330444335938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 170.0,
|
|
"epoch": 0.07146666666666666,
|
|
"grad_norm": 0.5508794784545898,
|
|
"learning_rate": 3.7222222222222225e-06,
|
|
"loss": 0.0023,
|
|
"num_tokens": 16764031.0,
|
|
"reward": 1.15234375,
|
|
"reward_std": 0.16439500451087952,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 67
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 11.3203125,
|
|
"calib/ece": 0.28082677165354336,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.014592363261093921,
|
|
"calib/mean_conf": 0.331771653543307,
|
|
"calib/mu_c": 0.33763157894736845,
|
|
"calib/mu_w": 0.32303921568627453,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0070866141732283455,
|
|
"calib/std_conf": 0.13205973886673128,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.26028397565922917,
|
|
"calib/step_q_c_n": 1479.0,
|
|
"calib/step_q_gap": 0.020410825553520917,
|
|
"calib/step_q_w": 0.23987315010570825,
|
|
"calib/step_q_w_n": 1419.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2425.0,
|
|
"completions/max_terminated_length": 2425.0,
|
|
"completions/mean_length": 642.328125,
|
|
"completions/mean_terminated_length": 647.3858032226562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 152.0,
|
|
"epoch": 0.07253333333333334,
|
|
"grad_norm": 0.3224004805088043,
|
|
"learning_rate": 3.694444444444445e-06,
|
|
"loss": -0.0131,
|
|
"num_tokens": 17032555.0,
|
|
"reward": 1.08984375,
|
|
"reward_std": 0.16255776584148407,
|
|
"rewards/accuracy_reward_step": 0.59375,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 68
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 11.8828125,
|
|
"calib/ece": 0.2305882352941176,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.037564408696745044,
|
|
"calib/mean_conf": 0.4050980392156863,
|
|
"calib/mu_c": 0.3890410958904109,
|
|
"calib/mu_w": 0.42660550458715596,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.031568627450980366,
|
|
"calib/std_conf": 0.14982044873414452,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.33201261829652995,
|
|
"calib/step_q_c_n": 1585.0,
|
|
"calib/step_q_gap": -0.028014835375398617,
|
|
"calib/step_q_w": 0.36002745367192857,
|
|
"calib/step_q_w_n": 1457.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2507.0,
|
|
"completions/max_terminated_length": 2507.0,
|
|
"completions/mean_length": 745.046875,
|
|
"completions/mean_terminated_length": 747.9686889648438,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 227.0,
|
|
"epoch": 0.0736,
|
|
"grad_norm": 0.31790482997894287,
|
|
"learning_rate": 3.6666666666666666e-06,
|
|
"loss": -0.0037,
|
|
"num_tokens": 17327783.0,
|
|
"reward": 1.068359375,
|
|
"reward_std": 0.2722419202327728,
|
|
"rewards/accuracy_reward_step": 0.5703125,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 69
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 12.109375,
|
|
"calib/ece": 0.1654032258064516,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.012714486638537248,
|
|
"calib/mean_conf": 0.5303225806451614,
|
|
"calib/mu_c": 0.5349367088607595,
|
|
"calib/mu_w": 0.5222222222222223,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.029314516129032256,
|
|
"calib/std_conf": 0.1149118560088766,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.464349617422013,
|
|
"calib/step_q_c_n": 1699.0,
|
|
"calib/step_q_gap": 0.023021994295674697,
|
|
"calib/step_q_w": 0.4413276231263383,
|
|
"calib/step_q_w_n": 1401.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2541.0,
|
|
"completions/max_terminated_length": 2541.0,
|
|
"completions/mean_length": 732.96484375,
|
|
"completions/mean_terminated_length": 744.5992431640625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 206.0,
|
|
"epoch": 0.07466666666666667,
|
|
"grad_norm": 0.7387703657150269,
|
|
"learning_rate": 3.638888888888889e-06,
|
|
"loss": -0.0024,
|
|
"num_tokens": 17622414.0,
|
|
"reward": 1.1015625,
|
|
"reward_std": 0.17702125012874603,
|
|
"rewards/accuracy_reward_step": 0.6171875,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 70
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 11.98046875,
|
|
"calib/ece": 0.09847656249999999,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0078125,
|
|
"calib/gap": 0.0018395275136847689,
|
|
"calib/mean_conf": 0.6123046875,
|
|
"calib/mu_c": 0.6128651685393258,
|
|
"calib/mu_w": 0.611025641025641,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0077343749999999965,
|
|
"calib/std_conf": 0.0873601613467337,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5387731481481481,
|
|
"calib/step_q_c_n": 2160.0,
|
|
"calib/step_q_gap": -0.0059126291396136565,
|
|
"calib/step_q_w": 0.5446857772877618,
|
|
"calib/step_q_w_n": 907.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2022.0,
|
|
"completions/max_terminated_length": 2022.0,
|
|
"completions/mean_length": 699.64453125,
|
|
"completions/mean_terminated_length": 705.153564453125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 226.0,
|
|
"epoch": 0.07573333333333333,
|
|
"grad_norm": 0.29512184858322144,
|
|
"learning_rate": 3.6111111111111115e-06,
|
|
"loss": 0.0128,
|
|
"num_tokens": 17905931.0,
|
|
"reward": 1.1953125,
|
|
"reward_std": 0.24605843424797058,
|
|
"rewards/accuracy_reward_step": 0.6953125,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 71
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 11.57421875,
|
|
"calib/ece": 0.09527559055118104,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.009160904255319013,
|
|
"calib/mean_conf": 0.6379527559055118,
|
|
"calib/mu_c": 0.6345625,
|
|
"calib/mu_w": 0.643723404255319,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.05165354330708659,
|
|
"calib/std_conf": 0.05090269523481802,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5643253747917824,
|
|
"calib/step_q_c_n": 1801.0,
|
|
"calib/step_q_gap": -0.020614384244362238,
|
|
"calib/step_q_w": 0.5849397590361446,
|
|
"calib/step_q_w_n": 1162.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1308.0,
|
|
"completions/max_terminated_length": 1308.0,
|
|
"completions/mean_length": 616.13671875,
|
|
"completions/mean_terminated_length": 620.9881591796875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 219.0,
|
|
"epoch": 0.0768,
|
|
"grad_norm": 0.22814252972602844,
|
|
"learning_rate": 3.5833333333333335e-06,
|
|
"loss": -0.0222,
|
|
"num_tokens": 18168070.0,
|
|
"reward": 1.12109375,
|
|
"reward_std": 0.10495803505182266,
|
|
"rewards/accuracy_reward_step": 0.625,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 72
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 10.890625,
|
|
"calib/ece": 0.1998823529411764,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.00392156862745098,
|
|
"calib/gap": 0.003478260869565153,
|
|
"calib/mean_conf": 0.6293725490196079,
|
|
"calib/mu_c": 0.6299999999999999,
|
|
"calib/mu_w": 0.6265217391304347,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.004823529411764709,
|
|
"calib/std_conf": 0.048462623877688904,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5721419009370817,
|
|
"calib/step_q_c_n": 2241.0,
|
|
"calib/step_q_gap": -0.01029466213421626,
|
|
"calib/step_q_w": 0.582436563071298,
|
|
"calib/step_q_w_n": 547.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0078125,
|
|
"completions/max_length": 1722.0,
|
|
"completions/max_terminated_length": 1722.0,
|
|
"completions/mean_length": 594.7109375,
|
|
"completions/mean_terminated_length": 599.3936767578125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 210.0,
|
|
"epoch": 0.07786666666666667,
|
|
"grad_norm": 0.34402838349342346,
|
|
"learning_rate": 3.555555555555556e-06,
|
|
"loss": -0.0233,
|
|
"num_tokens": 18427348.0,
|
|
"reward": 1.314453125,
|
|
"reward_std": 0.2204178273677826,
|
|
"rewards/accuracy_reward_step": 0.81640625,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 73
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 12.546875,
|
|
"calib/ece": 0.1037549407114625,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.003952569169960474,
|
|
"calib/gap": -0.004371177370030677,
|
|
"calib/mean_conf": 0.6729249011857708,
|
|
"calib/mu_c": 0.6710416666666666,
|
|
"calib/mu_w": 0.6754128440366973,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.1037549407114625,
|
|
"calib/std_conf": 0.07039233848004939,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.606,
|
|
"calib/step_q_c_n": 1625.0,
|
|
"calib/step_q_gap": -0.00197101449275372,
|
|
"calib/step_q_w": 0.6079710144927537,
|
|
"calib/step_q_w_n": 1587.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2889.0,
|
|
"completions/max_terminated_length": 2889.0,
|
|
"completions/mean_length": 705.6328125,
|
|
"completions/mean_terminated_length": 708.4000244140625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 223.0,
|
|
"epoch": 0.07893333333333333,
|
|
"grad_norm": 0.5026135444641113,
|
|
"learning_rate": 3.5277777777777784e-06,
|
|
"loss": 0.012,
|
|
"num_tokens": 18711918.0,
|
|
"reward": 1.056640625,
|
|
"reward_std": 0.23151721060276031,
|
|
"rewards/accuracy_reward_step": 0.5625,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 74
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 10.28125,
|
|
"calib/ece": 0.14257812499999994,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.01171875,
|
|
"calib/gap": -0.009701720452000573,
|
|
"calib/mean_conf": 0.7010156249999999,
|
|
"calib/mu_c": 0.6992344497607654,
|
|
"calib/mu_w": 0.708936170212766,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.013593750000000007,
|
|
"calib/std_conf": 0.061030727964357226,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6336624513618677,
|
|
"calib/step_q_c_n": 2056.0,
|
|
"calib/step_q_gap": -0.012257687527021055,
|
|
"calib/step_q_w": 0.6459201388888888,
|
|
"calib/step_q_w_n": 576.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0078125,
|
|
"completions/max_length": 1830.0,
|
|
"completions/max_terminated_length": 1830.0,
|
|
"completions/mean_length": 599.86328125,
|
|
"completions/mean_terminated_length": 604.5866088867188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 234.0,
|
|
"epoch": 0.08,
|
|
"grad_norm": 0.38530319929122925,
|
|
"learning_rate": 3.5e-06,
|
|
"loss": 0.0033,
|
|
"num_tokens": 18970235.0,
|
|
"reward": 1.31640625,
|
|
"reward_std": 0.152816042304039,
|
|
"rewards/accuracy_reward_step": 0.81640625,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 75
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 11.44140625,
|
|
"calib/ece": 0.06450988142292488,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.011857707509881422,
|
|
"calib/gap": -0.012663003663003769,
|
|
"calib/mean_conf": 0.7305889328063241,
|
|
"calib/mu_c": 0.7263846153846153,
|
|
"calib/mu_w": 0.7390476190476191,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06355731225296442,
|
|
"calib/std_conf": 0.05823786745392095,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6648797327394208,
|
|
"calib/step_q_c_n": 1796.0,
|
|
"calib/step_q_gap": -0.020451246960490943,
|
|
"calib/step_q_w": 0.6853309796999117,
|
|
"calib/step_q_w_n": 1133.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2525.0,
|
|
"completions/max_terminated_length": 2525.0,
|
|
"completions/mean_length": 653.90625,
|
|
"completions/mean_terminated_length": 659.0551147460938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 217.0,
|
|
"epoch": 0.08106666666666666,
|
|
"grad_norm": 0.2749916911125183,
|
|
"learning_rate": 3.4722222222222224e-06,
|
|
"loss": -0.0397,
|
|
"num_tokens": 19240691.0,
|
|
"reward": 1.154296875,
|
|
"reward_std": 0.14932094514369965,
|
|
"rewards/accuracy_reward_step": 0.66015625,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 76
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 11.51953125,
|
|
"calib/ece": 0.12937999999999994,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.028,
|
|
"calib/gap": -0.01767592592592604,
|
|
"calib/mean_conf": 0.7310679999999999,
|
|
"calib/mu_c": 0.72725,
|
|
"calib/mu_w": 0.744925925925926,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.03822399999999998,
|
|
"calib/std_conf": 0.07652042456756235,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.676183522446966,
|
|
"calib/step_q_c_n": 2027.0,
|
|
"calib/step_q_gap": -0.03730508926669995,
|
|
"calib/step_q_w": 0.713488611713666,
|
|
"calib/step_q_w_n": 922.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2514.0,
|
|
"completions/max_terminated_length": 2514.0,
|
|
"completions/mean_length": 622.73828125,
|
|
"completions/mean_terminated_length": 632.623046875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 185.0,
|
|
"epoch": 0.08213333333333334,
|
|
"grad_norm": 0.3323403000831604,
|
|
"learning_rate": 3.444444444444445e-06,
|
|
"loss": -0.0161,
|
|
"num_tokens": 19504776.0,
|
|
"reward": 1.25390625,
|
|
"reward_std": 0.2412024885416031,
|
|
"rewards/accuracy_reward_step": 0.765625,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 77
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 12.43359375,
|
|
"calib/ece": 0.11027450980392149,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.011764705882352941,
|
|
"calib/gap": -0.004929647906108503,
|
|
"calib/mean_conf": 0.7494901960784314,
|
|
"calib/mu_c": 0.7477116564417177,
|
|
"calib/mu_w": 0.7526413043478262,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.11027450980392149,
|
|
"calib/std_conf": 0.059183255630460015,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6824272493573265,
|
|
"calib/step_q_c_n": 1945.0,
|
|
"calib/step_q_gap": -0.016760957427810808,
|
|
"calib/step_q_w": 0.6991882067851373,
|
|
"calib/step_q_w_n": 1238.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2519.0,
|
|
"completions/max_terminated_length": 2519.0,
|
|
"completions/mean_length": 737.19140625,
|
|
"completions/mean_terminated_length": 740.0823974609375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 246.0,
|
|
"epoch": 0.0832,
|
|
"grad_norm": 0.3206815719604492,
|
|
"learning_rate": 3.416666666666667e-06,
|
|
"loss": 0.0029,
|
|
"num_tokens": 19801521.0,
|
|
"reward": 1.134765625,
|
|
"reward_std": 0.1259935051202774,
|
|
"rewards/accuracy_reward_step": 0.63671875,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 78
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 11.1171875,
|
|
"calib/ece": 0.054714062499999966,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0234375,
|
|
"calib/gap": 0.006358829253686582,
|
|
"calib/mean_conf": 0.7532140625,
|
|
"calib/mu_c": 0.7550273224043716,
|
|
"calib/mu_w": 0.7486684931506851,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0465421875,
|
|
"calib/std_conf": 0.06678973856249247,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.7061187961985216,
|
|
"calib/step_q_c_n": 1894.0,
|
|
"calib/step_q_gap": 0.014124258383395483,
|
|
"calib/step_q_w": 0.6919945378151261,
|
|
"calib/step_q_w_n": 952.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1911.0,
|
|
"completions/max_terminated_length": 1911.0,
|
|
"completions/mean_length": 710.62890625,
|
|
"completions/mean_terminated_length": 716.2244262695312,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 199.0,
|
|
"epoch": 0.08426666666666667,
|
|
"grad_norm": 0.2801198661327362,
|
|
"learning_rate": 3.3888888888888893e-06,
|
|
"loss": -0.0243,
|
|
"num_tokens": 20089818.0,
|
|
"reward": 1.21484375,
|
|
"reward_std": 0.1825442910194397,
|
|
"rewards/accuracy_reward_step": 0.71484375,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 79
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 10.17578125,
|
|
"calib/ece": 0.06495703125000002,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.00034867860187548505,
|
|
"calib/mean_conf": 0.7440351562499999,
|
|
"calib/mu_c": 0.7439411764705883,
|
|
"calib/mu_w": 0.7442898550724638,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.03926171874999998,
|
|
"calib/std_conf": 0.057322643717714956,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6887900410677618,
|
|
"calib/step_q_c_n": 1948.0,
|
|
"calib/step_q_gap": -0.003767036557808967,
|
|
"calib/step_q_w": 0.6925570776255707,
|
|
"calib/step_q_w_n": 657.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1356.0,
|
|
"completions/max_terminated_length": 1356.0,
|
|
"completions/mean_length": 559.10546875,
|
|
"completions/mean_terminated_length": 563.5078735351562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 231.0,
|
|
"epoch": 0.08533333333333333,
|
|
"grad_norm": 0.3234640657901764,
|
|
"learning_rate": 3.3611111111111117e-06,
|
|
"loss": -0.0066,
|
|
"num_tokens": 20335109.0,
|
|
"reward": 1.23046875,
|
|
"reward_std": 0.16728198528289795,
|
|
"rewards/accuracy_reward_step": 0.73046875,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 80
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 10.671875,
|
|
"calib/ece": 0.06084143426294811,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.01195219123505976,
|
|
"calib/gap": -0.01690155610155608,
|
|
"calib/mean_conf": 0.7262820717131473,
|
|
"calib/mu_c": 0.7218378378378378,
|
|
"calib/mu_w": 0.7387393939393939,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.025035856573705183,
|
|
"calib/std_conf": 0.055329716852905955,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.6718737238044062,
|
|
"calib/step_q_c_n": 1861.0,
|
|
"calib/step_q_gap": -0.018236035093412206,
|
|
"calib/step_q_w": 0.6901097588978184,
|
|
"calib/step_q_w_n": 871.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2525.0,
|
|
"completions/max_terminated_length": 2525.0,
|
|
"completions/mean_length": 644.9921875,
|
|
"completions/mean_terminated_length": 652.6403198242188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 239.0,
|
|
"epoch": 0.0864,
|
|
"grad_norm": 1.7586580514907837,
|
|
"learning_rate": 3.3333333333333333e-06,
|
|
"loss": -0.0175,
|
|
"num_tokens": 20606475.0,
|
|
"reward": 1.2109375,
|
|
"reward_std": 0.17270183563232422,
|
|
"rewards/accuracy_reward_step": 0.72265625,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 81
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 10.63671875,
|
|
"calib/ece": 0.07736259842519684,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.027559055118110236,
|
|
"calib/gap": -0.018181593462717083,
|
|
"calib/mean_conf": 0.7222688976377952,
|
|
"calib/mu_c": 0.7158981818181818,
|
|
"calib/mu_w": 0.7340797752808988,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.07501259842519682,
|
|
"calib/std_conf": 0.055923123705016065,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6629955128205127,
|
|
"calib/step_q_c_n": 1560.0,
|
|
"calib/step_q_gap": -0.021624865511387537,
|
|
"calib/step_q_w": 0.6846203783319003,
|
|
"calib/step_q_w_n": 1163.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2501.0,
|
|
"completions/max_terminated_length": 2501.0,
|
|
"completions/mean_length": 634.265625,
|
|
"completions/mean_terminated_length": 636.7529907226562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 234.0,
|
|
"epoch": 0.08746666666666666,
|
|
"grad_norm": 0.3052951991558075,
|
|
"learning_rate": 3.3055555555555558e-06,
|
|
"loss": 0.0178,
|
|
"num_tokens": 20874399.0,
|
|
"reward": 1.140625,
|
|
"reward_std": 0.19871768355369568,
|
|
"rewards/accuracy_reward_step": 0.64453125,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 82
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 12.33203125,
|
|
"calib/ece": 0.13502213438735175,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.019762845849802372,
|
|
"calib/gap": -0.02882232035327148,
|
|
"calib/mean_conf": 0.7339501976284584,
|
|
"calib/mu_c": 0.7232415094339623,
|
|
"calib/mu_w": 0.7520638297872337,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.1202569169960474,
|
|
"calib/std_conf": 0.06278977197273311,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6634853221957041,
|
|
"calib/step_q_c_n": 1676.0,
|
|
"calib/step_q_gap": -0.028898202449805677,
|
|
"calib/step_q_w": 0.6923835246455098,
|
|
"calib/step_q_w_n": 1481.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2434.0,
|
|
"completions/max_terminated_length": 2434.0,
|
|
"completions/mean_length": 747.3125,
|
|
"completions/mean_terminated_length": 753.1968383789062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 250.0,
|
|
"epoch": 0.08853333333333334,
|
|
"grad_norm": 0.2530120611190796,
|
|
"learning_rate": 3.277777777777778e-06,
|
|
"loss": -0.002,
|
|
"num_tokens": 21172975.0,
|
|
"reward": 1.115234375,
|
|
"reward_std": 0.1734684705734253,
|
|
"rewards/accuracy_reward_step": 0.62109375,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 83
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 10.45703125,
|
|
"calib/ece": 0.09926023437500002,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.00390625,
|
|
"calib/gap": -0.0005148772926892864,
|
|
"calib/mean_conf": 0.716447734375,
|
|
"calib/mu_c": 0.7162506329113923,
|
|
"calib/mu_w": 0.7167655102040816,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.09926023437500002,
|
|
"calib/std_conf": 0.05594260874317485,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6651108609271524,
|
|
"calib/step_q_c_n": 1510.0,
|
|
"calib/step_q_gap": -0.0043531236486830105,
|
|
"calib/step_q_w": 0.6694639845758354,
|
|
"calib/step_q_w_n": 1167.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1671.0,
|
|
"completions/max_terminated_length": 1671.0,
|
|
"completions/mean_length": 609.6953125,
|
|
"completions/mean_terminated_length": 614.4960327148438,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 161.0,
|
|
"epoch": 0.0896,
|
|
"grad_norm": 0.5088642835617065,
|
|
"learning_rate": 3.2500000000000002e-06,
|
|
"loss": 0.0052,
|
|
"num_tokens": 21434977.0,
|
|
"reward": 1.1171875,
|
|
"reward_std": 0.19530275464057922,
|
|
"rewards/accuracy_reward_step": 0.6171875,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 84
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 11.0390625,
|
|
"calib/ece": 0.08913231075697205,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.00398406374501992,
|
|
"calib/gap": 0.011304903846153791,
|
|
"calib/mean_conf": 0.7246701593625497,
|
|
"calib/mu_c": 0.72876875,
|
|
"calib/mu_w": 0.7174638461538462,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.08817613545816728,
|
|
"calib/std_conf": 0.05422468456992324,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.667843023255814,
|
|
"calib/step_q_c_n": 1720.0,
|
|
"calib/step_q_gap": -0.002186316708019609,
|
|
"calib/step_q_w": 0.6700293399638336,
|
|
"calib/step_q_w_n": 1106.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2519.0,
|
|
"completions/max_terminated_length": 2519.0,
|
|
"completions/mean_length": 682.765625,
|
|
"completions/mean_terminated_length": 690.8616943359375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 209.0,
|
|
"epoch": 0.09066666666666667,
|
|
"grad_norm": 0.3552238643169403,
|
|
"learning_rate": 3.2222222222222227e-06,
|
|
"loss": -0.0124,
|
|
"num_tokens": 21717589.0,
|
|
"reward": 1.11328125,
|
|
"reward_std": 0.25730133056640625,
|
|
"rewards/accuracy_reward_step": 0.625,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 85
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 11.50390625,
|
|
"calib/ece": 0.1369408730158731,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.01984126984126984,
|
|
"calib/gap": -0.014084684210526555,
|
|
"calib/mean_conf": 0.7401154761904762,
|
|
"calib/mu_c": 0.7345263157894737,
|
|
"calib/mu_w": 0.7486110000000002,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
|
"calib/pce": 0.1369408730158731,
|
|
"calib/std_conf": 0.06631348260175485,
|
|
"calib/step_conf_rate": 0.98828125,
|
|
"calib/step_q_c": 0.674241186345831,
|
|
"calib/step_q_c_n": 1787.0,
|
|
"calib/step_q_gap": -0.004901300700801081,
|
|
"calib/step_q_w": 0.679142487046632,
|
|
"calib/step_q_w_n": 1158.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2530.0,
|
|
"completions/max_terminated_length": 2530.0,
|
|
"completions/mean_length": 766.328125,
|
|
"completions/mean_terminated_length": 769.3333740234375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 179.0,
|
|
"epoch": 0.09173333333333333,
|
|
"grad_norm": 0.34955188632011414,
|
|
"learning_rate": 3.1944444444444443e-06,
|
|
"loss": 0.0287,
|
|
"num_tokens": 22019281.0,
|
|
"reward": 1.080078125,
|
|
"reward_std": 0.19237485527992249,
|
|
"rewards/accuracy_reward_step": 0.59375,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 86
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 9.87109375,
|
|
"calib/ece": 0.10698042968750003,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.015625,
|
|
"calib/gap": 0.006787857484457094,
|
|
"calib/mean_conf": 0.7179648046875,
|
|
"calib/mu_c": 0.7193170731707316,
|
|
"calib/mu_w": 0.7125292156862745,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.0120819921875,
|
|
"calib/std_conf": 0.05209170035673249,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.6690812720848056,
|
|
"calib/step_q_c_n": 1981.0,
|
|
"calib/step_q_gap": 0.00742588197491556,
|
|
"calib/step_q_w": 0.66165539010989,
|
|
"calib/step_q_w_n": 546.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0078125,
|
|
"completions/max_length": 1944.0,
|
|
"completions/max_terminated_length": 1944.0,
|
|
"completions/mean_length": 558.4609375,
|
|
"completions/mean_terminated_length": 562.8582763671875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 194.0,
|
|
"epoch": 0.0928,
|
|
"grad_norm": 0.40007147192955017,
|
|
"learning_rate": 3.1666666666666667e-06,
|
|
"loss": -0.0021,
|
|
"num_tokens": 22267743.0,
|
|
"reward": 1.296875,
|
|
"reward_std": 0.2075463831424713,
|
|
"rewards/accuracy_reward_step": 0.80078125,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 87
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 11.15625,
|
|
"calib/ece": 0.05912992125984258,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.05511811023622047,
|
|
"calib/gap": -0.009182278481012718,
|
|
"calib/mean_conf": 0.741255905511811,
|
|
"calib/mu_c": 0.7384,
|
|
"calib/mu_w": 0.7475822784810127,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.055704724409448875,
|
|
"calib/std_conf": 0.06912001334383905,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.673917751171265,
|
|
"calib/step_q_c_n": 1921.0,
|
|
"calib/step_q_gap": -0.015334976101462283,
|
|
"calib/step_q_w": 0.6892527272727272,
|
|
"calib/step_q_w_n": 935.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0078125,
|
|
"completions/max_length": 2875.0,
|
|
"completions/max_terminated_length": 2875.0,
|
|
"completions/mean_length": 725.66015625,
|
|
"completions/mean_terminated_length": 731.3740234375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 266.0,
|
|
"epoch": 0.09386666666666667,
|
|
"grad_norm": 0.5738656520843506,
|
|
"learning_rate": 3.138888888888889e-06,
|
|
"loss": 0.0035,
|
|
"num_tokens": 22563360.0,
|
|
"reward": 1.177734375,
|
|
"reward_std": 0.21606913208961487,
|
|
"rewards/accuracy_reward_step": 0.68359375,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 88
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 11.77734375,
|
|
"calib/ece": 0.14020318725099604,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0199203187250996,
|
|
"calib/gap": -0.017074783521385584,
|
|
"calib/mean_conf": 0.7298446215139442,
|
|
"calib/mu_c": 0.7228378378378378,
|
|
"calib/mu_w": 0.7399126213592234,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.14020318725099604,
|
|
"calib/std_conf": 0.06890072945100548,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.660207026348808,
|
|
"calib/step_q_c_n": 1594.0,
|
|
"calib/step_q_gap": -0.030072354368996357,
|
|
"calib/step_q_w": 0.6902793807178044,
|
|
"calib/step_q_w_n": 1421.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2518.0,
|
|
"completions/max_terminated_length": 2518.0,
|
|
"completions/mean_length": 767.21875,
|
|
"completions/mean_terminated_length": 776.3162231445312,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 196.0,
|
|
"epoch": 0.09493333333333333,
|
|
"grad_norm": 0.2824539840221405,
|
|
"learning_rate": 3.1111111111111116e-06,
|
|
"loss": -0.0148,
|
|
"num_tokens": 22868656.0,
|
|
"reward": 1.06640625,
|
|
"reward_std": 0.2012740671634674,
|
|
"rewards/accuracy_reward_step": 0.578125,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 89
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 11.89453125,
|
|
"calib/ece": 0.06254980079681267,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.03187250996015936,
|
|
"calib/gap": -0.002347365278399627,
|
|
"calib/mean_conf": 0.7262948207171314,
|
|
"calib/mu_c": 0.7255747126436782,
|
|
"calib/mu_w": 0.7279220779220779,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04780876494023897,
|
|
"calib/std_conf": 0.07368648411549109,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.660236537493709,
|
|
"calib/step_q_c_n": 1987.0,
|
|
"calib/step_q_gap": -0.015680286702888346,
|
|
"calib/step_q_w": 0.6759168241965974,
|
|
"calib/step_q_w_n": 1058.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1878.0,
|
|
"completions/max_terminated_length": 1878.0,
|
|
"completions/mean_length": 694.2421875,
|
|
"completions/mean_terminated_length": 708.0717163085938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 194.0,
|
|
"epoch": 0.096,
|
|
"grad_norm": 0.36388516426086426,
|
|
"learning_rate": 3.0833333333333336e-06,
|
|
"loss": -0.0381,
|
|
"num_tokens": 23149702.0,
|
|
"reward": 1.169921875,
|
|
"reward_std": 0.21346309781074524,
|
|
"rewards/accuracy_reward_step": 0.6796875,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 90
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 11.84375,
|
|
"calib/ece": 0.10066929133858277,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.010466200466200615,
|
|
"calib/mean_conf": 0.6818503937007874,
|
|
"calib/mu_c": 0.6786363636363636,
|
|
"calib/mu_w": 0.6891025641025642,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04480314960629926,
|
|
"calib/std_conf": 0.06787640211746589,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6107117969138874,
|
|
"calib/step_q_c_n": 2009.0,
|
|
"calib/step_q_gap": -0.022455358511332513,
|
|
"calib/step_q_w": 0.63316715542522,
|
|
"calib/step_q_w_n": 1023.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1942.0,
|
|
"completions/max_terminated_length": 1942.0,
|
|
"completions/mean_length": 734.11328125,
|
|
"completions/mean_terminated_length": 742.8182373046875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 334.0,
|
|
"epoch": 0.09706666666666666,
|
|
"grad_norm": 0.3484867215156555,
|
|
"learning_rate": 3.055555555555556e-06,
|
|
"loss": -0.0073,
|
|
"num_tokens": 23445347.0,
|
|
"reward": 1.18359375,
|
|
"reward_std": 0.2495465874671936,
|
|
"rewards/accuracy_reward_step": 0.6875,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 91
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 10.3359375,
|
|
"calib/ece": 0.14858203125000002,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.028630628535999292,
|
|
"calib/mean_conf": 0.63883984375,
|
|
"calib/mu_c": 0.6311229946524065,
|
|
"calib/mu_w": 0.6597536231884058,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.02847656250000002,
|
|
"calib/std_conf": 0.05463817180987653,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.5719755147468003,
|
|
"calib/step_q_c_n": 1797.0,
|
|
"calib/step_q_gap": -0.019971481719630813,
|
|
"calib/step_q_w": 0.5919469964664311,
|
|
"calib/step_q_w_n": 849.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0078125,
|
|
"completions/max_length": 1580.0,
|
|
"completions/max_terminated_length": 1580.0,
|
|
"completions/mean_length": 622.1640625,
|
|
"completions/mean_terminated_length": 627.06298828125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 231.0,
|
|
"epoch": 0.09813333333333334,
|
|
"grad_norm": 0.3795936703681946,
|
|
"learning_rate": 3.0277777777777776e-06,
|
|
"loss": -0.0023,
|
|
"num_tokens": 23711341.0,
|
|
"reward": 1.228515625,
|
|
"reward_std": 0.17531254887580872,
|
|
"rewards/accuracy_reward_step": 0.73046875,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 92
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 10.91015625,
|
|
"calib/ece": 0.1233858267716535,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.003937007874015748,
|
|
"calib/gap": -0.0200209542230817,
|
|
"calib/mean_conf": 0.6407874015748031,
|
|
"calib/mu_c": 0.6355851063829788,
|
|
"calib/mu_w": 0.6556060606060605,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.012007874015748009,
|
|
"calib/std_conf": 0.06036120017816887,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5698559670781893,
|
|
"calib/step_q_c_n": 1944.0,
|
|
"calib/step_q_gap": -0.017258284983059258,
|
|
"calib/step_q_w": 0.5871142520612486,
|
|
"calib/step_q_w_n": 849.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.01171875,
|
|
"completions/max_length": 2512.0,
|
|
"completions/max_terminated_length": 2512.0,
|
|
"completions/mean_length": 694.6953125,
|
|
"completions/mean_terminated_length": 702.932861328125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 197.0,
|
|
"epoch": 0.0992,
|
|
"grad_norm": 0.3373625576496124,
|
|
"learning_rate": 3e-06,
|
|
"loss": -0.0173,
|
|
"num_tokens": 23994959.0,
|
|
"reward": 1.23046875,
|
|
"reward_std": 0.19140824675559998,
|
|
"rewards/accuracy_reward_step": 0.734375,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 93
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 9.92578125,
|
|
"calib/ece": 0.09505882352941177,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.00392156862745098,
|
|
"calib/gap": -0.012319182389937122,
|
|
"calib/mean_conf": 0.620235294117647,
|
|
"calib/mu_c": 0.6155974842767296,
|
|
"calib/mu_w": 0.6279166666666667,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04588235294117648,
|
|
"calib/std_conf": 0.05204017612187577,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.564599589322382,
|
|
"calib/step_q_c_n": 1461.0,
|
|
"calib/step_q_gap": -0.01198374401095137,
|
|
"calib/step_q_w": 0.5765833333333333,
|
|
"calib/step_q_w_n": 1080.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2383.0,
|
|
"completions/max_terminated_length": 2383.0,
|
|
"completions/mean_length": 617.9765625,
|
|
"completions/mean_terminated_length": 622.842529296875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 186.0,
|
|
"epoch": 0.10026666666666667,
|
|
"grad_norm": 0.2551957368850708,
|
|
"learning_rate": 2.9722222222222225e-06,
|
|
"loss": -0.0174,
|
|
"num_tokens": 24261841.0,
|
|
"reward": 1.119140625,
|
|
"reward_std": 0.15191996097564697,
|
|
"rewards/accuracy_reward_step": 0.62109375,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 94
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 11.12890625,
|
|
"calib/ece": 0.12202380952380956,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.011131101250504094,
|
|
"calib/mean_conf": 0.6475000000000001,
|
|
"calib/mu_c": 0.6445405405405406,
|
|
"calib/mu_w": 0.6556716417910446,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.017698412698412692,
|
|
"calib/std_conf": 0.05655889611863664,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5712020725388601,
|
|
"calib/step_q_c_n": 1930.0,
|
|
"calib/step_q_gap": -0.01644754661239123,
|
|
"calib/step_q_w": 0.5876496191512514,
|
|
"calib/step_q_w_n": 919.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2507.0,
|
|
"completions/max_terminated_length": 2507.0,
|
|
"completions/mean_length": 749.59375,
|
|
"completions/mean_terminated_length": 755.4960327148438,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 210.0,
|
|
"epoch": 0.10133333333333333,
|
|
"grad_norm": 0.43740609288215637,
|
|
"learning_rate": 2.944444444444445e-06,
|
|
"loss": -0.024,
|
|
"num_tokens": 24559865.0,
|
|
"reward": 1.21484375,
|
|
"reward_std": 0.23373465240001678,
|
|
"rewards/accuracy_reward_step": 0.72265625,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 95
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 10.90625,
|
|
"calib/ece": 0.17896825396825397,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.032203846153846194,
|
|
"calib/mean_conf": 0.6380952380952382,
|
|
"calib/mu_c": 0.63145,
|
|
"calib/mu_w": 0.6636538461538461,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.011706349206349186,
|
|
"calib/std_conf": 0.0528732279257898,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5689752306945118,
|
|
"calib/step_q_c_n": 2059.0,
|
|
"calib/step_q_gap": -0.01829625634505161,
|
|
"calib/step_q_w": 0.5872714870395634,
|
|
"calib/step_q_w_n": 733.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2515.0,
|
|
"completions/max_terminated_length": 2515.0,
|
|
"completions/mean_length": 674.65234375,
|
|
"completions/mean_terminated_length": 679.9645385742188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 170.0,
|
|
"epoch": 0.1024,
|
|
"grad_norm": 0.4047119617462158,
|
|
"learning_rate": 2.916666666666667e-06,
|
|
"loss": 0.0177,
|
|
"num_tokens": 24838392.0,
|
|
"reward": 1.2734375,
|
|
"reward_std": 0.18096905946731567,
|
|
"rewards/accuracy_reward_step": 0.78125,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 96
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 10.21484375,
|
|
"calib/ece": 0.04666666666666672,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.003968253968253968,
|
|
"calib/gap": -0.0028923719958201533,
|
|
"calib/mean_conf": 0.633968253968254,
|
|
"calib/mu_c": 0.632969696969697,
|
|
"calib/mu_w": 0.6358620689655171,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.012936507936507934,
|
|
"calib/std_conf": 0.05331218874122451,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.565439459127228,
|
|
"calib/step_q_c_n": 1627.0,
|
|
"calib/step_q_gap": -0.020957302006375245,
|
|
"calib/step_q_w": 0.5863967611336033,
|
|
"calib/step_q_w_n": 988.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2204.0,
|
|
"completions/max_terminated_length": 2204.0,
|
|
"completions/mean_length": 662.5390625,
|
|
"completions/mean_terminated_length": 673.0556030273438,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 211.0,
|
|
"epoch": 0.10346666666666667,
|
|
"grad_norm": 0.3296211063861847,
|
|
"learning_rate": 2.888888888888889e-06,
|
|
"loss": -0.0314,
|
|
"num_tokens": 25113074.0,
|
|
"reward": 1.13671875,
|
|
"reward_std": 0.2406836599111557,
|
|
"rewards/accuracy_reward_step": 0.64453125,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 97
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 10.93359375,
|
|
"calib/ece": 0.05466135458167331,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.00398406374501992,
|
|
"calib/gap": 0.005932300631095844,
|
|
"calib/mean_conf": 0.6456573705179283,
|
|
"calib/mu_c": 0.6476190476190476,
|
|
"calib/mu_w": 0.6416867469879518,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0154980079681275,
|
|
"calib/std_conf": 0.05651804118633278,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5745659526493799,
|
|
"calib/step_q_c_n": 1774.0,
|
|
"calib/step_q_gap": -0.004487705887205373,
|
|
"calib/step_q_w": 0.5790536585365853,
|
|
"calib/step_q_w_n": 1025.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2535.0,
|
|
"completions/max_terminated_length": 2535.0,
|
|
"completions/mean_length": 750.25390625,
|
|
"completions/mean_terminated_length": 759.1502075195312,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 171.0,
|
|
"epoch": 0.10453333333333334,
|
|
"grad_norm": 0.5152229070663452,
|
|
"learning_rate": 2.861111111111111e-06,
|
|
"loss": -0.0273,
|
|
"num_tokens": 25411323.0,
|
|
"reward": 1.146484375,
|
|
"reward_std": 0.2555645704269409,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 98
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/avg_num_step_conf": 13.6875,
|
|
"calib/ece": 0.11594337349397595,
|
|
"calib/final_conf_rate": 0.97265625,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.028112449799196786,
|
|
"calib/gap": -0.029866776359088965,
|
|
"calib/mean_conf": 0.6661441767068274,
|
|
"calib/mu_c": 0.6533098591549297,
|
|
"calib/mu_w": 0.6831766355140186,
|
|
"calib/nonempty_final_conf_rate": 0.97265625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.10590321285140564,
|
|
"calib/std_conf": 0.0799204432245325,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5810451453308597,
|
|
"calib/step_q_c_n": 1617.0,
|
|
"calib/step_q_gap": -0.05463529982017368,
|
|
"calib/step_q_w": 0.6356804451510334,
|
|
"calib/step_q_w_n": 1887.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2943.0,
|
|
"completions/max_terminated_length": 2943.0,
|
|
"completions/mean_length": 885.05078125,
|
|
"completions/mean_terminated_length": 895.5454711914062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 220.0,
|
|
"epoch": 0.1056,
|
|
"grad_norm": 0.9046245217323303,
|
|
"learning_rate": 2.8333333333333335e-06,
|
|
"loss": -0.0087,
|
|
"num_tokens": 25743696.0,
|
|
"reward": 1.041015625,
|
|
"reward_std": 0.2987309396266937,
|
|
"rewards/accuracy_reward_step": 0.5546875,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 99
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9375,
|
|
"calib/avg_num_step_conf": 12.90625,
|
|
"calib/ece": 0.10766666666666669,
|
|
"calib/final_conf_rate": 0.9375,
|
|
"calib/format_rate": 0.9375,
|
|
"calib/frac_conf_gt_0.9": 0.008333333333333333,
|
|
"calib/gap": -0.031806100875868326,
|
|
"calib/mean_conf": 0.6669166666666667,
|
|
"calib/mu_c": 0.6555194805194805,
|
|
"calib/mu_w": 0.6873255813953488,
|
|
"calib/nonempty_final_conf_rate": 0.9375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06645833333333337,
|
|
"calib/std_conf": 0.06673699415533253,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5753517587939698,
|
|
"calib/step_q_c_n": 1592.0,
|
|
"calib/step_q_gap": -0.0632814187761237,
|
|
"calib/step_q_w": 0.6386331775700935,
|
|
"calib/step_q_w_n": 1712.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.05859375,
|
|
"completions/max_length": 2507.0,
|
|
"completions/max_terminated_length": 2507.0,
|
|
"completions/mean_length": 829.94140625,
|
|
"completions/mean_terminated_length": 881.5975341796875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 311.0,
|
|
"epoch": 0.10666666666666667,
|
|
"grad_norm": 0.6749600768089294,
|
|
"learning_rate": 2.805555555555556e-06,
|
|
"loss": -0.104,
|
|
"num_tokens": 26063569.0,
|
|
"reward": 1.0703125,
|
|
"reward_std": 0.24402156472206116,
|
|
"rewards/accuracy_reward_step": 0.6015625,
|
|
"rewards/format_reward_step": 0.9375,
|
|
"step": 100
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.94921875,
|
|
"calib/avg_num_step_conf": 11.8828125,
|
|
"calib/ece": 0.08617283950617284,
|
|
"calib/final_conf_rate": 0.94921875,
|
|
"calib/format_rate": 0.94921875,
|
|
"calib/frac_conf_gt_0.9": 0.024691358024691357,
|
|
"calib/gap": 0.008903061224489806,
|
|
"calib/mean_conf": 0.6758024691358024,
|
|
"calib/mu_c": 0.6793197278911565,
|
|
"calib/mu_w": 0.6704166666666667,
|
|
"calib/nonempty_final_conf_rate": 0.94921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.07851851851851853,
|
|
"calib/std_conf": 0.07632286085147473,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.587848347375243,
|
|
"calib/step_q_c_n": 1543.0,
|
|
"calib/step_q_gap": -0.039883473838899675,
|
|
"calib/step_q_w": 0.6277318212141427,
|
|
"calib/step_q_w_n": 1499.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3069.0,
|
|
"completions/max_terminated_length": 3069.0,
|
|
"completions/mean_length": 927.57421875,
|
|
"completions/mean_terminated_length": 949.8360595703125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 309.0,
|
|
"epoch": 0.10773333333333333,
|
|
"grad_norm": 0.6551363468170166,
|
|
"learning_rate": 2.7777777777777783e-06,
|
|
"loss": -0.0233,
|
|
"num_tokens": 26408020.0,
|
|
"reward": 1.048828125,
|
|
"reward_std": 0.27887511253356934,
|
|
"rewards/accuracy_reward_step": 0.57421875,
|
|
"rewards/format_reward_step": 0.94921875,
|
|
"step": 101
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96484375,
|
|
"calib/avg_num_step_conf": 11.37109375,
|
|
"calib/ece": 0.0396761133603239,
|
|
"calib/final_conf_rate": 0.96484375,
|
|
"calib/format_rate": 0.96484375,
|
|
"calib/frac_conf_gt_0.9": 0.004048582995951417,
|
|
"calib/gap": -0.003409460458240976,
|
|
"calib/mean_conf": 0.6502834008097166,
|
|
"calib/mu_c": 0.6491515151515151,
|
|
"calib/mu_w": 0.6525609756097561,
|
|
"calib/nonempty_final_conf_rate": 0.96484375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.010971659919028343,
|
|
"calib/std_conf": 0.06104621472093361,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5718552875695733,
|
|
"calib/step_q_c_n": 1617.0,
|
|
"calib/step_q_gap": -0.05240553159580541,
|
|
"calib/step_q_w": 0.6242608191653787,
|
|
"calib/step_q_w_n": 1294.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2830.0,
|
|
"completions/max_terminated_length": 2830.0,
|
|
"completions/mean_length": 728.40625,
|
|
"completions/mean_terminated_length": 745.8880615234375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 240.0,
|
|
"epoch": 0.1088,
|
|
"grad_norm": 0.5937004089355469,
|
|
"learning_rate": 2.7500000000000004e-06,
|
|
"loss": -0.0515,
|
|
"num_tokens": 26701188.0,
|
|
"reward": 1.126953125,
|
|
"reward_std": 0.18155640363693237,
|
|
"rewards/accuracy_reward_step": 0.64453125,
|
|
"rewards/format_reward_step": 0.96484375,
|
|
"step": 102
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9296875,
|
|
"calib/avg_num_step_conf": 11.8828125,
|
|
"calib/ece": 0.15508474576271183,
|
|
"calib/final_conf_rate": 0.921875,
|
|
"calib/format_rate": 0.921875,
|
|
"calib/frac_conf_gt_0.9": 0.01694915254237288,
|
|
"calib/gap": -0.045967343687773665,
|
|
"calib/mean_conf": 0.6633898305084746,
|
|
"calib/mu_c": 0.6476129032258064,
|
|
"calib/mu_w": 0.6935802469135801,
|
|
"calib/nonempty_final_conf_rate": 0.921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.08084745762711865,
|
|
"calib/std_conf": 0.0752591042860422,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5754819277108434,
|
|
"calib/step_q_c_n": 1660.0,
|
|
"calib/step_q_gap": -0.041833557093787554,
|
|
"calib/step_q_w": 0.6173154848046309,
|
|
"calib/step_q_w_n": 1382.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2903.0,
|
|
"completions/max_terminated_length": 2903.0,
|
|
"completions/mean_length": 894.046875,
|
|
"completions/mean_terminated_length": 938.016357421875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 336.0,
|
|
"epoch": 0.10986666666666667,
|
|
"grad_norm": 0.4124567210674286,
|
|
"learning_rate": 2.7222222222222224e-06,
|
|
"loss": -0.0684,
|
|
"num_tokens": 27034616.0,
|
|
"reward": 1.07421875,
|
|
"reward_std": 0.23740246891975403,
|
|
"rewards/accuracy_reward_step": 0.61328125,
|
|
"rewards/format_reward_step": 0.921875,
|
|
"step": 103
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9453125,
|
|
"calib/avg_num_step_conf": 11.72265625,
|
|
"calib/ece": 0.08705394190871377,
|
|
"calib/final_conf_rate": 0.94140625,
|
|
"calib/format_rate": 0.94140625,
|
|
"calib/frac_conf_gt_0.9": 0.012448132780082987,
|
|
"calib/gap": -0.01802863591479753,
|
|
"calib/mean_conf": 0.6638174273858921,
|
|
"calib/mu_c": 0.6561870503597123,
|
|
"calib/mu_w": 0.6742156862745098,
|
|
"calib/nonempty_final_conf_rate": 0.94140625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.08705394190871377,
|
|
"calib/std_conf": 0.06525048979339483,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5755367231638419,
|
|
"calib/step_q_c_n": 1416.0,
|
|
"calib/step_q_gap": -0.04505633677306675,
|
|
"calib/step_q_w": 0.6205930599369086,
|
|
"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": 0.04296875,
|
|
"completions/max_length": 2775.0,
|
|
"completions/max_terminated_length": 2775.0,
|
|
"completions/mean_length": 834.640625,
|
|
"completions/mean_terminated_length": 872.1142578125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 211.0,
|
|
"epoch": 0.11093333333333333,
|
|
"grad_norm": 0.4785301983356476,
|
|
"learning_rate": 2.6944444444444444e-06,
|
|
"loss": -0.0446,
|
|
"num_tokens": 27354964.0,
|
|
"reward": 1.013671875,
|
|
"reward_std": 0.2309260368347168,
|
|
"rewards/accuracy_reward_step": 0.54296875,
|
|
"rewards/format_reward_step": 0.94140625,
|
|
"step": 104
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9453125,
|
|
"calib/avg_num_step_conf": 9.25,
|
|
"calib/ece": 0.1797499999999999,
|
|
"calib/final_conf_rate": 0.9375,
|
|
"calib/format_rate": 0.9375,
|
|
"calib/frac_conf_gt_0.9": 0.004166666666666667,
|
|
"calib/gap": -0.038034682080924975,
|
|
"calib/mean_conf": 0.6525833333333334,
|
|
"calib/mu_c": 0.6419653179190752,
|
|
"calib/mu_w": 0.6800000000000002,
|
|
"calib/nonempty_final_conf_rate": 0.9375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.055749999999999994,
|
|
"calib/std_conf": 0.05867986357251429,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5645699937225361,
|
|
"calib/step_q_c_n": 1593.0,
|
|
"calib/step_q_gap": -0.01497839337423812,
|
|
"calib/step_q_w": 0.5795483870967743,
|
|
"calib/step_q_w_n": 775.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2929.0,
|
|
"completions/max_terminated_length": 2929.0,
|
|
"completions/mean_length": 840.31640625,
|
|
"completions/mean_terminated_length": 870.9352416992188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 302.0,
|
|
"epoch": 0.112,
|
|
"grad_norm": 0.31713542342185974,
|
|
"learning_rate": 2.666666666666667e-06,
|
|
"loss": -0.023,
|
|
"num_tokens": 27675845.0,
|
|
"reward": 1.14453125,
|
|
"reward_std": 0.2606049180030823,
|
|
"rewards/accuracy_reward_step": 0.67578125,
|
|
"rewards/format_reward_step": 0.9375,
|
|
"step": 105
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 10.19140625,
|
|
"calib/ece": 0.07927999999999995,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.027988570669147173,
|
|
"calib/mean_conf": 0.6452,
|
|
"calib/mu_c": 0.6338926174496645,
|
|
"calib/mu_w": 0.6618811881188117,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06423999999999995,
|
|
"calib/std_conf": 0.05042380390252208,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5630168453292496,
|
|
"calib/step_q_c_n": 1306.0,
|
|
"calib/step_q_gap": -0.031081389513421098,
|
|
"calib/step_q_w": 0.5940982348426707,
|
|
"calib/step_q_w_n": 1303.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2632.0,
|
|
"completions/max_terminated_length": 2632.0,
|
|
"completions/mean_length": 782.90234375,
|
|
"completions/mean_terminated_length": 795.3294067382812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 295.0,
|
|
"epoch": 0.11306666666666666,
|
|
"grad_norm": 0.3023886978626251,
|
|
"learning_rate": 2.6388888888888893e-06,
|
|
"loss": -0.0247,
|
|
"num_tokens": 27980852.0,
|
|
"reward": 1.0703125,
|
|
"reward_std": 0.16276384890079498,
|
|
"rewards/accuracy_reward_step": 0.58203125,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 106
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 10.2734375,
|
|
"calib/ece": 0.14872509960159358,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.00398406374501992,
|
|
"calib/gap": 0.005709268590624705,
|
|
"calib/mean_conf": 0.6515936254980079,
|
|
"calib/mu_c": 0.6532768361581921,
|
|
"calib/mu_w": 0.6475675675675674,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.047569721115537825,
|
|
"calib/std_conf": 0.056390872287046694,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5655207706201084,
|
|
"calib/step_q_c_n": 1661.0,
|
|
"calib/step_q_gap": -0.04727592700631056,
|
|
"calib/step_q_w": 0.612796697626419,
|
|
"calib/step_q_w_n": 969.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2508.0,
|
|
"completions/max_terminated_length": 2508.0,
|
|
"completions/mean_length": 800.51953125,
|
|
"completions/mean_terminated_length": 813.2262573242188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 237.0,
|
|
"epoch": 0.11413333333333334,
|
|
"grad_norm": 0.30182257294654846,
|
|
"learning_rate": 2.6111111111111113e-06,
|
|
"loss": 0.0129,
|
|
"num_tokens": 28290401.0,
|
|
"reward": 1.181640625,
|
|
"reward_std": 0.21702617406845093,
|
|
"rewards/accuracy_reward_step": 0.69140625,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 107
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 10.63671875,
|
|
"calib/ece": 0.1460080645161291,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.009232899189826838,
|
|
"calib/mean_conf": 0.6650403225806452,
|
|
"calib/mu_c": 0.66321608040201,
|
|
"calib/mu_w": 0.6724489795918368,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.004314516129032264,
|
|
"calib/std_conf": 0.0597778008199004,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5796858124693177,
|
|
"calib/step_q_c_n": 2037.0,
|
|
"calib/step_q_gap": -0.003389989279953487,
|
|
"calib/step_q_w": 0.5830758017492712,
|
|
"calib/step_q_w_n": 686.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2514.0,
|
|
"completions/max_terminated_length": 2514.0,
|
|
"completions/mean_length": 862.4609375,
|
|
"completions/mean_terminated_length": 886.706787109375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 194.0,
|
|
"epoch": 0.1152,
|
|
"grad_norm": 0.2606714367866516,
|
|
"learning_rate": 2.5833333333333337e-06,
|
|
"loss": -0.0707,
|
|
"num_tokens": 28614423.0,
|
|
"reward": 1.26171875,
|
|
"reward_std": 0.16206827759742737,
|
|
"rewards/accuracy_reward_step": 0.77734375,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 108
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.91796875,
|
|
"calib/avg_num_step_conf": 11.21875,
|
|
"calib/ece": 0.12272340425531918,
|
|
"calib/final_conf_rate": 0.91796875,
|
|
"calib/format_rate": 0.91796875,
|
|
"calib/frac_conf_gt_0.9": 0.00851063829787234,
|
|
"calib/gap": -0.02565572541966432,
|
|
"calib/mean_conf": 0.6565957446808511,
|
|
"calib/mu_c": 0.6461151079136691,
|
|
"calib/mu_w": 0.6717708333333334,
|
|
"calib/nonempty_final_conf_rate": 0.91796875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.09391489361702125,
|
|
"calib/std_conf": 0.06526139244296268,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5648293515358362,
|
|
"calib/step_q_c_n": 1172.0,
|
|
"calib/step_q_gap": -0.06726806022886955,
|
|
"calib/step_q_w": 0.6320974117647058,
|
|
"calib/step_q_w_n": 1700.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0625,
|
|
"completions/max_length": 3062.0,
|
|
"completions/max_terminated_length": 3062.0,
|
|
"completions/mean_length": 821.3515625,
|
|
"completions/mean_terminated_length": 876.1083984375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 257.0,
|
|
"epoch": 0.11626666666666667,
|
|
"grad_norm": 0.2767605185508728,
|
|
"learning_rate": 2.5555555555555557e-06,
|
|
"loss": -0.0747,
|
|
"num_tokens": 28929289.0,
|
|
"reward": 1.001953125,
|
|
"reward_std": 0.20017710328102112,
|
|
"rewards/accuracy_reward_step": 0.54296875,
|
|
"rewards/format_reward_step": 0.91796875,
|
|
"step": 109
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 8.90234375,
|
|
"calib/ece": 0.0916269841269841,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.003968253968253968,
|
|
"calib/gap": -0.008398117218341938,
|
|
"calib/mean_conf": 0.6412301587301587,
|
|
"calib/mu_c": 0.6387640449438201,
|
|
"calib/mu_w": 0.647162162162162,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.013253968253968247,
|
|
"calib/std_conf": 0.054486005586175154,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.567722705961152,
|
|
"calib/step_q_c_n": 1493.0,
|
|
"calib/step_q_gap": -0.023867624827652034,
|
|
"calib/step_q_w": 0.5915903307888041,
|
|
"calib/step_q_w_n": 786.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2500.0,
|
|
"completions/max_terminated_length": 2500.0,
|
|
"completions/mean_length": 749.65234375,
|
|
"completions/mean_terminated_length": 758.54150390625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 206.0,
|
|
"epoch": 0.11733333333333333,
|
|
"grad_norm": 0.33444392681121826,
|
|
"learning_rate": 2.5277777777777778e-06,
|
|
"loss": -0.0207,
|
|
"num_tokens": 29226120.0,
|
|
"reward": 1.1875,
|
|
"reward_std": 0.21375533938407898,
|
|
"rewards/accuracy_reward_step": 0.6953125,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 110
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96484375,
|
|
"calib/avg_num_step_conf": 9.0,
|
|
"calib/ece": 0.17157894736842103,
|
|
"calib/final_conf_rate": 0.96484375,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.03143781094527354,
|
|
"calib/mean_conf": 0.6361943319838056,
|
|
"calib/mu_c": 0.6276666666666668,
|
|
"calib/mu_w": 0.6591044776119404,
|
|
"calib/nonempty_final_conf_rate": 0.96484375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.03951417004048582,
|
|
"calib/std_conf": 0.05609501309554516,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.5615014367816092,
|
|
"calib/step_q_c_n": 1392.0,
|
|
"calib/step_q_gap": -0.041788036902601355,
|
|
"calib/step_q_w": 0.6032894736842106,
|
|
"calib/step_q_w_n": 912.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2341.0,
|
|
"completions/max_terminated_length": 2341.0,
|
|
"completions/mean_length": 783.45703125,
|
|
"completions/mean_terminated_length": 805.4818725585938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 238.0,
|
|
"epoch": 0.1184,
|
|
"grad_norm": 0.29022684693336487,
|
|
"learning_rate": 2.5e-06,
|
|
"loss": -0.0271,
|
|
"num_tokens": 29534093.0,
|
|
"reward": 1.18359375,
|
|
"reward_std": 0.2522103786468506,
|
|
"rewards/accuracy_reward_step": 0.703125,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 111
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 8.8984375,
|
|
"calib/ece": 0.07762096774193554,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.011590909090909096,
|
|
"calib/mean_conf": 0.6503629032258064,
|
|
"calib/mu_c": 0.64625,
|
|
"calib/mu_w": 0.6578409090909091,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04141129032258063,
|
|
"calib/std_conf": 0.05729336428916774,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5658276125095346,
|
|
"calib/step_q_c_n": 1311.0,
|
|
"calib/step_q_gap": -0.014096172392223383,
|
|
"calib/step_q_w": 0.579923784901758,
|
|
"calib/step_q_w_n": 967.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.03125,
|
|
"completions/max_length": 2238.0,
|
|
"completions/max_terminated_length": 2238.0,
|
|
"completions/mean_length": 860.03125,
|
|
"completions/mean_terminated_length": 887.774169921875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 375.0,
|
|
"epoch": 0.11946666666666667,
|
|
"grad_norm": 0.27546238899230957,
|
|
"learning_rate": 2.4722222222222226e-06,
|
|
"loss": -0.0407,
|
|
"num_tokens": 29862181.0,
|
|
"reward": 1.109375,
|
|
"reward_std": 0.23978866636753082,
|
|
"rewards/accuracy_reward_step": 0.625,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 112
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.94921875,
|
|
"calib/avg_num_step_conf": 7.68359375,
|
|
"calib/ece": 0.06102880658436222,
|
|
"calib/final_conf_rate": 0.94921875,
|
|
"calib/format_rate": 0.94921875,
|
|
"calib/frac_conf_gt_0.9": 0.00411522633744856,
|
|
"calib/gap": -0.01115797546012276,
|
|
"calib/mean_conf": 0.6287654320987656,
|
|
"calib/mu_c": 0.6250920245398772,
|
|
"calib/mu_w": 0.63625,
|
|
"calib/nonempty_final_conf_rate": 0.94921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.00950617283950617,
|
|
"calib/std_conf": 0.05716302690171662,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5484249767008388,
|
|
"calib/step_q_c_n": 1073.0,
|
|
"calib/step_q_gap": -0.04518799869066015,
|
|
"calib/step_q_w": 0.5936129753914989,
|
|
"calib/step_q_w_n": 894.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.03125,
|
|
"completions/max_length": 2942.0,
|
|
"completions/max_terminated_length": 2942.0,
|
|
"completions/mean_length": 726.50390625,
|
|
"completions/mean_terminated_length": 749.9395141601562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 283.0,
|
|
"epoch": 0.12053333333333334,
|
|
"grad_norm": 0.2952890694141388,
|
|
"learning_rate": 2.4444444444444447e-06,
|
|
"loss": -0.0271,
|
|
"num_tokens": 30153366.0,
|
|
"reward": 1.111328125,
|
|
"reward_std": 0.22342383861541748,
|
|
"rewards/accuracy_reward_step": 0.63671875,
|
|
"rewards/format_reward_step": 0.94921875,
|
|
"step": 113
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.92578125,
|
|
"calib/avg_num_step_conf": 7.85546875,
|
|
"calib/ece": 0.23105485232067513,
|
|
"calib/final_conf_rate": 0.92578125,
|
|
"calib/format_rate": 0.92578125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.026989010989010964,
|
|
"calib/mean_conf": 0.6216033755274262,
|
|
"calib/mu_c": 0.6168205128205129,
|
|
"calib/mu_w": 0.6438095238095238,
|
|
"calib/nonempty_final_conf_rate": 0.92578125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.014936708860759488,
|
|
"calib/std_conf": 0.05321680508106623,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5515306122448981,
|
|
"calib/step_q_c_n": 1274.0,
|
|
"calib/step_q_gap": -0.08099991692742226,
|
|
"calib/step_q_w": 0.6325305291723203,
|
|
"calib/step_q_w_n": 737.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0625,
|
|
"completions/max_length": 2523.0,
|
|
"completions/max_terminated_length": 2523.0,
|
|
"completions/mean_length": 683.72265625,
|
|
"completions/mean_terminated_length": 729.30419921875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 251.0,
|
|
"epoch": 0.1216,
|
|
"grad_norm": 0.3420001268386841,
|
|
"learning_rate": 2.4166666666666667e-06,
|
|
"loss": -0.0587,
|
|
"num_tokens": 30433423.0,
|
|
"reward": 1.224609375,
|
|
"reward_std": 0.18992431461811066,
|
|
"rewards/accuracy_reward_step": 0.76171875,
|
|
"rewards/format_reward_step": 0.92578125,
|
|
"step": 114
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/avg_num_step_conf": 8.0703125,
|
|
"calib/ece": 0.06385542168674699,
|
|
"calib/final_conf_rate": 0.97265625,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.001348588120740013,
|
|
"calib/mean_conf": 0.6306827309236948,
|
|
"calib/mu_c": 0.6301898734177215,
|
|
"calib/mu_w": 0.6315384615384615,
|
|
"calib/nonempty_final_conf_rate": 0.97265625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.030000000000000006,
|
|
"calib/std_conf": 0.04578559846807956,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5574569319114028,
|
|
"calib/step_q_c_n": 1219.0,
|
|
"calib/step_q_gap": -0.02704129713228076,
|
|
"calib/step_q_w": 0.5844982290436835,
|
|
"calib/step_q_w_n": 847.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3010.0,
|
|
"completions/max_terminated_length": 3010.0,
|
|
"completions/mean_length": 732.890625,
|
|
"completions/mean_terminated_length": 744.5238647460938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 337.0,
|
|
"epoch": 0.12266666666666666,
|
|
"grad_norm": 0.3784162700176239,
|
|
"learning_rate": 2.388888888888889e-06,
|
|
"loss": -0.0124,
|
|
"num_tokens": 30726307.0,
|
|
"reward": 1.103515625,
|
|
"reward_std": 0.17752531170845032,
|
|
"rewards/accuracy_reward_step": 0.6171875,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 115
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96484375,
|
|
"calib/avg_num_step_conf": 7.40625,
|
|
"calib/ece": 0.06744939271255054,
|
|
"calib/final_conf_rate": 0.96484375,
|
|
"calib/format_rate": 0.96484375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.003655038759689977,
|
|
"calib/mean_conf": 0.6301214574898786,
|
|
"calib/mu_c": 0.6290116279069767,
|
|
"calib/mu_w": 0.6326666666666667,
|
|
"calib/nonempty_final_conf_rate": 0.96484375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0006072874493927086,
|
|
"calib/std_conf": 0.05397166584111702,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5624370430544273,
|
|
"calib/step_q_c_n": 1231.0,
|
|
"calib/step_q_gap": -0.013833633637302056,
|
|
"calib/step_q_w": 0.5762706766917294,
|
|
"calib/step_q_w_n": 665.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2520.0,
|
|
"completions/max_terminated_length": 2520.0,
|
|
"completions/mean_length": 802.13671875,
|
|
"completions/mean_terminated_length": 821.3880615234375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 175.0,
|
|
"epoch": 0.12373333333333333,
|
|
"grad_norm": 0.31041327118873596,
|
|
"learning_rate": 2.361111111111111e-06,
|
|
"loss": -0.0078,
|
|
"num_tokens": 31036174.0,
|
|
"reward": 1.154296875,
|
|
"reward_std": 0.23412039875984192,
|
|
"rewards/accuracy_reward_step": 0.671875,
|
|
"rewards/format_reward_step": 0.96484375,
|
|
"step": 116
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 6.58984375,
|
|
"calib/ece": 0.11896414342629488,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.00398406374501992,
|
|
"calib/gap": -0.017238835725677748,
|
|
"calib/mean_conf": 0.6310756972111553,
|
|
"calib/mu_c": 0.6242763157894737,
|
|
"calib/mu_w": 0.6415151515151515,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.07223107569721121,
|
|
"calib/std_conf": 0.05846678362848867,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5513297297297297,
|
|
"calib/step_q_c_n": 925.0,
|
|
"calib/step_q_gap": -0.007476044548485605,
|
|
"calib/step_q_w": 0.5588057742782153,
|
|
"calib/step_q_w_n": 762.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2023.0,
|
|
"completions/max_terminated_length": 2023.0,
|
|
"completions/mean_length": 713.640625,
|
|
"completions/mean_terminated_length": 730.7680053710938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 256.0,
|
|
"epoch": 0.1248,
|
|
"grad_norm": 0.33048129081726074,
|
|
"learning_rate": 2.3333333333333336e-06,
|
|
"loss": -0.0263,
|
|
"num_tokens": 31325466.0,
|
|
"reward": 1.083984375,
|
|
"reward_std": 0.2014136016368866,
|
|
"rewards/accuracy_reward_step": 0.59375,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 117
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 7.7109375,
|
|
"calib/ece": 0.1069047619047619,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.003968253968253968,
|
|
"calib/gap": -0.0239285714285713,
|
|
"calib/mean_conf": 0.6409523809523808,
|
|
"calib/mu_c": 0.6329761904761905,
|
|
"calib/mu_w": 0.6569047619047618,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04059523809523809,
|
|
"calib/std_conf": 0.059240277099613596,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5579061976549413,
|
|
"calib/step_q_c_n": 1194.0,
|
|
"calib/step_q_gap": -0.030504058755315144,
|
|
"calib/step_q_w": 0.5884102564102565,
|
|
"calib/step_q_w_n": 780.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3072.0,
|
|
"completions/max_terminated_length": 3072.0,
|
|
"completions/mean_length": 796.66015625,
|
|
"completions/mean_terminated_length": 802.9330444335938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 274.0,
|
|
"epoch": 0.12586666666666665,
|
|
"grad_norm": 0.5396471619606018,
|
|
"learning_rate": 2.305555555555556e-06,
|
|
"loss": -0.0079,
|
|
"num_tokens": 31633419.0,
|
|
"reward": 1.1484375,
|
|
"reward_std": 0.20124970376491547,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 118
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 8.42578125,
|
|
"calib/ece": 0.07161507936507941,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.027777777777777776,
|
|
"calib/gap": -0.005440476190476162,
|
|
"calib/mean_conf": 0.6469325396825396,
|
|
"calib/mu_c": 0.6451190476190476,
|
|
"calib/mu_w": 0.6505595238095238,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.02594047619047618,
|
|
"calib/std_conf": 0.07483415136827791,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5731582840236688,
|
|
"calib/step_q_c_n": 1352.0,
|
|
"calib/step_q_gap": -0.0018603495167039963,
|
|
"calib/step_q_w": 0.5750186335403727,
|
|
"calib/step_q_w_n": 805.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2675.0,
|
|
"completions/max_terminated_length": 2675.0,
|
|
"completions/mean_length": 833.93359375,
|
|
"completions/mean_terminated_length": 840.5,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 220.0,
|
|
"epoch": 0.12693333333333334,
|
|
"grad_norm": 0.5779143571853638,
|
|
"learning_rate": 2.277777777777778e-06,
|
|
"loss": -0.019,
|
|
"num_tokens": 31951970.0,
|
|
"reward": 1.1484375,
|
|
"reward_std": 0.23095275461673737,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 119
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 8.29296875,
|
|
"calib/ece": 0.0986693548387097,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.019189539308557935,
|
|
"calib/mean_conf": 0.6338306451612903,
|
|
"calib/mu_c": 0.6284916201117319,
|
|
"calib/mu_w": 0.6476811594202898,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.005362903225806452,
|
|
"calib/std_conf": 0.042162363161067036,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5576158940397351,
|
|
"calib/step_q_c_n": 1359.0,
|
|
"calib/step_q_gap": -0.0209443153843486,
|
|
"calib/step_q_w": 0.5785602094240837,
|
|
"calib/step_q_w_n": 764.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2951.0,
|
|
"completions/max_terminated_length": 2951.0,
|
|
"completions/mean_length": 764.48828125,
|
|
"completions/mean_terminated_length": 773.5534057617188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 320.0,
|
|
"epoch": 0.128,
|
|
"grad_norm": 0.2819494605064392,
|
|
"learning_rate": 2.25e-06,
|
|
"loss": 0.0236,
|
|
"num_tokens": 32254367.0,
|
|
"reward": 1.18359375,
|
|
"reward_std": 0.1648695021867752,
|
|
"rewards/accuracy_reward_step": 0.69921875,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 120
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 9.04296875,
|
|
"calib/ece": 0.10322580645161307,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.010199800199800269,
|
|
"calib/mean_conf": 0.6502419354838709,
|
|
"calib/mu_c": 0.6475274725274724,
|
|
"calib/mu_w": 0.6577272727272727,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.009798387096774209,
|
|
"calib/std_conf": 0.06128284122421192,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5725629438347322,
|
|
"calib/step_q_c_n": 1549.0,
|
|
"calib/step_q_gap": -0.013585881230542052,
|
|
"calib/step_q_w": 0.5861488250652742,
|
|
"calib/step_q_w_n": 766.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.03125,
|
|
"completions/max_length": 2167.0,
|
|
"completions/max_terminated_length": 2167.0,
|
|
"completions/mean_length": 828.71875,
|
|
"completions/mean_terminated_length": 855.4515991210938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 221.0,
|
|
"epoch": 0.12906666666666666,
|
|
"grad_norm": 0.3823467791080475,
|
|
"learning_rate": 2.222222222222222e-06,
|
|
"loss": -0.0604,
|
|
"num_tokens": 32571575.0,
|
|
"reward": 1.1953125,
|
|
"reward_std": 0.2390308976173401,
|
|
"rewards/accuracy_reward_step": 0.7109375,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 121
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9375,
|
|
"calib/avg_num_step_conf": 9.24609375,
|
|
"calib/ece": 0.13908333333333334,
|
|
"calib/final_conf_rate": 0.9375,
|
|
"calib/format_rate": 0.9375,
|
|
"calib/frac_conf_gt_0.9": 0.004166666666666667,
|
|
"calib/gap": -0.014493844049247584,
|
|
"calib/mean_conf": 0.6385833333333333,
|
|
"calib/mu_c": 0.6344767441860465,
|
|
"calib/mu_w": 0.6489705882352941,
|
|
"calib/nonempty_final_conf_rate": 0.9375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.030500000000000013,
|
|
"calib/std_conf": 0.05275408093745504,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5678004216444132,
|
|
"calib/step_q_c_n": 1423.0,
|
|
"calib/step_q_gap": -0.034244069881010364,
|
|
"calib/step_q_w": 0.6020444915254236,
|
|
"calib/step_q_w_n": 944.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.05859375,
|
|
"completions/max_length": 2797.0,
|
|
"completions/max_terminated_length": 2797.0,
|
|
"completions/mean_length": 725.13671875,
|
|
"completions/mean_terminated_length": 770.269775390625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 346.0,
|
|
"epoch": 0.13013333333333332,
|
|
"grad_norm": 0.3606873154640198,
|
|
"learning_rate": 2.1944444444444445e-06,
|
|
"loss": -0.0522,
|
|
"num_tokens": 32864554.0,
|
|
"reward": 1.140625,
|
|
"reward_std": 0.1336795687675476,
|
|
"rewards/accuracy_reward_step": 0.671875,
|
|
"rewards/format_reward_step": 0.9375,
|
|
"step": 122
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9609375,
|
|
"calib/avg_num_step_conf": 9.19140625,
|
|
"calib/ece": 0.05191056910569104,
|
|
"calib/final_conf_rate": 0.9609375,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.024390243902439025,
|
|
"calib/gap": -8.377425044092668e-05,
|
|
"calib/mean_conf": 0.664349593495935,
|
|
"calib/mu_c": 0.664320987654321,
|
|
"calib/mu_w": 0.6644047619047619,
|
|
"calib/nonempty_final_conf_rate": 0.9609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.028861788617886158,
|
|
"calib/std_conf": 0.08075472084565177,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5770537491705374,
|
|
"calib/step_q_c_n": 1507.0,
|
|
"calib/step_q_gap": -0.017237030971306577,
|
|
"calib/step_q_w": 0.594290780141844,
|
|
"calib/step_q_w_n": 846.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2890.0,
|
|
"completions/max_terminated_length": 2890.0,
|
|
"completions/mean_length": 900.34375,
|
|
"completions/mean_terminated_length": 933.1498413085938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 335.0,
|
|
"epoch": 0.1312,
|
|
"grad_norm": 0.40366265177726746,
|
|
"learning_rate": 2.166666666666667e-06,
|
|
"loss": -0.0764,
|
|
"num_tokens": 33200330.0,
|
|
"reward": 1.11328125,
|
|
"reward_std": 0.28023582696914673,
|
|
"rewards/accuracy_reward_step": 0.6328125,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 123
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 8.88671875,
|
|
"calib/ece": 0.10948000000000005,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.008,
|
|
"calib/gap": -0.02430859156057341,
|
|
"calib/mean_conf": 0.6535599999999999,
|
|
"calib/mu_c": 0.6468508287292817,
|
|
"calib/mu_w": 0.6711594202898551,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.019520000000000003,
|
|
"calib/std_conf": 0.06392281595799734,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5701692708333334,
|
|
"calib/step_q_c_n": 1536.0,
|
|
"calib/step_q_gap": -0.013836141886558462,
|
|
"calib/step_q_w": 0.5840054127198918,
|
|
"calib/step_q_w_n": 739.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2598.0,
|
|
"completions/max_terminated_length": 2598.0,
|
|
"completions/mean_length": 829.16015625,
|
|
"completions/mean_terminated_length": 845.6773071289062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 319.0,
|
|
"epoch": 0.13226666666666667,
|
|
"grad_norm": 0.22802260518074036,
|
|
"learning_rate": 2.138888888888889e-06,
|
|
"loss": -0.0256,
|
|
"num_tokens": 33519411.0,
|
|
"reward": 1.1953125,
|
|
"reward_std": 0.17892926931381226,
|
|
"rewards/accuracy_reward_step": 0.70703125,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 124
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 9.03125,
|
|
"calib/ece": 0.08841897233201577,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.003952569169960474,
|
|
"calib/gap": 0.0006696195935385063,
|
|
"calib/mean_conf": 0.6468379446640317,
|
|
"calib/mu_c": 0.6471052631578948,
|
|
"calib/mu_w": 0.6464356435643563,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0672332015810276,
|
|
"calib/std_conf": 0.0648644179068933,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5699325337331334,
|
|
"calib/step_q_c_n": 1334.0,
|
|
"calib/step_q_gap": 8.590796626228858e-05,
|
|
"calib/step_q_w": 0.5698466257668711,
|
|
"calib/step_q_w_n": 978.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2461.0,
|
|
"completions/max_terminated_length": 2461.0,
|
|
"completions/mean_length": 806.55078125,
|
|
"completions/mean_terminated_length": 816.1146850585938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 250.0,
|
|
"epoch": 0.13333333333333333,
|
|
"grad_norm": 0.2709088623523712,
|
|
"learning_rate": 2.1111111111111114e-06,
|
|
"loss": -0.0183,
|
|
"num_tokens": 33830696.0,
|
|
"reward": 1.087890625,
|
|
"reward_std": 0.2635945975780487,
|
|
"rewards/accuracy_reward_step": 0.59375,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 125
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9453125,
|
|
"calib/avg_num_step_conf": 9.6328125,
|
|
"calib/ece": 0.1561570247933885,
|
|
"calib/final_conf_rate": 0.9453125,
|
|
"calib/format_rate": 0.9453125,
|
|
"calib/frac_conf_gt_0.9": 0.004132231404958678,
|
|
"calib/gap": -0.04367865185990527,
|
|
"calib/mean_conf": 0.6445041322314048,
|
|
"calib/mu_c": 0.6302453987730061,
|
|
"calib/mu_w": 0.6739240506329114,
|
|
"calib/nonempty_final_conf_rate": 0.9453125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06355371900826447,
|
|
"calib/std_conf": 0.05899259289082415,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5620327421555252,
|
|
"calib/step_q_c_n": 1466.0,
|
|
"calib/step_q_gap": -0.061277257844474686,
|
|
"calib/step_q_w": 0.6233099999999999,
|
|
"calib/step_q_w_n": 1000.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2977.0,
|
|
"completions/max_terminated_length": 2977.0,
|
|
"completions/mean_length": 783.8984375,
|
|
"completions/mean_terminated_length": 819.0938720703125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 306.0,
|
|
"epoch": 0.1344,
|
|
"grad_norm": 0.4443349242210388,
|
|
"learning_rate": 2.0833333333333334e-06,
|
|
"loss": -0.0611,
|
|
"num_tokens": 34136838.0,
|
|
"reward": 1.109375,
|
|
"reward_std": 0.26804327964782715,
|
|
"rewards/accuracy_reward_step": 0.63671875,
|
|
"rewards/format_reward_step": 0.9453125,
|
|
"step": 126
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.95703125,
|
|
"calib/avg_num_step_conf": 9.62890625,
|
|
"calib/ece": 0.07832653061224497,
|
|
"calib/final_conf_rate": 0.95703125,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.004081632653061225,
|
|
"calib/gap": -0.01936545664073752,
|
|
"calib/mean_conf": 0.6517142857142858,
|
|
"calib/mu_c": 0.6446794871794871,
|
|
"calib/mu_w": 0.6640449438202246,
|
|
"calib/nonempty_final_conf_rate": 0.95703125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.046653061224489804,
|
|
"calib/std_conf": 0.0549166344079267,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5608229426433916,
|
|
"calib/step_q_c_n": 1203.0,
|
|
"calib/step_q_gap": -0.04648292106500784,
|
|
"calib/step_q_w": 0.6073058637083995,
|
|
"calib/step_q_w_n": 1262.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2831.0,
|
|
"completions/max_terminated_length": 2831.0,
|
|
"completions/mean_length": 780.3515625,
|
|
"completions/mean_terminated_length": 812.0731201171875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 292.0,
|
|
"epoch": 0.13546666666666668,
|
|
"grad_norm": 0.25136062502861023,
|
|
"learning_rate": 2.0555555555555555e-06,
|
|
"loss": -0.0525,
|
|
"num_tokens": 34440280.0,
|
|
"reward": 1.087890625,
|
|
"reward_std": 0.2381621152162552,
|
|
"rewards/accuracy_reward_step": 0.609375,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 127
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 9.46875,
|
|
"calib/ece": 0.12346774193548388,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.033895837749982505,
|
|
"calib/mean_conf": 0.6504032258064515,
|
|
"calib/mu_c": 0.6382389937106917,
|
|
"calib/mu_w": 0.6721348314606742,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06637096774193549,
|
|
"calib/std_conf": 0.06796463968608007,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5672322070452911,
|
|
"calib/step_q_c_n": 1391.0,
|
|
"calib/step_q_gap": -0.018198577078619826,
|
|
"calib/step_q_w": 0.5854307841239109,
|
|
"calib/step_q_w_n": 1033.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.03125,
|
|
"completions/max_length": 3007.0,
|
|
"completions/max_terminated_length": 3007.0,
|
|
"completions/mean_length": 781.79296875,
|
|
"completions/mean_terminated_length": 807.0120849609375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 291.0,
|
|
"epoch": 0.13653333333333334,
|
|
"grad_norm": 0.31424975395202637,
|
|
"learning_rate": 2.027777777777778e-06,
|
|
"loss": -0.045,
|
|
"num_tokens": 34747083.0,
|
|
"reward": 1.10546875,
|
|
"reward_std": 0.2466757893562317,
|
|
"rewards/accuracy_reward_step": 0.62109375,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 128
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 8.7421875,
|
|
"calib/ece": 0.021764705882352867,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.011764705882352941,
|
|
"calib/gap": 0.02230303030303027,
|
|
"calib/mean_conf": 0.6444313725490196,
|
|
"calib/mu_c": 0.6523030303030303,
|
|
"calib/mu_w": 0.63,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.009568627450980395,
|
|
"calib/std_conf": 0.06580698303191819,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5802972972972973,
|
|
"calib/step_q_c_n": 1480.0,
|
|
"calib/step_q_gap": 0.016247165371175876,
|
|
"calib/step_q_w": 0.5640501319261214,
|
|
"calib/step_q_w_n": 758.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2272.0,
|
|
"completions/max_terminated_length": 2272.0,
|
|
"completions/mean_length": 745.8671875,
|
|
"completions/mean_terminated_length": 748.7921752929688,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 252.0,
|
|
"epoch": 0.1376,
|
|
"grad_norm": 0.3294617235660553,
|
|
"learning_rate": 2.0000000000000003e-06,
|
|
"loss": 0.0198,
|
|
"num_tokens": 35040409.0,
|
|
"reward": 1.142578125,
|
|
"reward_std": 0.21977511048316956,
|
|
"rewards/accuracy_reward_step": 0.64453125,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 129
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 9.9296875,
|
|
"calib/ece": 0.13968253968253969,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.003968253968253968,
|
|
"calib/gap": -0.03529761904761908,
|
|
"calib/mean_conf": 0.6415873015873016,
|
|
"calib/mu_c": 0.6298214285714285,
|
|
"calib/mu_w": 0.6651190476190476,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.057301587301587305,
|
|
"calib/std_conf": 0.04980776854070341,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5638549361987911,
|
|
"calib/step_q_c_n": 1489.0,
|
|
"calib/step_q_gap": -0.029535377191522327,
|
|
"calib/step_q_w": 0.5933903133903135,
|
|
"calib/step_q_w_n": 1053.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2988.0,
|
|
"completions/max_terminated_length": 2988.0,
|
|
"completions/mean_length": 718.37109375,
|
|
"completions/mean_terminated_length": 726.8893432617188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 276.0,
|
|
"epoch": 0.13866666666666666,
|
|
"grad_norm": 0.5593166351318359,
|
|
"learning_rate": 1.9722222222222224e-06,
|
|
"loss": -0.0497,
|
|
"num_tokens": 35329600.0,
|
|
"reward": 1.1484375,
|
|
"reward_std": 0.12416224181652069,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 130
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9609375,
|
|
"calib/avg_num_step_conf": 11.15625,
|
|
"calib/ece": 0.17669387755102034,
|
|
"calib/final_conf_rate": 0.95703125,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.004081632653061225,
|
|
"calib/gap": -0.024859370834444228,
|
|
"calib/mean_conf": 0.6449387755102041,
|
|
"calib/mu_c": 0.6326612903225806,
|
|
"calib/mu_w": 0.6575206611570248,
|
|
"calib/nonempty_final_conf_rate": 0.95703125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.15775510204081622,
|
|
"calib/std_conf": 0.057906472150079845,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5670183852917666,
|
|
"calib/step_q_c_n": 1251.0,
|
|
"calib/step_q_gap": -0.031018997885803423,
|
|
"calib/step_q_w": 0.59803738317757,
|
|
"calib/step_q_w_n": 1605.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3018.0,
|
|
"completions/max_terminated_length": 3018.0,
|
|
"completions/mean_length": 752.140625,
|
|
"completions/mean_terminated_length": 779.5465698242188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 370.0,
|
|
"epoch": 0.13973333333333332,
|
|
"grad_norm": 0.8481810092926025,
|
|
"learning_rate": 1.944444444444445e-06,
|
|
"loss": -0.0373,
|
|
"num_tokens": 35628356.0,
|
|
"reward": 0.962890625,
|
|
"reward_std": 0.2541370987892151,
|
|
"rewards/accuracy_reward_step": 0.484375,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 131
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 8.69921875,
|
|
"calib/ece": 0.17771811023622053,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.003937007874015748,
|
|
"calib/gap": -0.010880097087378693,
|
|
"calib/mean_conf": 0.6447259842519685,
|
|
"calib/mu_c": 0.6426699029126213,
|
|
"calib/mu_w": 0.65355,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.005710236220472421,
|
|
"calib/std_conf": 0.06773103036588879,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5663284457478006,
|
|
"calib/step_q_c_n": 1705.0,
|
|
"calib/step_q_gap": -0.006469255401624685,
|
|
"calib/step_q_w": 0.5727977011494253,
|
|
"calib/step_q_w_n": 522.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2595.0,
|
|
"completions/max_terminated_length": 2595.0,
|
|
"completions/mean_length": 764.26953125,
|
|
"completions/mean_terminated_length": 770.2874145507812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 279.0,
|
|
"epoch": 0.1408,
|
|
"grad_norm": 0.27393683791160583,
|
|
"learning_rate": 1.916666666666667e-06,
|
|
"loss": 0.0111,
|
|
"num_tokens": 35929601.0,
|
|
"reward": 1.30078125,
|
|
"reward_std": 0.21899119019508362,
|
|
"rewards/accuracy_reward_step": 0.8046875,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 132
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9609375,
|
|
"calib/avg_num_step_conf": 11.84765625,
|
|
"calib/ece": 0.12471544715447158,
|
|
"calib/final_conf_rate": 0.9609375,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.008130081300813009,
|
|
"calib/gap": -0.01428610224726734,
|
|
"calib/mean_conf": 0.6692682926829268,
|
|
"calib/mu_c": 0.6632867132867133,
|
|
"calib/mu_w": 0.6775728155339806,
|
|
"calib/nonempty_final_conf_rate": 0.9609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.10634146341463417,
|
|
"calib/std_conf": 0.06447626060526396,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5788989053444945,
|
|
"calib/step_q_c_n": 1553.0,
|
|
"calib/step_q_gap": -0.022053797358208227,
|
|
"calib/step_q_w": 0.6009527027027027,
|
|
"calib/step_q_w_n": 1480.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2494.0,
|
|
"completions/max_terminated_length": 2494.0,
|
|
"completions/mean_length": 911.9453125,
|
|
"completions/mean_terminated_length": 945.1741333007812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 356.0,
|
|
"epoch": 0.14186666666666667,
|
|
"grad_norm": 0.6605852842330933,
|
|
"learning_rate": 1.888888888888889e-06,
|
|
"loss": -0.0547,
|
|
"num_tokens": 36269403.0,
|
|
"reward": 1.0390625,
|
|
"reward_std": 0.33969852328300476,
|
|
"rewards/accuracy_reward_step": 0.55859375,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 133
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 12.14453125,
|
|
"calib/ece": 0.06972222222222227,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.011904761904761904,
|
|
"calib/gap": 0.00014432989690715825,
|
|
"calib/mean_conf": 0.6719444444444443,
|
|
"calib/mu_c": 0.6719999999999999,
|
|
"calib/mu_w": 0.6718556701030928,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06329365079365083,
|
|
"calib/std_conf": 0.07139614563821317,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5890830945558739,
|
|
"calib/step_q_c_n": 1745.0,
|
|
"calib/step_q_gap": -0.010902242687527819,
|
|
"calib/step_q_w": 0.5999853372434018,
|
|
"calib/step_q_w_n": 1364.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2429.0,
|
|
"completions/max_terminated_length": 2429.0,
|
|
"completions/mean_length": 906.65625,
|
|
"completions/mean_terminated_length": 921.0476684570312,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 355.0,
|
|
"epoch": 0.14293333333333333,
|
|
"grad_norm": 1.3199166059494019,
|
|
"learning_rate": 1.8611111111111113e-06,
|
|
"loss": -0.0304,
|
|
"num_tokens": 36610459.0,
|
|
"reward": 1.09765625,
|
|
"reward_std": 0.30026212334632874,
|
|
"rewards/accuracy_reward_step": 0.60546875,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 134
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/avg_num_step_conf": 10.6484375,
|
|
"calib/ece": 0.075301204819277,
|
|
"calib/final_conf_rate": 0.97265625,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.004016064257028112,
|
|
"calib/gap": -0.016020104244229505,
|
|
"calib/mean_conf": 0.651847389558233,
|
|
"calib/mu_c": 0.6467647058823529,
|
|
"calib/mu_w": 0.6627848101265824,
|
|
"calib/nonempty_final_conf_rate": 0.97265625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.02220883534136544,
|
|
"calib/std_conf": 0.05533215514313524,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5762599145820623,
|
|
"calib/step_q_c_n": 1639.0,
|
|
"calib/step_q_gap": -0.01798718753385309,
|
|
"calib/step_q_w": 0.5942471021159154,
|
|
"calib/step_q_w_n": 1087.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2484.0,
|
|
"completions/max_terminated_length": 2484.0,
|
|
"completions/mean_length": 779.2734375,
|
|
"completions/mean_terminated_length": 801.1806640625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 339.0,
|
|
"epoch": 0.144,
|
|
"grad_norm": 22.018888473510742,
|
|
"learning_rate": 1.8333333333333333e-06,
|
|
"loss": -0.0541,
|
|
"num_tokens": 36915833.0,
|
|
"reward": 1.150390625,
|
|
"reward_std": 0.17979279160499573,
|
|
"rewards/accuracy_reward_step": 0.6640625,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 135
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 11.046875,
|
|
"calib/ece": 0.09971999999999998,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.004,
|
|
"calib/gap": -0.027167981017516718,
|
|
"calib/mean_conf": 0.6575599999999999,
|
|
"calib/mu_c": 0.647888198757764,
|
|
"calib/mu_w": 0.6750561797752808,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.056639999999999996,
|
|
"calib/std_conf": 0.06205196531939985,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5717223974763407,
|
|
"calib/step_q_c_n": 1585.0,
|
|
"calib/step_q_gap": -0.04463319383500286,
|
|
"calib/step_q_w": 0.6163555913113435,
|
|
"calib/step_q_w_n": 1243.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2909.0,
|
|
"completions/max_terminated_length": 2909.0,
|
|
"completions/mean_length": 792.5234375,
|
|
"completions/mean_terminated_length": 808.310791015625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 259.0,
|
|
"epoch": 0.14506666666666668,
|
|
"grad_norm": 8.103403091430664,
|
|
"learning_rate": 1.8055555555555557e-06,
|
|
"loss": -0.0398,
|
|
"num_tokens": 37227207.0,
|
|
"reward": 1.1171875,
|
|
"reward_std": 0.19905364513397217,
|
|
"rewards/accuracy_reward_step": 0.62890625,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 136
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 11.69921875,
|
|
"calib/ece": 0.05221774193548386,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.008064516129032258,
|
|
"calib/gap": 0.005273477812177574,
|
|
"calib/mean_conf": 0.6664112903225806,
|
|
"calib/mu_c": 0.6684313725490196,
|
|
"calib/mu_w": 0.6631578947368421,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.050846774193548376,
|
|
"calib/std_conf": 0.06967974113141867,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5858468957203135,
|
|
"calib/step_q_c_n": 1659.0,
|
|
"calib/step_q_gap": -0.026376158171902198,
|
|
"calib/step_q_w": 0.6122230538922157,
|
|
"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": 0.02734375,
|
|
"completions/max_length": 2503.0,
|
|
"completions/max_terminated_length": 2503.0,
|
|
"completions/mean_length": 805.234375,
|
|
"completions/mean_terminated_length": 827.8714599609375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 300.0,
|
|
"epoch": 0.14613333333333334,
|
|
"grad_norm": 6.265161514282227,
|
|
"learning_rate": 1.777777777777778e-06,
|
|
"loss": -0.0268,
|
|
"num_tokens": 37540331.0,
|
|
"reward": 1.08203125,
|
|
"reward_std": 0.1679389476776123,
|
|
"rewards/accuracy_reward_step": 0.59765625,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 137
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 10.09375,
|
|
"calib/ece": 0.10435483870967743,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.020161290322580645,
|
|
"calib/gap": 0.02305802170208937,
|
|
"calib/mean_conf": 0.6577419354838709,
|
|
"calib/mu_c": 0.6632275132275132,
|
|
"calib/mu_w": 0.6401694915254238,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0,
|
|
"calib/std_conf": 0.08042566904549314,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.598025641025641,
|
|
"calib/step_q_c_n": 1950.0,
|
|
"calib/step_q_gap": 0.00047043597832252537,
|
|
"calib/step_q_w": 0.5975552050473185,
|
|
"calib/step_q_w_n": 634.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.03125,
|
|
"completions/max_length": 2803.0,
|
|
"completions/max_terminated_length": 2803.0,
|
|
"completions/mean_length": 752.515625,
|
|
"completions/mean_terminated_length": 776.790283203125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 207.0,
|
|
"epoch": 0.1472,
|
|
"grad_norm": 4.048952579498291,
|
|
"learning_rate": 1.75e-06,
|
|
"loss": -0.0587,
|
|
"num_tokens": 37837311.0,
|
|
"reward": 1.22265625,
|
|
"reward_std": 0.16601943969726562,
|
|
"rewards/accuracy_reward_step": 0.73828125,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 138
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 9.109375,
|
|
"calib/ece": 0.08968253968253964,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.013015030946065376,
|
|
"calib/mean_conf": 0.6362698412698412,
|
|
"calib/mu_c": 0.6322413793103447,
|
|
"calib/mu_w": 0.6452564102564101,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.017738095238095233,
|
|
"calib/std_conf": 0.0440730657226954,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5634498680738786,
|
|
"calib/step_q_c_n": 1516.0,
|
|
"calib/step_q_gap": -0.01664817114180761,
|
|
"calib/step_q_w": 0.5800980392156863,
|
|
"calib/step_q_w_n": 816.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2489.0,
|
|
"completions/max_terminated_length": 2489.0,
|
|
"completions/mean_length": 678.41015625,
|
|
"completions/mean_terminated_length": 689.1785888671875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 278.0,
|
|
"epoch": 0.14826666666666666,
|
|
"grad_norm": 3.1196072101593018,
|
|
"learning_rate": 1.7222222222222224e-06,
|
|
"loss": -0.0342,
|
|
"num_tokens": 38114080.0,
|
|
"reward": 1.171875,
|
|
"reward_std": 0.21596798300743103,
|
|
"rewards/accuracy_reward_step": 0.6796875,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 139
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 10.58203125,
|
|
"calib/ece": 0.08480314960629928,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.003937007874015748,
|
|
"calib/gap": -0.018391330891330937,
|
|
"calib/mean_conf": 0.6554330708661418,
|
|
"calib/mu_c": 0.6502197802197802,
|
|
"calib/mu_w": 0.6686111111111112,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.011850393700787384,
|
|
"calib/std_conf": 0.0541978862578768,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5736004331348132,
|
|
"calib/step_q_c_n": 1847.0,
|
|
"calib/step_q_gap": -0.02481023971901508,
|
|
"calib/step_q_w": 0.5984106728538283,
|
|
"calib/step_q_w_n": 862.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3044.0,
|
|
"completions/max_terminated_length": 3044.0,
|
|
"completions/mean_length": 797.0,
|
|
"completions/mean_terminated_length": 803.2755737304688,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 268.0,
|
|
"epoch": 0.14933333333333335,
|
|
"grad_norm": 0.7780373096466064,
|
|
"learning_rate": 1.6944444444444446e-06,
|
|
"loss": -0.0067,
|
|
"num_tokens": 38423128.0,
|
|
"reward": 1.20703125,
|
|
"reward_std": 0.18714842200279236,
|
|
"rewards/accuracy_reward_step": 0.7109375,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 140
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.97265625,
|
|
"calib/avg_num_step_conf": 13.09765625,
|
|
"calib/ece": 0.11514056224899591,
|
|
"calib/final_conf_rate": 0.97265625,
|
|
"calib/format_rate": 0.97265625,
|
|
"calib/frac_conf_gt_0.9": 0.008032128514056224,
|
|
"calib/gap": -0.014134290540540495,
|
|
"calib/mean_conf": 0.6630923694779117,
|
|
"calib/mu_c": 0.6594594594594595,
|
|
"calib/mu_w": 0.67359375,
|
|
"calib/nonempty_final_conf_rate": 0.97265625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.01763052208835339,
|
|
"calib/std_conf": 0.06142611964552913,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5822583686940123,
|
|
"calib/step_q_c_n": 2121.0,
|
|
"calib/step_q_gap": -0.04230331961767608,
|
|
"calib/step_q_w": 0.6245616883116883,
|
|
"calib/step_q_w_n": 1232.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2920.0,
|
|
"completions/max_terminated_length": 2920.0,
|
|
"completions/mean_length": 835.42578125,
|
|
"completions/mean_terminated_length": 852.0677490234375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 322.0,
|
|
"epoch": 0.1504,
|
|
"grad_norm": 1.050325870513916,
|
|
"learning_rate": 1.6666666666666667e-06,
|
|
"loss": -0.0231,
|
|
"num_tokens": 38744093.0,
|
|
"reward": 1.208984375,
|
|
"reward_std": 0.14863857626914978,
|
|
"rewards/accuracy_reward_step": 0.72265625,
|
|
"rewards/format_reward_step": 0.97265625,
|
|
"step": 141
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 11.23046875,
|
|
"calib/ece": 0.10089843749999997,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.00390625,
|
|
"calib/gap": -0.020271330500163387,
|
|
"calib/mean_conf": 0.6598828125,
|
|
"calib/mu_c": 0.652360248447205,
|
|
"calib/mu_w": 0.6726315789473684,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06593750000000001,
|
|
"calib/std_conf": 0.06411343963702029,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5813268236645606,
|
|
"calib/step_q_c_n": 1741.0,
|
|
"calib/step_q_gap": -0.0040788200744165115,
|
|
"calib/step_q_w": 0.5854056437389771,
|
|
"calib/step_q_w_n": 1134.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2251.0,
|
|
"completions/max_terminated_length": 2251.0,
|
|
"completions/mean_length": 833.3828125,
|
|
"completions/mean_terminated_length": 839.9448852539062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 218.0,
|
|
"epoch": 0.15146666666666667,
|
|
"grad_norm": 0.21071504056453705,
|
|
"learning_rate": 1.638888888888889e-06,
|
|
"loss": 0.0005,
|
|
"num_tokens": 39062599.0,
|
|
"reward": 1.12890625,
|
|
"reward_std": 0.17084911465644836,
|
|
"rewards/accuracy_reward_step": 0.62890625,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 142
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96484375,
|
|
"calib/avg_num_step_conf": 12.16015625,
|
|
"calib/ece": 0.1629149797570851,
|
|
"calib/final_conf_rate": 0.96484375,
|
|
"calib/format_rate": 0.96484375,
|
|
"calib/frac_conf_gt_0.9": 0.012145748987854251,
|
|
"calib/gap": -0.041809460062472215,
|
|
"calib/mean_conf": 0.6544939271255061,
|
|
"calib/mu_c": 0.6407831325301203,
|
|
"calib/mu_w": 0.6825925925925925,
|
|
"calib/nonempty_final_conf_rate": 0.96484375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.07267206477732792,
|
|
"calib/std_conf": 0.06410112648434106,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5699192618223761,
|
|
"calib/step_q_c_n": 1734.0,
|
|
"calib/step_q_gap": -0.05538893252135124,
|
|
"calib/step_q_w": 0.6253081943437273,
|
|
"calib/step_q_w_n": 1379.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2411.0,
|
|
"completions/max_terminated_length": 2411.0,
|
|
"completions/mean_length": 781.33203125,
|
|
"completions/mean_terminated_length": 809.8016357421875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 296.0,
|
|
"epoch": 0.15253333333333333,
|
|
"grad_norm": 0.2845354378223419,
|
|
"learning_rate": 1.6111111111111113e-06,
|
|
"loss": -0.0763,
|
|
"num_tokens": 39369956.0,
|
|
"reward": 1.130859375,
|
|
"reward_std": 0.22284802794456482,
|
|
"rewards/accuracy_reward_step": 0.6484375,
|
|
"rewards/format_reward_step": 0.96484375,
|
|
"step": 143
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 10.984375,
|
|
"calib/ece": 0.11803921568627448,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.005136408730158748,
|
|
"calib/mean_conf": 0.6484705882352939,
|
|
"calib/mu_c": 0.6497395833333334,
|
|
"calib/mu_w": 0.6446031746031746,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0067843137254901984,
|
|
"calib/std_conf": 0.053089031008104394,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.577141472868217,
|
|
"calib/step_q_c_n": 2064.0,
|
|
"calib/step_q_gap": -0.005759596650499521,
|
|
"calib/step_q_w": 0.5829010695187166,
|
|
"calib/step_q_w_n": 748.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.0078125,
|
|
"completions/max_length": 1968.0,
|
|
"completions/max_terminated_length": 1968.0,
|
|
"completions/mean_length": 760.26953125,
|
|
"completions/mean_terminated_length": 766.2559204101562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 199.0,
|
|
"epoch": 0.1536,
|
|
"grad_norm": 0.31746000051498413,
|
|
"learning_rate": 1.5833333333333333e-06,
|
|
"loss": -0.0168,
|
|
"num_tokens": 39668713.0,
|
|
"reward": 1.248046875,
|
|
"reward_std": 0.1931958794593811,
|
|
"rewards/accuracy_reward_step": 0.75,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 144
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 11.6953125,
|
|
"calib/ece": 0.11761904761904765,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.011904761904761904,
|
|
"calib/gap": -0.00792102015631424,
|
|
"calib/mean_conf": 0.663968253968254,
|
|
"calib/mu_c": 0.6619251336898396,
|
|
"calib/mu_w": 0.6698461538461539,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.019761904761904737,
|
|
"calib/std_conf": 0.06784762766295559,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5882366697848456,
|
|
"calib/step_q_c_n": 2138.0,
|
|
"calib/step_q_gap": -0.011109124607677767,
|
|
"calib/step_q_w": 0.5993457943925233,
|
|
"calib/step_q_w_n": 856.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2386.0,
|
|
"completions/max_terminated_length": 2386.0,
|
|
"completions/mean_length": 785.81640625,
|
|
"completions/mean_terminated_length": 798.2897338867188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 253.0,
|
|
"epoch": 0.15466666666666667,
|
|
"grad_norm": 0.18704430758953094,
|
|
"learning_rate": 1.5555555555555558e-06,
|
|
"loss": -0.0404,
|
|
"num_tokens": 39972586.0,
|
|
"reward": 1.22265625,
|
|
"reward_std": 0.16047047078609467,
|
|
"rewards/accuracy_reward_step": 0.73046875,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 145
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.95703125,
|
|
"calib/avg_num_step_conf": 12.0390625,
|
|
"calib/ece": 0.11702040816326532,
|
|
"calib/final_conf_rate": 0.95703125,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.004081632653061225,
|
|
"calib/gap": -0.0030535279805352644,
|
|
"calib/mean_conf": 0.6599591836734694,
|
|
"calib/mu_c": 0.6586131386861314,
|
|
"calib/mu_w": 0.6616666666666666,
|
|
"calib/nonempty_final_conf_rate": 0.95703125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.10889795918367348,
|
|
"calib/std_conf": 0.0667343069326609,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5916322580645161,
|
|
"calib/step_q_c_n": 1550.0,
|
|
"calib/step_q_gap": -0.02187296386759885,
|
|
"calib/step_q_w": 0.613505221932115,
|
|
"calib/step_q_w_n": 1532.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2737.0,
|
|
"completions/max_terminated_length": 2737.0,
|
|
"completions/mean_length": 779.09765625,
|
|
"completions/mean_terminated_length": 814.0775146484375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 279.0,
|
|
"epoch": 0.15573333333333333,
|
|
"grad_norm": 0.2517707347869873,
|
|
"learning_rate": 1.527777777777778e-06,
|
|
"loss": -0.064,
|
|
"num_tokens": 40279251.0,
|
|
"reward": 1.013671875,
|
|
"reward_std": 0.20814210176467896,
|
|
"rewards/accuracy_reward_step": 0.53515625,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 146
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 11.53515625,
|
|
"calib/ece": 0.039439999999999954,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.004,
|
|
"calib/gap": 0.003721590909090855,
|
|
"calib/mean_conf": 0.6544800000000001,
|
|
"calib/mu_c": 0.6559090909090909,
|
|
"calib/mu_w": 0.6521875,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.03895999999999995,
|
|
"calib/std_conf": 0.05965844114624518,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5829066171923315,
|
|
"calib/step_q_c_n": 1617.0,
|
|
"calib/step_q_gap": -0.026127813945393097,
|
|
"calib/step_q_w": 0.6090344311377246,
|
|
"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": 0.015625,
|
|
"completions/max_length": 3068.0,
|
|
"completions/max_terminated_length": 3068.0,
|
|
"completions/mean_length": 786.4453125,
|
|
"completions/mean_terminated_length": 798.9285888671875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 346.0,
|
|
"epoch": 0.1568,
|
|
"grad_norm": 0.27295610308647156,
|
|
"learning_rate": 1.5e-06,
|
|
"loss": 0.0041,
|
|
"num_tokens": 40584261.0,
|
|
"reward": 1.08984375,
|
|
"reward_std": 0.14193546772003174,
|
|
"rewards/accuracy_reward_step": 0.6015625,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 147
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 12.4453125,
|
|
"calib/ece": 0.1605533596837945,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.003952569169960474,
|
|
"calib/gap": -0.020801169590643376,
|
|
"calib/mean_conf": 0.656600790513834,
|
|
"calib/mu_c": 0.6514210526315789,
|
|
"calib/mu_w": 0.6722222222222223,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.03308300395256916,
|
|
"calib/std_conf": 0.06462652448769993,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5864745011086474,
|
|
"calib/step_q_c_n": 2255.0,
|
|
"calib/step_q_gap": -0.01205396290853844,
|
|
"calib/step_q_w": 0.5985284640171858,
|
|
"calib/step_q_w_n": 931.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2960.0,
|
|
"completions/max_terminated_length": 2960.0,
|
|
"completions/mean_length": 754.83984375,
|
|
"completions/mean_terminated_length": 760.783447265625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 185.0,
|
|
"epoch": 0.15786666666666666,
|
|
"grad_norm": 0.25415608286857605,
|
|
"learning_rate": 1.4722222222222225e-06,
|
|
"loss": 0.006,
|
|
"num_tokens": 40882612.0,
|
|
"reward": 1.236328125,
|
|
"reward_std": 0.17967022955417633,
|
|
"rewards/accuracy_reward_step": 0.7421875,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 148
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 12.25390625,
|
|
"calib/ece": 0.14382470119521903,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.01593625498007968,
|
|
"calib/gap": -0.048063143752798854,
|
|
"calib/mean_conf": 0.6673306772908366,
|
|
"calib/mu_c": 0.6525862068965518,
|
|
"calib/mu_w": 0.7006493506493506,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.058964143426294795,
|
|
"calib/std_conf": 0.07435008054622753,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5768194217130387,
|
|
"calib/step_q_c_n": 1833.0,
|
|
"calib/step_q_gap": -0.04203418258143987,
|
|
"calib/step_q_w": 0.6188536042944786,
|
|
"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": 0.01953125,
|
|
"completions/max_length": 2585.0,
|
|
"completions/max_terminated_length": 2585.0,
|
|
"completions/mean_length": 835.87890625,
|
|
"completions/mean_terminated_length": 852.5299072265625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 281.0,
|
|
"epoch": 0.15893333333333334,
|
|
"grad_norm": 0.9107939600944519,
|
|
"learning_rate": 1.4444444444444445e-06,
|
|
"loss": -0.0662,
|
|
"num_tokens": 41201053.0,
|
|
"reward": 1.169921875,
|
|
"reward_std": 0.20084260404109955,
|
|
"rewards/accuracy_reward_step": 0.6796875,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 149
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 11.91015625,
|
|
"calib/ece": 0.12536000000000005,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.02,
|
|
"calib/gap": -0.04635688011410555,
|
|
"calib/mean_conf": 0.64896,
|
|
"calib/mu_c": 0.6346820809248556,
|
|
"calib/mu_w": 0.6810389610389611,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04115999999999999,
|
|
"calib/std_conf": 0.07070868687792187,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5711497476163769,
|
|
"calib/step_q_c_n": 1783.0,
|
|
"calib/step_q_gap": -0.05457695064586632,
|
|
"calib/step_q_w": 0.6257266982622433,
|
|
"calib/step_q_w_n": 1266.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2558.0,
|
|
"completions/max_terminated_length": 2558.0,
|
|
"completions/mean_length": 701.390625,
|
|
"completions/mean_terminated_length": 721.1083984375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 243.0,
|
|
"epoch": 0.16,
|
|
"grad_norm": 0.2597368657588959,
|
|
"learning_rate": 1.4166666666666667e-06,
|
|
"loss": -0.069,
|
|
"num_tokens": 41485569.0,
|
|
"reward": 1.1640625,
|
|
"reward_std": 0.1759205013513565,
|
|
"rewards/accuracy_reward_step": 0.67578125,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 150
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.94140625,
|
|
"calib/avg_num_step_conf": 14.2578125,
|
|
"calib/ece": 0.10184647302904559,
|
|
"calib/final_conf_rate": 0.94140625,
|
|
"calib/format_rate": 0.94140625,
|
|
"calib/frac_conf_gt_0.9": 0.03319502074688797,
|
|
"calib/gap": -0.020073888404533613,
|
|
"calib/mean_conf": 0.670253112033195,
|
|
"calib/mu_c": 0.6625067567567566,
|
|
"calib/mu_w": 0.6825806451612902,
|
|
"calib/nonempty_final_conf_rate": 0.94140625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.07899585062240659,
|
|
"calib/std_conf": 0.1014594242664176,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6037099841521395,
|
|
"calib/step_q_c_n": 1893.0,
|
|
"calib/step_q_gap": -0.04592575859117287,
|
|
"calib/step_q_w": 0.6496357427433124,
|
|
"calib/step_q_w_n": 1757.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2994.0,
|
|
"completions/max_terminated_length": 2994.0,
|
|
"completions/mean_length": 826.16015625,
|
|
"completions/mean_terminated_length": 873.9545288085938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 308.0,
|
|
"epoch": 0.16106666666666666,
|
|
"grad_norm": 0.4585876762866974,
|
|
"learning_rate": 1.3888888888888892e-06,
|
|
"loss": -0.1048,
|
|
"num_tokens": 41804090.0,
|
|
"reward": 1.048828125,
|
|
"reward_std": 0.20975670218467712,
|
|
"rewards/accuracy_reward_step": 0.578125,
|
|
"rewards/format_reward_step": 0.94140625,
|
|
"step": 151
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.94140625,
|
|
"calib/avg_num_step_conf": 14.57421875,
|
|
"calib/ece": 0.1368181818181818,
|
|
"calib/final_conf_rate": 0.9453125,
|
|
"calib/format_rate": 0.94140625,
|
|
"calib/frac_conf_gt_0.9": 0.012396694214876033,
|
|
"calib/gap": -0.040026748804409484,
|
|
"calib/mean_conf": 0.6561570247933884,
|
|
"calib/mu_c": 0.6440828402366864,
|
|
"calib/mu_w": 0.6841095890410959,
|
|
"calib/nonempty_final_conf_rate": 0.9453125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04731404958677685,
|
|
"calib/std_conf": 0.07951105835709585,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5904891578416541,
|
|
"calib/step_q_c_n": 1983.0,
|
|
"calib/step_q_gap": -0.08152457213546249,
|
|
"calib/step_q_w": 0.6720137299771166,
|
|
"calib/step_q_w_n": 1748.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2633.0,
|
|
"completions/max_terminated_length": 2633.0,
|
|
"completions/mean_length": 767.5,
|
|
"completions/mean_terminated_length": 811.9008178710938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 306.0,
|
|
"epoch": 0.16213333333333332,
|
|
"grad_norm": 0.5712304711341858,
|
|
"learning_rate": 1.3611111111111112e-06,
|
|
"loss": -0.0978,
|
|
"num_tokens": 42105962.0,
|
|
"reward": 1.130859375,
|
|
"reward_std": 0.284695029258728,
|
|
"rewards/accuracy_reward_step": 0.66015625,
|
|
"rewards/format_reward_step": 0.94140625,
|
|
"step": 152
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.95703125,
|
|
"calib/avg_num_step_conf": 14.12109375,
|
|
"calib/ece": 0.06963265306122443,
|
|
"calib/final_conf_rate": 0.95703125,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.0163265306122449,
|
|
"calib/gap": -0.01282914493440801,
|
|
"calib/mean_conf": 0.6707755102040817,
|
|
"calib/mu_c": 0.6669005847953217,
|
|
"calib/mu_w": 0.6797297297297297,
|
|
"calib/nonempty_final_conf_rate": 0.95703125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.021224489795918337,
|
|
"calib/std_conf": 0.06483110405747476,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.591789375582479,
|
|
"calib/step_q_c_n": 2146.0,
|
|
"calib/step_q_gap": -0.05262178847470278,
|
|
"calib/step_q_w": 0.6444111640571818,
|
|
"calib/step_q_w_n": 1469.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2854.0,
|
|
"completions/max_terminated_length": 2854.0,
|
|
"completions/mean_length": 878.5859375,
|
|
"completions/mean_terminated_length": 910.5992431640625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 342.0,
|
|
"epoch": 0.1632,
|
|
"grad_norm": 0.36940431594848633,
|
|
"learning_rate": 1.3333333333333334e-06,
|
|
"loss": -0.022,
|
|
"num_tokens": 42438200.0,
|
|
"reward": 1.146484375,
|
|
"reward_std": 0.23738202452659607,
|
|
"rewards/accuracy_reward_step": 0.66796875,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 153
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9453125,
|
|
"calib/avg_num_step_conf": 13.66796875,
|
|
"calib/ece": 0.06636363636363632,
|
|
"calib/final_conf_rate": 0.9453125,
|
|
"calib/format_rate": 0.9453125,
|
|
"calib/frac_conf_gt_0.9": 0.008264462809917356,
|
|
"calib/gap": -0.0007320205479453001,
|
|
"calib/mean_conf": 0.660495867768595,
|
|
"calib/mu_c": 0.6602054794520548,
|
|
"calib/mu_w": 0.6609375000000001,
|
|
"calib/nonempty_final_conf_rate": 0.9453125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06177685950413221,
|
|
"calib/std_conf": 0.060968102209736655,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5861912658927585,
|
|
"calib/step_q_c_n": 1809.0,
|
|
"calib/step_q_gap": -0.0726430536338688,
|
|
"calib/step_q_w": 0.6588343195266273,
|
|
"calib/step_q_w_n": 1690.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2580.0,
|
|
"completions/max_terminated_length": 2580.0,
|
|
"completions/mean_length": 780.71875,
|
|
"completions/mean_terminated_length": 819.1146850585938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 346.0,
|
|
"epoch": 0.16426666666666667,
|
|
"grad_norm": 0.7874165773391724,
|
|
"learning_rate": 1.3055555555555556e-06,
|
|
"loss": -0.0603,
|
|
"num_tokens": 42742504.0,
|
|
"reward": 1.04296875,
|
|
"reward_std": 0.20134729146957397,
|
|
"rewards/accuracy_reward_step": 0.5703125,
|
|
"rewards/format_reward_step": 0.9453125,
|
|
"step": 154
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9609375,
|
|
"calib/avg_num_step_conf": 14.3828125,
|
|
"calib/ece": 0.054918699186991944,
|
|
"calib/final_conf_rate": 0.9609375,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.011882698768197142,
|
|
"calib/mean_conf": 0.6622357723577236,
|
|
"calib/mu_c": 0.6667763157894736,
|
|
"calib/mu_w": 0.6548936170212765,
|
|
"calib/nonempty_final_conf_rate": 0.9609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.049634146341463424,
|
|
"calib/std_conf": 0.054945291534414825,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5891995997998999,
|
|
"calib/step_q_c_n": 1999.0,
|
|
"calib/step_q_gap": -0.05192339485250663,
|
|
"calib/step_q_w": 0.6411229946524065,
|
|
"calib/step_q_w_n": 1683.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2807.0,
|
|
"completions/max_terminated_length": 2807.0,
|
|
"completions/mean_length": 771.58984375,
|
|
"completions/mean_terminated_length": 793.2810668945312,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 266.0,
|
|
"epoch": 0.16533333333333333,
|
|
"grad_norm": 0.4282694458961487,
|
|
"learning_rate": 1.2777777777777779e-06,
|
|
"loss": -0.0058,
|
|
"num_tokens": 43047247.0,
|
|
"reward": 1.07421875,
|
|
"reward_std": 0.2485809326171875,
|
|
"rewards/accuracy_reward_step": 0.59375,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 155
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 12.7265625,
|
|
"calib/ece": 0.08524556451612905,
|
|
"calib/final_conf_rate": 0.96875,
|
|
"calib/format_rate": 0.96875,
|
|
"calib/frac_conf_gt_0.9": 0.008064516129032258,
|
|
"calib/gap": -0.01273744047619041,
|
|
"calib/mean_conf": 0.6636326612903225,
|
|
"calib/mu_c": 0.6595238095238095,
|
|
"calib/mu_w": 0.6722612499999999,
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.03572943548387098,
|
|
"calib/std_conf": 0.07622542294139767,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.584796833773087,
|
|
"calib/step_q_c_n": 1895.0,
|
|
"calib/step_q_gap": -0.03450778838391966,
|
|
"calib/step_q_w": 0.6193046221570067,
|
|
"calib/step_q_w_n": 1363.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3023.0,
|
|
"completions/max_terminated_length": 3023.0,
|
|
"completions/mean_length": 821.12109375,
|
|
"completions/mean_terminated_length": 837.4780883789062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 355.0,
|
|
"epoch": 0.1664,
|
|
"grad_norm": 0.3223394453525543,
|
|
"learning_rate": 1.25e-06,
|
|
"loss": -0.0287,
|
|
"num_tokens": 43362214.0,
|
|
"reward": 1.140625,
|
|
"reward_std": 0.16987359523773193,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.96875,
|
|
"step": 156
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 12.8671875,
|
|
"calib/ece": 0.08476377952755902,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.007874015748031496,
|
|
"calib/gap": -0.012634310134310267,
|
|
"calib/mean_conf": 0.6706692913385827,
|
|
"calib/mu_c": 0.667087912087912,
|
|
"calib/mu_w": 0.6797222222222222,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.019448818897637794,
|
|
"calib/std_conf": 0.06856189285212524,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5906270022883295,
|
|
"calib/step_q_c_n": 2185.0,
|
|
"calib/step_q_gap": -0.03149202386135497,
|
|
"calib/step_q_w": 0.6221190261496845,
|
|
"calib/step_q_w_n": 1109.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2546.0,
|
|
"completions/max_terminated_length": 2546.0,
|
|
"completions/mean_length": 817.640625,
|
|
"completions/mean_terminated_length": 824.0787353515625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 252.0,
|
|
"epoch": 0.16746666666666668,
|
|
"grad_norm": 0.7573793530464172,
|
|
"learning_rate": 1.2222222222222223e-06,
|
|
"loss": -0.023,
|
|
"num_tokens": 43675258.0,
|
|
"reward": 1.20703125,
|
|
"reward_std": 0.2888256311416626,
|
|
"rewards/accuracy_reward_step": 0.7109375,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 157
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 12.45703125,
|
|
"calib/ece": 0.1071653543307087,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.007874015748031496,
|
|
"calib/gap": -0.008466954022988582,
|
|
"calib/mean_conf": 0.6615748031496064,
|
|
"calib/mu_c": 0.6589080459770115,
|
|
"calib/mu_w": 0.667375,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04185039370078742,
|
|
"calib/std_conf": 0.067938429708527,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5890804044294655,
|
|
"calib/step_q_c_n": 2077.0,
|
|
"calib/step_q_gap": -0.018104847369095678,
|
|
"calib/step_q_w": 0.6071852517985612,
|
|
"calib/step_q_w_n": 1112.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2511.0,
|
|
"completions/max_terminated_length": 2511.0,
|
|
"completions/mean_length": 786.75,
|
|
"completions/mean_terminated_length": 792.9448852539062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 301.0,
|
|
"epoch": 0.16853333333333334,
|
|
"grad_norm": 0.33515551686286926,
|
|
"learning_rate": 1.1944444444444446e-06,
|
|
"loss": -0.0143,
|
|
"num_tokens": 43981906.0,
|
|
"reward": 1.17578125,
|
|
"reward_std": 0.18715086579322815,
|
|
"rewards/accuracy_reward_step": 0.6796875,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 158
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 13.65625,
|
|
"calib/ece": 0.09509599999999999,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.004,
|
|
"calib/gap": -0.013936188315371578,
|
|
"calib/mean_conf": 0.6574159999999999,
|
|
"calib/mu_c": 0.6526219512195122,
|
|
"calib/mu_w": 0.6665581395348837,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04825599999999999,
|
|
"calib/std_conf": 0.07192125516146114,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5841757105943153,
|
|
"calib/step_q_c_n": 1935.0,
|
|
"calib/step_q_gap": -0.04489924135956047,
|
|
"calib/step_q_w": 0.6290749519538757,
|
|
"calib/step_q_w_n": 1561.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2849.0,
|
|
"completions/max_terminated_length": 2849.0,
|
|
"completions/mean_length": 776.61328125,
|
|
"completions/mean_terminated_length": 785.8221435546875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 262.0,
|
|
"epoch": 0.1696,
|
|
"grad_norm": 0.5414705872535706,
|
|
"learning_rate": 1.1666666666666668e-06,
|
|
"loss": -0.0035,
|
|
"num_tokens": 44285503.0,
|
|
"reward": 1.12890625,
|
|
"reward_std": 0.18605078756809235,
|
|
"rewards/accuracy_reward_step": 0.640625,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 159
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96875,
|
|
"calib/avg_num_step_conf": 13.859375,
|
|
"calib/ece": 0.1265587044534413,
|
|
"calib/final_conf_rate": 0.96484375,
|
|
"calib/format_rate": 0.96484375,
|
|
"calib/frac_conf_gt_0.9": 0.008097165991902834,
|
|
"calib/gap": -0.03826023391812872,
|
|
"calib/mean_conf": 0.6573279352226721,
|
|
"calib/mu_c": 0.6455555555555554,
|
|
"calib/mu_w": 0.6838157894736842,
|
|
"calib/nonempty_final_conf_rate": 0.96484375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04578947368421053,
|
|
"calib/std_conf": 0.0572103313497274,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5782809330628803,
|
|
"calib/step_q_c_n": 1972.0,
|
|
"calib/step_q_gap": -0.0440436227746831,
|
|
"calib/step_q_w": 0.6223245558375634,
|
|
"calib/step_q_w_n": 1576.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.03125,
|
|
"completions/max_length": 3011.0,
|
|
"completions/max_terminated_length": 3011.0,
|
|
"completions/mean_length": 755.3125,
|
|
"completions/mean_terminated_length": 779.6773681640625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 313.0,
|
|
"epoch": 0.17066666666666666,
|
|
"grad_norm": 0.48375532031059265,
|
|
"learning_rate": 1.138888888888889e-06,
|
|
"loss": -0.0286,
|
|
"num_tokens": 44583703.0,
|
|
"reward": 1.150390625,
|
|
"reward_std": 0.2401934415102005,
|
|
"rewards/accuracy_reward_step": 0.66796875,
|
|
"rewards/format_reward_step": 0.96484375,
|
|
"step": 160
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 13.19921875,
|
|
"calib/ece": 0.15460317460317463,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.003968253968253968,
|
|
"calib/gap": -0.010598170302603793,
|
|
"calib/mean_conf": 0.655952380952381,
|
|
"calib/mu_c": 0.6538916256157636,
|
|
"calib/mu_w": 0.6644897959183674,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0024999999999999996,
|
|
"calib/std_conf": 0.053186980317399546,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5826842105263158,
|
|
"calib/step_q_c_n": 2470.0,
|
|
"calib/step_q_gap": -0.0590759654912858,
|
|
"calib/step_q_w": 0.6417601760176016,
|
|
"calib/step_q_w_n": 909.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2037.0,
|
|
"completions/max_terminated_length": 2037.0,
|
|
"completions/mean_length": 740.15625,
|
|
"completions/mean_terminated_length": 754.900390625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 300.0,
|
|
"epoch": 0.17173333333333332,
|
|
"grad_norm": 0.218248188495636,
|
|
"learning_rate": 1.111111111111111e-06,
|
|
"loss": -0.0208,
|
|
"num_tokens": 44877103.0,
|
|
"reward": 1.28515625,
|
|
"reward_std": 0.15109506249427795,
|
|
"rewards/accuracy_reward_step": 0.79296875,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 161
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 11.36328125,
|
|
"calib/ece": 0.14329411764705888,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.00392156862745098,
|
|
"calib/gap": 0.0014897733554450898,
|
|
"calib/mean_conf": 0.6524705882352941,
|
|
"calib/mu_c": 0.6527860696517412,
|
|
"calib/mu_w": 0.6512962962962962,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.003764705882352941,
|
|
"calib/std_conf": 0.05308312125712916,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5773848238482384,
|
|
"calib/step_q_c_n": 2214.0,
|
|
"calib/step_q_gap": -0.01855762219492707,
|
|
"calib/step_q_w": 0.5959424460431655,
|
|
"calib/step_q_w_n": 695.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 3007.0,
|
|
"completions/max_terminated_length": 3007.0,
|
|
"completions/mean_length": 757.2890625,
|
|
"completions/mean_terminated_length": 760.2588500976562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 311.0,
|
|
"epoch": 0.1728,
|
|
"grad_norm": 0.23682500422000885,
|
|
"learning_rate": 1.0833333333333335e-06,
|
|
"loss": 0.0092,
|
|
"num_tokens": 45175113.0,
|
|
"reward": 1.283203125,
|
|
"reward_std": 0.15493015944957733,
|
|
"rewards/accuracy_reward_step": 0.78515625,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 162
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.94140625,
|
|
"calib/avg_num_step_conf": 16.49609375,
|
|
"calib/ece": 0.08908713692946066,
|
|
"calib/final_conf_rate": 0.94140625,
|
|
"calib/format_rate": 0.94140625,
|
|
"calib/frac_conf_gt_0.9": 0.02074688796680498,
|
|
"calib/gap": -0.029056327160493778,
|
|
"calib/mean_conf": 0.6812033195020748,
|
|
"calib/mu_c": 0.6714375,
|
|
"calib/mu_w": 0.7004938271604938,
|
|
"calib/nonempty_final_conf_rate": 0.94140625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.05319502074688799,
|
|
"calib/std_conf": 0.07385505141617801,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5897835099032704,
|
|
"calib/step_q_c_n": 2171.0,
|
|
"calib/step_q_gap": -0.07958783512596934,
|
|
"calib/step_q_w": 0.6693713450292398,
|
|
"calib/step_q_w_n": 2052.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3000.0,
|
|
"completions/max_terminated_length": 3000.0,
|
|
"completions/mean_length": 876.32421875,
|
|
"completions/mean_terminated_length": 915.6693725585938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 343.0,
|
|
"epoch": 0.17386666666666667,
|
|
"grad_norm": 0.3832489848136902,
|
|
"learning_rate": 1.0555555555555557e-06,
|
|
"loss": -0.0285,
|
|
"num_tokens": 45504284.0,
|
|
"reward": 1.095703125,
|
|
"reward_std": 0.20130212604999542,
|
|
"rewards/accuracy_reward_step": 0.625,
|
|
"rewards/format_reward_step": 0.94140625,
|
|
"step": 163
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.953125,
|
|
"calib/avg_num_step_conf": 15.13671875,
|
|
"calib/ece": 0.2001639344262296,
|
|
"calib/final_conf_rate": 0.953125,
|
|
"calib/format_rate": 0.953125,
|
|
"calib/frac_conf_gt_0.9": 0.012295081967213115,
|
|
"calib/gap": -0.04190548780487813,
|
|
"calib/mean_conf": 0.6774590163934424,
|
|
"calib/mu_c": 0.6637195121951219,
|
|
"calib/mu_w": 0.7056250000000001,
|
|
"calib/nonempty_final_conf_rate": 0.953125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.10274590163934427,
|
|
"calib/std_conf": 0.06997144002165918,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5873583093179635,
|
|
"calib/step_q_c_n": 2082.0,
|
|
"calib/step_q_gap": -0.050770692355210034,
|
|
"calib/step_q_w": 0.6381290016731735,
|
|
"calib/step_q_w_n": 1793.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 3004.0,
|
|
"completions/max_terminated_length": 3004.0,
|
|
"completions/mean_length": 882.9140625,
|
|
"completions/mean_terminated_length": 915.0850219726562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 387.0,
|
|
"epoch": 0.17493333333333333,
|
|
"grad_norm": 0.20860423147678375,
|
|
"learning_rate": 1.0277777777777777e-06,
|
|
"loss": -0.07,
|
|
"num_tokens": 45836446.0,
|
|
"reward": 1.1171875,
|
|
"reward_std": 0.23892061412334442,
|
|
"rewards/accuracy_reward_step": 0.640625,
|
|
"rewards/format_reward_step": 0.953125,
|
|
"step": 164
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 14.03515625,
|
|
"calib/ece": 0.13533596837944664,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.015810276679841896,
|
|
"calib/gap": -0.02319058495114834,
|
|
"calib/mean_conf": 0.668695652173913,
|
|
"calib/mu_c": 0.6585211267605633,
|
|
"calib/mu_w": 0.6817117117117116,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.12138339920948618,
|
|
"calib/std_conf": 0.06808125711257992,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5844966063348417,
|
|
"calib/step_q_c_n": 1768.0,
|
|
"calib/step_q_gap": -0.03118010599392551,
|
|
"calib/step_q_w": 0.6156767123287672,
|
|
"calib/step_q_w_n": 1825.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2477.0,
|
|
"completions/max_terminated_length": 2477.0,
|
|
"completions/mean_length": 856.7890625,
|
|
"completions/mean_terminated_length": 870.388916015625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 330.0,
|
|
"epoch": 0.176,
|
|
"grad_norm": 0.20547796785831451,
|
|
"learning_rate": 1.0000000000000002e-06,
|
|
"loss": -0.0252,
|
|
"num_tokens": 46161360.0,
|
|
"reward": 1.048828125,
|
|
"reward_std": 0.19807085394859314,
|
|
"rewards/accuracy_reward_step": 0.5546875,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 165
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.95703125,
|
|
"calib/avg_num_step_conf": 14.03515625,
|
|
"calib/ece": 0.17277551020408166,
|
|
"calib/final_conf_rate": 0.95703125,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.0163265306122449,
|
|
"calib/gap": -0.0472100122100122,
|
|
"calib/mean_conf": 0.6703265306122449,
|
|
"calib/mu_c": 0.6581868131868132,
|
|
"calib/mu_w": 0.7053968253968254,
|
|
"calib/nonempty_final_conf_rate": 0.95703125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.050122448979591866,
|
|
"calib/std_conf": 0.0650078135822061,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5863866666666667,
|
|
"calib/step_q_c_n": 2250.0,
|
|
"calib/step_q_gap": -0.05549717547778599,
|
|
"calib/step_q_w": 0.6418838421444527,
|
|
"calib/step_q_w_n": 1343.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2158.0,
|
|
"completions/max_terminated_length": 2158.0,
|
|
"completions/mean_length": 778.828125,
|
|
"completions/mean_terminated_length": 817.131103515625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 307.0,
|
|
"epoch": 0.17706666666666668,
|
|
"grad_norm": 0.2849808633327484,
|
|
"learning_rate": 9.722222222222224e-07,
|
|
"loss": -0.0281,
|
|
"num_tokens": 46466924.0,
|
|
"reward": 1.189453125,
|
|
"reward_std": 0.17397813498973846,
|
|
"rewards/accuracy_reward_step": 0.7109375,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 166
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 13.97265625,
|
|
"calib/ece": 0.2165354330708662,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.015748031496062992,
|
|
"calib/gap": -0.024596296296296782,
|
|
"calib/mean_conf": 0.6769291338582677,
|
|
"calib/mu_c": 0.6717,
|
|
"calib/mu_w": 0.6962962962962967,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.05303149606299212,
|
|
"calib/std_conf": 0.07389087304750322,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5909087550794238,
|
|
"calib/step_q_c_n": 2707.0,
|
|
"calib/step_q_gap": -0.04283837135735791,
|
|
"calib/step_q_w": 0.6337471264367817,
|
|
"calib/step_q_w_n": 870.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2335.0,
|
|
"completions/max_terminated_length": 2335.0,
|
|
"completions/mean_length": 841.6796875,
|
|
"completions/mean_terminated_length": 851.6600952148438,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 316.0,
|
|
"epoch": 0.17813333333333334,
|
|
"grad_norm": 0.2347254455089569,
|
|
"learning_rate": 9.444444444444445e-07,
|
|
"loss": -0.007,
|
|
"num_tokens": 46788002.0,
|
|
"reward": 1.27734375,
|
|
"reward_std": 0.09994982928037643,
|
|
"rewards/accuracy_reward_step": 0.78125,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 167
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.94140625,
|
|
"calib/avg_num_step_conf": 16.671875,
|
|
"calib/ece": 0.14452282157676355,
|
|
"calib/final_conf_rate": 0.94140625,
|
|
"calib/format_rate": 0.94140625,
|
|
"calib/frac_conf_gt_0.9": 0.008298755186721992,
|
|
"calib/gap": -0.04282077922077909,
|
|
"calib/mean_conf": 0.6852697095435684,
|
|
"calib/mu_c": 0.6735428571428572,
|
|
"calib/mu_w": 0.7163636363636363,
|
|
"calib/nonempty_final_conf_rate": 0.94140625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.05182572614107889,
|
|
"calib/std_conf": 0.07231809254864334,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5977239413680782,
|
|
"calib/step_q_c_n": 2456.0,
|
|
"calib/step_q_gap": -0.05827666569593948,
|
|
"calib/step_q_w": 0.6560006070640176,
|
|
"calib/step_q_w_n": 1812.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2822.0,
|
|
"completions/max_terminated_length": 2822.0,
|
|
"completions/mean_length": 873.77734375,
|
|
"completions/mean_terminated_length": 924.326416015625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 328.0,
|
|
"epoch": 0.1792,
|
|
"grad_norm": 0.23246829211711884,
|
|
"learning_rate": 9.166666666666666e-07,
|
|
"loss": -0.0611,
|
|
"num_tokens": 47116361.0,
|
|
"reward": 1.154296875,
|
|
"reward_std": 0.25926464796066284,
|
|
"rewards/accuracy_reward_step": 0.68359375,
|
|
"rewards/format_reward_step": 0.94140625,
|
|
"step": 168
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 12.6171875,
|
|
"calib/ece": 0.07893700787401581,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.005826411960132782,
|
|
"calib/mean_conf": 0.6689370078740159,
|
|
"calib/mu_c": 0.6669642857142858,
|
|
"calib/mu_w": 0.6727906976744186,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04322834645669292,
|
|
"calib/std_conf": 0.05315306200489301,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5849169921874999,
|
|
"calib/step_q_c_n": 2048.0,
|
|
"calib/step_q_gap": -0.007663380062923109,
|
|
"calib/step_q_w": 0.592580372250423,
|
|
"calib/step_q_w_n": 1182.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2639.0,
|
|
"completions/max_terminated_length": 2639.0,
|
|
"completions/mean_length": 846.12109375,
|
|
"completions/mean_terminated_length": 852.783447265625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 316.0,
|
|
"epoch": 0.18026666666666666,
|
|
"grad_norm": 0.30217358469963074,
|
|
"learning_rate": 8.88888888888889e-07,
|
|
"loss": 0.0012,
|
|
"num_tokens": 47437152.0,
|
|
"reward": 1.15234375,
|
|
"reward_std": 0.17056959867477417,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 169
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 13.07421875,
|
|
"calib/ece": 0.0821428571428572,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.007936507936507936,
|
|
"calib/gap": -0.0013406593406593004,
|
|
"calib/mean_conf": 0.6703174603174604,
|
|
"calib/mu_c": 0.669945054945055,
|
|
"calib/mu_w": 0.6712857142857143,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.015119047619047627,
|
|
"calib/std_conf": 0.05874259842516748,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5910180412371134,
|
|
"calib/step_q_c_n": 2328.0,
|
|
"calib/step_q_gap": -0.013663018625496903,
|
|
"calib/step_q_w": 0.6046810598626103,
|
|
"calib/step_q_w_n": 1019.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2513.0,
|
|
"completions/max_terminated_length": 2513.0,
|
|
"completions/mean_length": 826.29296875,
|
|
"completions/mean_terminated_length": 832.7991943359375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 355.0,
|
|
"epoch": 0.18133333333333335,
|
|
"grad_norm": 0.25517553091049194,
|
|
"learning_rate": 8.611111111111112e-07,
|
|
"loss": -0.0104,
|
|
"num_tokens": 47752835.0,
|
|
"reward": 1.20703125,
|
|
"reward_std": 0.18111488223075867,
|
|
"rewards/accuracy_reward_step": 0.71484375,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 170
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 13.5859375,
|
|
"calib/ece": 0.06722222222222217,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.010637990951412823,
|
|
"calib/mean_conf": 0.6609523809523808,
|
|
"calib/mu_c": 0.6566887417218544,
|
|
"calib/mu_w": 0.6673267326732673,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06448412698412694,
|
|
"calib/std_conf": 0.053257812176460355,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5830900372935536,
|
|
"calib/step_q_c_n": 1877.0,
|
|
"calib/step_q_gap": -0.013262242531555724,
|
|
"calib/step_q_w": 0.5963522798251093,
|
|
"calib/step_q_w_n": 1601.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1948.0,
|
|
"completions/max_terminated_length": 1948.0,
|
|
"completions/mean_length": 794.51171875,
|
|
"completions/mean_terminated_length": 807.123046875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 183.0,
|
|
"epoch": 0.1824,
|
|
"grad_norm": 0.20195415616035461,
|
|
"learning_rate": 8.333333333333333e-07,
|
|
"loss": -0.0151,
|
|
"num_tokens": 48063126.0,
|
|
"reward": 1.08203125,
|
|
"reward_std": 0.14921081066131592,
|
|
"rewards/accuracy_reward_step": 0.58984375,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 171
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 12.57421875,
|
|
"calib/ece": 0.1586799999999999,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.02331443688586554,
|
|
"calib/mean_conf": 0.65524,
|
|
"calib/mu_c": 0.6502040816326531,
|
|
"calib/mu_w": 0.6735185185185186,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.014960000000000001,
|
|
"calib/std_conf": 0.05578657903116126,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5797955555555555,
|
|
"calib/step_q_c_n": 2250.0,
|
|
"calib/step_q_gap": -0.038842215342277386,
|
|
"calib/step_q_w": 0.6186377708978329,
|
|
"calib/step_q_w_n": 969.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2883.0,
|
|
"completions/max_terminated_length": 2883.0,
|
|
"completions/mean_length": 767.859375,
|
|
"completions/mean_terminated_length": 780.0476684570312,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 361.0,
|
|
"epoch": 0.18346666666666667,
|
|
"grad_norm": 0.30305296182632446,
|
|
"learning_rate": 8.055555555555557e-07,
|
|
"loss": -0.0186,
|
|
"num_tokens": 48363050.0,
|
|
"reward": 1.25390625,
|
|
"reward_std": 0.1857965886592865,
|
|
"rewards/accuracy_reward_step": 0.765625,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 172
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 13.6953125,
|
|
"calib/ece": 0.07822134387351778,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.03557312252964427,
|
|
"calib/gap": 0.011060606060606215,
|
|
"calib/mean_conf": 0.6734782608695652,
|
|
"calib/mu_c": 0.6763636363636365,
|
|
"calib/mu_w": 0.6653030303030303,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0062845849802371494,
|
|
"calib/std_conf": 0.08751598680459782,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.602466456195738,
|
|
"calib/step_q_c_n": 2534.0,
|
|
"calib/step_q_gap": -0.00825370841331552,
|
|
"calib/step_q_w": 0.6107201646090535,
|
|
"calib/step_q_w_n": 972.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2514.0,
|
|
"completions/max_terminated_length": 2514.0,
|
|
"completions/mean_length": 839.54296875,
|
|
"completions/mean_terminated_length": 849.498046875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 276.0,
|
|
"epoch": 0.18453333333333333,
|
|
"grad_norm": 0.3388940393924713,
|
|
"learning_rate": 7.777777777777779e-07,
|
|
"loss": -0.0148,
|
|
"num_tokens": 48681133.0,
|
|
"reward": 1.22265625,
|
|
"reward_std": 0.17308899760246277,
|
|
"rewards/accuracy_reward_step": 0.73046875,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 173
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9296875,
|
|
"calib/avg_num_step_conf": 15.359375,
|
|
"calib/ece": 0.06844537815126057,
|
|
"calib/final_conf_rate": 0.9296875,
|
|
"calib/format_rate": 0.92578125,
|
|
"calib/frac_conf_gt_0.9": 0.008403361344537815,
|
|
"calib/gap": 0.007239896180941785,
|
|
"calib/mean_conf": 0.6717226890756303,
|
|
"calib/mu_c": 0.674551724137931,
|
|
"calib/mu_w": 0.6673118279569892,
|
|
"calib/nonempty_final_conf_rate": 0.9296875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
|
"calib/pce": 0.06546218487394966,
|
|
"calib/std_conf": 0.05931661583213658,
|
|
"calib/step_conf_rate": 0.99609375,
|
|
"calib/step_q_c": 0.5958893657606105,
|
|
"calib/step_q_c_n": 2097.0,
|
|
"calib/step_q_gap": -0.051729162849743826,
|
|
"calib/step_q_w": 0.6476185286103543,
|
|
"calib/step_q_w_n": 1835.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.05859375,
|
|
"completions/max_length": 2882.0,
|
|
"completions/max_terminated_length": 2882.0,
|
|
"completions/mean_length": 860.96875,
|
|
"completions/mean_terminated_length": 914.5560913085938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 369.0,
|
|
"epoch": 0.1856,
|
|
"grad_norm": 0.2765766978263855,
|
|
"learning_rate": 7.5e-07,
|
|
"loss": -0.0608,
|
|
"num_tokens": 49005773.0,
|
|
"reward": 1.029296875,
|
|
"reward_std": 0.3664853870868683,
|
|
"rewards/accuracy_reward_step": 0.56640625,
|
|
"rewards/format_reward_step": 0.92578125,
|
|
"step": 174
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9609375,
|
|
"calib/avg_num_step_conf": 14.53125,
|
|
"calib/ece": 0.1834146341463415,
|
|
"calib/final_conf_rate": 0.9609375,
|
|
"calib/format_rate": 0.9609375,
|
|
"calib/frac_conf_gt_0.9": 0.012195121951219513,
|
|
"calib/gap": -0.02279263771186446,
|
|
"calib/mean_conf": 0.6778861788617886,
|
|
"calib/mu_c": 0.6669531249999999,
|
|
"calib/mu_w": 0.6897457627118644,
|
|
"calib/nonempty_final_conf_rate": 0.9609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.1704878048780488,
|
|
"calib/std_conf": 0.07265300124195796,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5895738295318127,
|
|
"calib/step_q_c_n": 1666.0,
|
|
"calib/step_q_gap": -0.03702792314588932,
|
|
"calib/step_q_w": 0.626601752677702,
|
|
"calib/step_q_w_n": 2054.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2944.0,
|
|
"completions/max_terminated_length": 2944.0,
|
|
"completions/mean_length": 846.40625,
|
|
"completions/mean_terminated_length": 877.2470092773438,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 232.0,
|
|
"epoch": 0.18666666666666668,
|
|
"grad_norm": 0.24171149730682373,
|
|
"learning_rate": 7.222222222222222e-07,
|
|
"loss": -0.062,
|
|
"num_tokens": 49328277.0,
|
|
"reward": 0.98046875,
|
|
"reward_std": 0.25995826721191406,
|
|
"rewards/accuracy_reward_step": 0.5,
|
|
"rewards/format_reward_step": 0.9609375,
|
|
"step": 175
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9765625,
|
|
"calib/avg_num_step_conf": 15.5625,
|
|
"calib/ece": 0.08387999999999995,
|
|
"calib/final_conf_rate": 0.9765625,
|
|
"calib/format_rate": 0.9765625,
|
|
"calib/frac_conf_gt_0.9": 0.008,
|
|
"calib/gap": -0.016378672621955692,
|
|
"calib/mean_conf": 0.6831600000000001,
|
|
"calib/mu_c": 0.6773291925465837,
|
|
"calib/mu_w": 0.6937078651685394,
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.06152000000000002,
|
|
"calib/std_conf": 0.07869951969357881,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6036913043478261,
|
|
"calib/step_q_c_n": 2300.0,
|
|
"calib/step_q_gap": -0.02549189042652067,
|
|
"calib/step_q_w": 0.6291831947743468,
|
|
"calib/step_q_w_n": 1684.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2249.0,
|
|
"completions/max_terminated_length": 2249.0,
|
|
"completions/mean_length": 873.48046875,
|
|
"completions/mean_terminated_length": 898.0361328125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 270.0,
|
|
"epoch": 0.18773333333333334,
|
|
"grad_norm": 0.20827040076255798,
|
|
"learning_rate": 6.944444444444446e-07,
|
|
"loss": -0.0165,
|
|
"num_tokens": 49655952.0,
|
|
"reward": 1.1171875,
|
|
"reward_std": 0.2059280127286911,
|
|
"rewards/accuracy_reward_step": 0.62890625,
|
|
"rewards/format_reward_step": 0.9765625,
|
|
"step": 176
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 13.60546875,
|
|
"calib/ece": 0.15513944223107565,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.00398406374501992,
|
|
"calib/gap": -0.03370410212277686,
|
|
"calib/mean_conf": 0.6635856573705179,
|
|
"calib/mu_c": 0.6524404761904763,
|
|
"calib/mu_w": 0.6861445783132532,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0747011952191235,
|
|
"calib/std_conf": 0.059284341111813736,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.583358208955224,
|
|
"calib/step_q_c_n": 2010.0,
|
|
"calib/step_q_gap": -0.034155097222644315,
|
|
"calib/step_q_w": 0.6175133061778683,
|
|
"calib/step_q_w_n": 1473.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2894.0,
|
|
"completions/max_terminated_length": 2894.0,
|
|
"completions/mean_length": 824.33984375,
|
|
"completions/mean_terminated_length": 834.1146850585938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 336.0,
|
|
"epoch": 0.1888,
|
|
"grad_norm": 0.21935386955738068,
|
|
"learning_rate": 6.666666666666667e-07,
|
|
"loss": -0.0015,
|
|
"num_tokens": 49970815.0,
|
|
"reward": 1.146484375,
|
|
"reward_std": 0.12562449276447296,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 177
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 13.89453125,
|
|
"calib/ece": 0.060912698412698404,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.003968253968253968,
|
|
"calib/gap": -0.003250896319601848,
|
|
"calib/mean_conf": 0.6672619047619048,
|
|
"calib/mu_c": 0.6662427745664741,
|
|
"calib/mu_w": 0.6694936708860759,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.020833333333333322,
|
|
"calib/std_conf": 0.05504726375729799,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5884535186794092,
|
|
"calib/step_q_c_n": 2302.0,
|
|
"calib/step_q_gap": -0.011267596858439366,
|
|
"calib/step_q_w": 0.5997211155378486,
|
|
"calib/step_q_w_n": 1255.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 1764.0,
|
|
"completions/max_terminated_length": 1764.0,
|
|
"completions/mean_length": 773.0234375,
|
|
"completions/mean_terminated_length": 785.293701171875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 291.0,
|
|
"epoch": 0.18986666666666666,
|
|
"grad_norm": 0.2755891978740692,
|
|
"learning_rate": 6.388888888888889e-07,
|
|
"loss": -0.0395,
|
|
"num_tokens": 50274781.0,
|
|
"reward": 1.16796875,
|
|
"reward_std": 0.20019932091236115,
|
|
"rewards/accuracy_reward_step": 0.67578125,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 178
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 13.35546875,
|
|
"calib/ece": 0.11400793650793654,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.011904761904761904,
|
|
"calib/gap": -0.0018976911853059164,
|
|
"calib/mean_conf": 0.6651190476190476,
|
|
"calib/mu_c": 0.6646596858638745,
|
|
"calib/mu_w": 0.6665573770491804,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.010595238095238095,
|
|
"calib/std_conf": 0.060411828330779516,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5900040832993059,
|
|
"calib/step_q_c_n": 2449.0,
|
|
"calib/step_q_gap": -0.007501071339869325,
|
|
"calib/step_q_w": 0.5975051546391752,
|
|
"calib/step_q_w_n": 970.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2469.0,
|
|
"completions/max_terminated_length": 2469.0,
|
|
"completions/mean_length": 827.62890625,
|
|
"completions/mean_terminated_length": 840.7659301757812,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 276.0,
|
|
"epoch": 0.19093333333333334,
|
|
"grad_norm": 0.2493710219860077,
|
|
"learning_rate": 6.111111111111112e-07,
|
|
"loss": -0.0272,
|
|
"num_tokens": 50592918.0,
|
|
"reward": 1.23828125,
|
|
"reward_std": 0.16652068495750427,
|
|
"rewards/accuracy_reward_step": 0.74609375,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 179
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96484375,
|
|
"calib/avg_num_step_conf": 17.390625,
|
|
"calib/ece": 0.043724696356275315,
|
|
"calib/final_conf_rate": 0.96484375,
|
|
"calib/format_rate": 0.96484375,
|
|
"calib/frac_conf_gt_0.9": 0.020242914979757085,
|
|
"calib/gap": -0.0009728308501313343,
|
|
"calib/mean_conf": 0.7036437246963563,
|
|
"calib/mu_c": 0.703312883435583,
|
|
"calib/mu_w": 0.7042857142857143,
|
|
"calib/nonempty_final_conf_rate": 0.96484375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.043724696356275315,
|
|
"calib/std_conf": 0.07650568626887001,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6103323262839879,
|
|
"calib/step_q_c_n": 2648.0,
|
|
"calib/step_q_gap": -0.026618893228207208,
|
|
"calib/step_q_w": 0.6369512195121951,
|
|
"calib/step_q_w_n": 1804.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2547.0,
|
|
"completions/max_terminated_length": 2547.0,
|
|
"completions/mean_length": 965.7421875,
|
|
"completions/mean_terminated_length": 1004.9999389648438,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 390.0,
|
|
"epoch": 0.192,
|
|
"grad_norm": 0.2207210808992386,
|
|
"learning_rate": 5.833333333333334e-07,
|
|
"loss": -0.0376,
|
|
"num_tokens": 50944004.0,
|
|
"reward": 1.119140625,
|
|
"reward_std": 0.2057252824306488,
|
|
"rewards/accuracy_reward_step": 0.63671875,
|
|
"rewards/format_reward_step": 0.96484375,
|
|
"step": 180
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 13.69140625,
|
|
"calib/ece": 0.05204724409448826,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.003937007874015748,
|
|
"calib/gap": -0.007761627906976765,
|
|
"calib/mean_conf": 0.6663779527559055,
|
|
"calib/mu_c": 0.66375,
|
|
"calib/mu_w": 0.6715116279069767,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.028503937007874056,
|
|
"calib/std_conf": 0.05793247890729539,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5896074933095451,
|
|
"calib/step_q_c_n": 2242.0,
|
|
"calib/step_q_gap": -0.008777304790217322,
|
|
"calib/step_q_w": 0.5983847980997624,
|
|
"calib/step_q_w_n": 1263.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.00390625,
|
|
"completions/max_length": 2532.0,
|
|
"completions/max_terminated_length": 2532.0,
|
|
"completions/mean_length": 820.90234375,
|
|
"completions/mean_terminated_length": 824.1216430664062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 363.0,
|
|
"epoch": 0.19306666666666666,
|
|
"grad_norm": 0.27567002177238464,
|
|
"learning_rate": 5.555555555555555e-07,
|
|
"loss": 0.0064,
|
|
"num_tokens": 51260419.0,
|
|
"reward": 1.15234375,
|
|
"reward_std": 0.2041134089231491,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 181
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9921875,
|
|
"calib/avg_num_step_conf": 14.12890625,
|
|
"calib/ece": 0.1293700787401575,
|
|
"calib/final_conf_rate": 0.9921875,
|
|
"calib/format_rate": 0.9921875,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.036599099099099086,
|
|
"calib/mean_conf": 0.6714960629921259,
|
|
"calib/mu_c": 0.6608333333333333,
|
|
"calib/mu_w": 0.6974324324324324,
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.046102362204724384,
|
|
"calib/std_conf": 0.05996821651820224,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5833461210571185,
|
|
"calib/step_q_c_n": 2346.0,
|
|
"calib/step_q_gap": -0.03431713622061561,
|
|
"calib/step_q_w": 0.6176632572777341,
|
|
"calib/step_q_w_n": 1271.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2130.0,
|
|
"completions/max_terminated_length": 2130.0,
|
|
"completions/mean_length": 847.31640625,
|
|
"completions/mean_terminated_length": 857.3636474609375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 371.0,
|
|
"epoch": 0.19413333333333332,
|
|
"grad_norm": 0.2295883148908615,
|
|
"learning_rate": 5.277777777777779e-07,
|
|
"loss": -0.0192,
|
|
"num_tokens": 51583492.0,
|
|
"reward": 1.19921875,
|
|
"reward_std": 0.19453556835651398,
|
|
"rewards/accuracy_reward_step": 0.703125,
|
|
"rewards/format_reward_step": 0.9921875,
|
|
"step": 182
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 15.28515625,
|
|
"calib/ece": 0.10800796812748993,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.027888446215139442,
|
|
"calib/gap": -0.012878316288720715,
|
|
"calib/mean_conf": 0.6839442231075696,
|
|
"calib/mu_c": 0.6799421965317919,
|
|
"calib/mu_w": 0.6928205128205126,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.05135458167330675,
|
|
"calib/std_conf": 0.07437375246865623,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.6010756972111554,
|
|
"calib/step_q_c_n": 2510.0,
|
|
"calib/step_q_gap": -0.03647954156290023,
|
|
"calib/step_q_w": 0.6375552387740556,
|
|
"calib/step_q_w_n": 1403.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2978.0,
|
|
"completions/max_terminated_length": 2978.0,
|
|
"completions/mean_length": 914.36328125,
|
|
"completions/mean_terminated_length": 928.8770141601562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 361.0,
|
|
"epoch": 0.1952,
|
|
"grad_norm": 0.2984082102775574,
|
|
"learning_rate": 5.000000000000001e-07,
|
|
"loss": -0.0115,
|
|
"num_tokens": 51924249.0,
|
|
"reward": 1.166015625,
|
|
"reward_std": 0.258495569229126,
|
|
"rewards/accuracy_reward_step": 0.67578125,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 183
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.984375,
|
|
"calib/avg_num_step_conf": 14.609375,
|
|
"calib/ece": 0.07559523809523806,
|
|
"calib/final_conf_rate": 0.984375,
|
|
"calib/format_rate": 0.984375,
|
|
"calib/frac_conf_gt_0.9": 0.01984126984126984,
|
|
"calib/gap": -0.014779980102548329,
|
|
"calib/mean_conf": 0.6736111111111112,
|
|
"calib/mu_c": 0.6693296089385475,
|
|
"calib/mu_w": 0.6841095890410959,
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.019444444444444445,
|
|
"calib/std_conf": 0.07324707830158014,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5995702347791484,
|
|
"calib/step_q_c_n": 2513.0,
|
|
"calib/step_q_gap": -0.01725128111327212,
|
|
"calib/step_q_w": 0.6168215158924205,
|
|
"calib/step_q_w_n": 1227.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2498.0,
|
|
"completions/max_terminated_length": 2498.0,
|
|
"completions/mean_length": 845.875,
|
|
"completions/mean_terminated_length": 862.72509765625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 292.0,
|
|
"epoch": 0.19626666666666667,
|
|
"grad_norm": 0.21070480346679688,
|
|
"learning_rate": 4.7222222222222226e-07,
|
|
"loss": -0.0355,
|
|
"num_tokens": 52246073.0,
|
|
"reward": 1.19140625,
|
|
"reward_std": 0.193123459815979,
|
|
"rewards/accuracy_reward_step": 0.69921875,
|
|
"rewards/format_reward_step": 0.984375,
|
|
"step": 184
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.95703125,
|
|
"calib/avg_num_step_conf": 14.80859375,
|
|
"calib/ece": 0.17412244897959178,
|
|
"calib/final_conf_rate": 0.95703125,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.03939504827156637,
|
|
"calib/mean_conf": 0.6626938775510205,
|
|
"calib/mu_c": 0.6504733727810651,
|
|
"calib/mu_w": 0.6898684210526315,
|
|
"calib/nonempty_final_conf_rate": 0.95703125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.07351020408163263,
|
|
"calib/std_conf": 0.056605402526337355,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5756875973015049,
|
|
"calib/step_q_c_n": 1927.0,
|
|
"calib/step_q_gap": -0.06673847565986846,
|
|
"calib/step_q_w": 0.6424260729613733,
|
|
"calib/step_q_w_n": 1864.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2965.0,
|
|
"completions/max_terminated_length": 2965.0,
|
|
"completions/mean_length": 818.75,
|
|
"completions/mean_terminated_length": 848.5830078125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 380.0,
|
|
"epoch": 0.19733333333333333,
|
|
"grad_norm": 0.1729198396205902,
|
|
"learning_rate": 4.444444444444445e-07,
|
|
"loss": -0.0276,
|
|
"num_tokens": 52562593.0,
|
|
"reward": 1.138671875,
|
|
"reward_std": 0.12429915368556976,
|
|
"rewards/accuracy_reward_step": 0.66015625,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 185
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 15.1953125,
|
|
"calib/ece": 0.1281422924901186,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.023715415019762844,
|
|
"calib/gap": -0.04608823529411754,
|
|
"calib/mean_conf": 0.6743873517786562,
|
|
"calib/mu_c": 0.662,
|
|
"calib/mu_w": 0.7080882352941176,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.035652173913043476,
|
|
"calib/std_conf": 0.06845067002867303,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5861823924200552,
|
|
"calib/step_q_c_n": 2533.0,
|
|
"calib/step_q_gap": -0.04835703278259762,
|
|
"calib/step_q_w": 0.6345394252026528,
|
|
"calib/step_q_w_n": 1357.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 894.87890625,
|
|
"completions/mean_terminated_length": 901.9251708984375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 259.0,
|
|
"epoch": 0.1984,
|
|
"grad_norm": 0.2917207181453705,
|
|
"learning_rate": 4.1666666666666667e-07,
|
|
"loss": -0.0176,
|
|
"num_tokens": 52896722.0,
|
|
"reward": 1.216796875,
|
|
"reward_std": 0.1938859224319458,
|
|
"rewards/accuracy_reward_step": 0.72265625,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 186
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.9453125,
|
|
"calib/avg_num_step_conf": 16.36328125,
|
|
"calib/ece": 0.09082644628099174,
|
|
"calib/final_conf_rate": 0.9453125,
|
|
"calib/format_rate": 0.9453125,
|
|
"calib/frac_conf_gt_0.9": 0.008264462809917356,
|
|
"calib/gap": -0.01920346320346311,
|
|
"calib/mean_conf": 0.6867768595041321,
|
|
"calib/mu_c": 0.6806666666666666,
|
|
"calib/mu_w": 0.6998701298701298,
|
|
"calib/nonempty_final_conf_rate": 0.9453125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.04789256198347107,
|
|
"calib/std_conf": 0.0702264916229942,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5987443392342529,
|
|
"calib/step_q_c_n": 2429.0,
|
|
"calib/step_q_gap": -0.05200003576574719,
|
|
"calib/step_q_w": 0.650744375,
|
|
"calib/step_q_w_n": 1760.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2662.0,
|
|
"completions/max_terminated_length": 2662.0,
|
|
"completions/mean_length": 917.1640625,
|
|
"completions/mean_terminated_length": 958.3428344726562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 338.0,
|
|
"epoch": 0.19946666666666665,
|
|
"grad_norm": 0.2757011353969574,
|
|
"learning_rate": 3.8888888888888895e-07,
|
|
"loss": -0.0491,
|
|
"num_tokens": 53233060.0,
|
|
"reward": 1.1171875,
|
|
"reward_std": 0.3008151352405548,
|
|
"rewards/accuracy_reward_step": 0.64453125,
|
|
"rewards/format_reward_step": 0.9453125,
|
|
"step": 187
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 15.234375,
|
|
"calib/ece": 0.08810355731225292,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.015810276679841896,
|
|
"calib/gap": -0.00777241653418137,
|
|
"calib/mean_conf": 0.6781430830039525,
|
|
"calib/mu_c": 0.676054054054054,
|
|
"calib/mu_w": 0.6838264705882354,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.017510671936758898,
|
|
"calib/std_conf": 0.0817776717111406,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5998210957882968,
|
|
"calib/step_q_c_n": 2683.0,
|
|
"calib/step_q_gap": -0.014320317523124682,
|
|
"calib/step_q_w": 0.6141414133114215,
|
|
"calib/step_q_w_n": 1217.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2373.0,
|
|
"completions/max_terminated_length": 2373.0,
|
|
"completions/mean_length": 909.578125,
|
|
"completions/mean_terminated_length": 920.3636474609375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 341.0,
|
|
"epoch": 0.20053333333333334,
|
|
"grad_norm": 0.2577819526195526,
|
|
"learning_rate": 3.611111111111111e-07,
|
|
"loss": 0.0013,
|
|
"num_tokens": 53569984.0,
|
|
"reward": 1.216796875,
|
|
"reward_std": 0.12154898792505264,
|
|
"rewards/accuracy_reward_step": 0.72265625,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 188
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 13.40625,
|
|
"calib/ece": 0.10988142292490109,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.02540476190476182,
|
|
"calib/mean_conf": 0.6591304347826086,
|
|
"calib/mu_c": 0.6505952380952381,
|
|
"calib/mu_w": 0.6759999999999999,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.052490118577075084,
|
|
"calib/std_conf": 0.05169875862043467,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5799900199600799,
|
|
"calib/step_q_c_n": 2004.0,
|
|
"calib/step_q_gap": -0.02961082037605456,
|
|
"calib/step_q_w": 0.6096008403361345,
|
|
"calib/step_q_w_n": 1428.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2418.0,
|
|
"completions/max_terminated_length": 2418.0,
|
|
"completions/mean_length": 806.65234375,
|
|
"completions/mean_terminated_length": 816.2174072265625,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 376.0,
|
|
"epoch": 0.2016,
|
|
"grad_norm": 0.2876899838447571,
|
|
"learning_rate": 3.3333333333333335e-07,
|
|
"loss": -0.0313,
|
|
"num_tokens": 53884255.0,
|
|
"reward": 1.150390625,
|
|
"reward_std": 0.2516007423400879,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 189
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 13.890625,
|
|
"calib/ece": 0.0443873517786561,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.007500000000000062,
|
|
"calib/mean_conf": 0.6751383399209486,
|
|
"calib/mu_c": 0.6725,
|
|
"calib/mu_w": 0.68,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.03565217391304347,
|
|
"calib/std_conf": 0.05668441721105008,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5904176334106728,
|
|
"calib/step_q_c_n": 2155.0,
|
|
"calib/step_q_gap": -0.017084151029013173,
|
|
"calib/step_q_w": 0.6075017844396859,
|
|
"calib/step_q_w_n": 1401.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2056.0,
|
|
"completions/max_terminated_length": 2056.0,
|
|
"completions/mean_length": 877.1640625,
|
|
"completions/mean_terminated_length": 891.0873413085938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 266.0,
|
|
"epoch": 0.20266666666666666,
|
|
"grad_norm": 0.20597021281719208,
|
|
"learning_rate": 3.055555555555556e-07,
|
|
"loss": -0.0274,
|
|
"num_tokens": 54214417.0,
|
|
"reward": 1.134765625,
|
|
"reward_std": 0.16795945167541504,
|
|
"rewards/accuracy_reward_step": 0.640625,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 190
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96484375,
|
|
"calib/avg_num_step_conf": 14.15625,
|
|
"calib/ece": 0.02862348178137656,
|
|
"calib/final_conf_rate": 0.96484375,
|
|
"calib/format_rate": 0.96484375,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": 0.001968152866241968,
|
|
"calib/mean_conf": 0.664251012145749,
|
|
"calib/mu_c": 0.664968152866242,
|
|
"calib/mu_w": 0.663,
|
|
"calib/nonempty_final_conf_rate": 0.96484375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.02862348178137656,
|
|
"calib/std_conf": 0.05040871162286916,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5865340364333653,
|
|
"calib/step_q_c_n": 2086.0,
|
|
"calib/step_q_gap": -0.024348798417089812,
|
|
"calib/step_q_w": 0.6108828348504551,
|
|
"calib/step_q_w_n": 1538.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 0.0,
|
|
"clip_ratio/low_mean": 0.0,
|
|
"clip_ratio/low_min": 0.0,
|
|
"clip_ratio/region_mean": 0.0,
|
|
"completions/clipped_ratio": 0.03125,
|
|
"completions/max_length": 2986.0,
|
|
"completions/max_terminated_length": 2986.0,
|
|
"completions/mean_length": 822.21875,
|
|
"completions/mean_terminated_length": 848.7418823242188,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 276.0,
|
|
"epoch": 0.20373333333333332,
|
|
"grad_norm": 0.22264082729816437,
|
|
"learning_rate": 2.7777777777777776e-07,
|
|
"loss": -0.051,
|
|
"num_tokens": 54529073.0,
|
|
"reward": 1.095703125,
|
|
"reward_std": 0.16742515563964844,
|
|
"rewards/accuracy_reward_step": 0.61328125,
|
|
"rewards/format_reward_step": 0.96484375,
|
|
"step": 191
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 14.57421875,
|
|
"calib/ece": 0.12988047808764938,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.01593625498007968,
|
|
"calib/gap": -0.032950000000000035,
|
|
"calib/mean_conf": 0.6760956175298806,
|
|
"calib/mu_c": 0.66625,
|
|
"calib/mu_w": 0.6992,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.05239043824701198,
|
|
"calib/std_conf": 0.0739056603663103,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5940485312899106,
|
|
"calib/step_q_c_n": 2349.0,
|
|
"calib/step_q_gap": -0.03642180156103003,
|
|
"calib/step_q_w": 0.6304703328509407,
|
|
"calib/step_q_w_n": 1382.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2530.0,
|
|
"completions/max_terminated_length": 2530.0,
|
|
"completions/mean_length": 879.79296875,
|
|
"completions/mean_terminated_length": 897.3187255859375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 324.0,
|
|
"epoch": 0.2048,
|
|
"grad_norm": 0.2249525636434555,
|
|
"learning_rate": 2.5000000000000004e-07,
|
|
"loss": -0.027,
|
|
"num_tokens": 54859276.0,
|
|
"reward": 1.177734375,
|
|
"reward_std": 0.17005354166030884,
|
|
"rewards/accuracy_reward_step": 0.6875,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 192
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.95703125,
|
|
"calib/avg_num_step_conf": 15.44921875,
|
|
"calib/ece": 0.11563265306122451,
|
|
"calib/final_conf_rate": 0.95703125,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.004081632653061225,
|
|
"calib/gap": -0.02923701298701309,
|
|
"calib/mean_conf": 0.6700816326530613,
|
|
"calib/mu_c": 0.6608928571428572,
|
|
"calib/mu_w": 0.6901298701298703,
|
|
"calib/nonempty_final_conf_rate": 0.95703125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.05000000000000001,
|
|
"calib/std_conf": 0.06241396911102182,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5874590909090909,
|
|
"calib/step_q_c_n": 2200.0,
|
|
"calib/step_q_gap": -0.062181934731934674,
|
|
"calib/step_q_w": 0.6496410256410255,
|
|
"calib/step_q_w_n": 1755.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2508.0,
|
|
"completions/max_terminated_length": 2508.0,
|
|
"completions/mean_length": 807.765625,
|
|
"completions/mean_terminated_length": 840.6016235351562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 380.0,
|
|
"epoch": 0.20586666666666667,
|
|
"grad_norm": 0.28256791830062866,
|
|
"learning_rate": 2.2222222222222224e-07,
|
|
"loss": -0.0349,
|
|
"num_tokens": 55171776.0,
|
|
"reward": 1.134765625,
|
|
"reward_std": 0.21509431302547455,
|
|
"rewards/accuracy_reward_step": 0.65625,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 193
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 14.4375,
|
|
"calib/ece": 0.09863281249999997,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0078125,
|
|
"calib/gap": -0.015414866032843477,
|
|
"calib/mean_conf": 0.6708203125,
|
|
"calib/mu_c": 0.6661235955056181,
|
|
"calib/mu_w": 0.6815384615384615,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.037070312499999994,
|
|
"calib/std_conf": 0.06361283449432469,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5915540820555326,
|
|
"calib/step_q_c_n": 2413.0,
|
|
"calib/step_q_gap": -0.009856673984997455,
|
|
"calib/step_q_w": 0.6014107560405301,
|
|
"calib/step_q_w_n": 1283.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2169.0,
|
|
"completions/max_terminated_length": 2169.0,
|
|
"completions/mean_length": 830.1953125,
|
|
"completions/mean_terminated_length": 836.7322998046875,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 396.0,
|
|
"epoch": 0.20693333333333333,
|
|
"grad_norm": 0.7657372355461121,
|
|
"learning_rate": 1.9444444444444447e-07,
|
|
"loss": 0.0052,
|
|
"num_tokens": 55490250.0,
|
|
"reward": 1.1953125,
|
|
"reward_std": 0.1434774398803711,
|
|
"rewards/accuracy_reward_step": 0.6953125,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 194
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98828125,
|
|
"calib/avg_num_step_conf": 15.44140625,
|
|
"calib/ece": 0.1362055335968379,
|
|
"calib/final_conf_rate": 0.98828125,
|
|
"calib/format_rate": 0.98828125,
|
|
"calib/frac_conf_gt_0.9": 0.015810276679841896,
|
|
"calib/gap": -0.028996151804670967,
|
|
"calib/mean_conf": 0.6756521739130434,
|
|
"calib/mu_c": 0.6646496815286624,
|
|
"calib/mu_w": 0.6936458333333334,
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0956521739130435,
|
|
"calib/std_conf": 0.0671679327575762,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5884119570294255,
|
|
"calib/step_q_c_n": 2141.0,
|
|
"calib/step_q_gap": -0.02956927917366514,
|
|
"calib/step_q_w": 0.6179812362030906,
|
|
"calib/step_q_w_n": 1812.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2339.0,
|
|
"completions/max_terminated_length": 2339.0,
|
|
"completions/mean_length": 861.03125,
|
|
"completions/mean_terminated_length": 874.698486328125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 374.0,
|
|
"epoch": 0.208,
|
|
"grad_norm": 0.24327611923217773,
|
|
"learning_rate": 1.6666666666666668e-07,
|
|
"loss": -0.016,
|
|
"num_tokens": 55816658.0,
|
|
"reward": 1.107421875,
|
|
"reward_std": 0.24357575178146362,
|
|
"rewards/accuracy_reward_step": 0.61328125,
|
|
"rewards/format_reward_step": 0.98828125,
|
|
"step": 195
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 1.0,
|
|
"calib/avg_num_step_conf": 12.58203125,
|
|
"calib/ece": 0.11406249999999982,
|
|
"calib/final_conf_rate": 1.0,
|
|
"calib/format_rate": 1.0,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.01078947368421046,
|
|
"calib/mean_conf": 0.655703125,
|
|
"calib/mu_c": 0.6525,
|
|
"calib/mu_w": 0.6632894736842104,
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0333203125,
|
|
"calib/std_conf": 0.044338604682989016,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5792476489028213,
|
|
"calib/step_q_c_n": 2233.0,
|
|
"calib/step_q_gap": -0.006470974578960065,
|
|
"calib/step_q_w": 0.5857186234817814,
|
|
"calib/step_q_w_n": 988.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2000.0,
|
|
"completions/max_terminated_length": 2000.0,
|
|
"completions/mean_length": 739.484375,
|
|
"completions/mean_terminated_length": 745.3070678710938,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 368.0,
|
|
"epoch": 0.20906666666666668,
|
|
"grad_norm": 0.4479708671569824,
|
|
"learning_rate": 1.3888888888888888e-07,
|
|
"loss": 0.0002,
|
|
"num_tokens": 56108510.0,
|
|
"reward": 1.203125,
|
|
"reward_std": 0.15190494060516357,
|
|
"rewards/accuracy_reward_step": 0.703125,
|
|
"rewards/format_reward_step": 1.0,
|
|
"step": 196
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.99609375,
|
|
"calib/avg_num_step_conf": 14.3203125,
|
|
"calib/ece": 0.05960784313725492,
|
|
"calib/final_conf_rate": 0.99609375,
|
|
"calib/format_rate": 0.99609375,
|
|
"calib/frac_conf_gt_0.9": 0.011764705882352941,
|
|
"calib/gap": -0.014470895830466346,
|
|
"calib/mean_conf": 0.6685490196078431,
|
|
"calib/mu_c": 0.6636686390532545,
|
|
"calib/mu_w": 0.6781395348837208,
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.03270588235294117,
|
|
"calib/std_conf": 0.060803867822500295,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5914689507494646,
|
|
"calib/step_q_c_n": 2335.0,
|
|
"calib/step_q_gap": -0.020138862924464762,
|
|
"calib/step_q_w": 0.6116078136739294,
|
|
"calib/step_q_w_n": 1331.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2246.0,
|
|
"completions/max_terminated_length": 2246.0,
|
|
"completions/mean_length": 834.4453125,
|
|
"completions/mean_terminated_length": 841.0157470703125,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 354.0,
|
|
"epoch": 0.21013333333333334,
|
|
"grad_norm": 0.24074199795722961,
|
|
"learning_rate": 1.1111111111111112e-07,
|
|
"loss": -0.018,
|
|
"num_tokens": 56427184.0,
|
|
"reward": 1.158203125,
|
|
"reward_std": 0.19163835048675537,
|
|
"rewards/accuracy_reward_step": 0.66015625,
|
|
"rewards/format_reward_step": 0.99609375,
|
|
"step": 197
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.95703125,
|
|
"calib/avg_num_step_conf": 15.453125,
|
|
"calib/ece": 0.1080408163265306,
|
|
"calib/final_conf_rate": 0.95703125,
|
|
"calib/format_rate": 0.95703125,
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
|
"calib/gap": -0.022599967621823036,
|
|
"calib/mean_conf": 0.6618367346938775,
|
|
"calib/mu_c": 0.655287356321839,
|
|
"calib/mu_w": 0.677887323943662,
|
|
"calib/nonempty_final_conf_rate": 0.95703125,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.029836734693877567,
|
|
"calib/std_conf": 0.06065201746018707,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5824627202892003,
|
|
"calib/step_q_c_n": 2213.0,
|
|
"calib/step_q_gap": -0.0915590812024808,
|
|
"calib/step_q_w": 0.6740218014916811,
|
|
"calib/step_q_w_n": 1743.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2851.0,
|
|
"completions/max_terminated_length": 2851.0,
|
|
"completions/mean_length": 781.96875,
|
|
"completions/mean_terminated_length": 817.0775146484375,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 306.0,
|
|
"epoch": 0.2112,
|
|
"grad_norm": 0.29830166697502136,
|
|
"learning_rate": 8.333333333333334e-08,
|
|
"loss": -0.0408,
|
|
"num_tokens": 56732752.0,
|
|
"reward": 1.158203125,
|
|
"reward_std": 0.26918303966522217,
|
|
"rewards/accuracy_reward_step": 0.6796875,
|
|
"rewards/format_reward_step": 0.95703125,
|
|
"step": 198
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.96484375,
|
|
"calib/avg_num_step_conf": 14.7265625,
|
|
"calib/ece": 0.2073279352226721,
|
|
"calib/final_conf_rate": 0.96484375,
|
|
"calib/format_rate": 0.96484375,
|
|
"calib/frac_conf_gt_0.9": 0.004048582995951417,
|
|
"calib/gap": -0.03895469769081594,
|
|
"calib/mean_conf": 0.6691902834008097,
|
|
"calib/mu_c": 0.6595698924731184,
|
|
"calib/mu_w": 0.6985245901639343,
|
|
"calib/nonempty_final_conf_rate": 0.96484375,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.0617408906882591,
|
|
"calib/std_conf": 0.07207840020445129,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5907238493723849,
|
|
"calib/step_q_c_n": 2390.0,
|
|
"calib/step_q_gap": -0.0657415129464557,
|
|
"calib/step_q_w": 0.6564653623188406,
|
|
"calib/step_q_w_n": 1380.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2444.0,
|
|
"completions/max_terminated_length": 2444.0,
|
|
"completions/mean_length": 811.703125,
|
|
"completions/mean_terminated_length": 841.2793579101562,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 317.0,
|
|
"epoch": 0.21226666666666666,
|
|
"grad_norm": 0.3444276750087738,
|
|
"learning_rate": 5.555555555555556e-08,
|
|
"loss": -0.0413,
|
|
"num_tokens": 57044748.0,
|
|
"reward": 1.208984375,
|
|
"reward_std": 0.21363122761249542,
|
|
"rewards/accuracy_reward_step": 0.7265625,
|
|
"rewards/format_reward_step": 0.96484375,
|
|
"step": 199
|
|
},
|
|
{
|
|
"calib/answer_extract_rate": 0.98046875,
|
|
"calib/avg_num_step_conf": 14.8671875,
|
|
"calib/ece": 0.20741035856573703,
|
|
"calib/final_conf_rate": 0.98046875,
|
|
"calib/format_rate": 0.98046875,
|
|
"calib/frac_conf_gt_0.9": 0.03187250996015936,
|
|
"calib/gap": -0.06286187845303881,
|
|
"calib/mean_conf": 0.6726693227091634,
|
|
"calib/mu_c": 0.6551381215469613,
|
|
"calib/mu_w": 0.7180000000000001,
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
|
"calib/pce": 0.07948207171314742,
|
|
"calib/std_conf": 0.07831674256807128,
|
|
"calib/step_conf_rate": 1.0,
|
|
"calib/step_q_c": 0.5828472222222223,
|
|
"calib/step_q_c_n": 2304.0,
|
|
"calib/step_q_gap": -0.0629184235833703,
|
|
"calib/step_q_w": 0.6457656458055926,
|
|
"calib/step_q_w_n": 1502.0,
|
|
"clip_ratio/high_max": 0.0,
|
|
"clip_ratio/high_mean": 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": 2474.0,
|
|
"completions/max_terminated_length": 2474.0,
|
|
"completions/mean_length": 855.87890625,
|
|
"completions/mean_terminated_length": 872.9282836914062,
|
|
"completions/min_length": 0.0,
|
|
"completions/min_terminated_length": 334.0,
|
|
"epoch": 0.21333333333333335,
|
|
"grad_norm": 0.12666837871074677,
|
|
"learning_rate": 2.777777777777778e-08,
|
|
"loss": -0.0451,
|
|
"num_tokens": 57371901.0,
|
|
"reward": 1.197265625,
|
|
"reward_std": 0.10271690785884857,
|
|
"rewards/accuracy_reward_step": 0.70703125,
|
|
"rewards/format_reward_step": 0.98046875,
|
|
"step": 200
|
|
},
|
|
{
|
|
"epoch": 0.21333333333333335,
|
|
"step": 200,
|
|
"total_flos": 0.0,
|
|
"train_loss": -0.012864412539638579,
|
|
"train_runtime": 7417.8224,
|
|
"train_samples_per_second": 6.902,
|
|
"train_steps_per_second": 0.027
|
|
}
|
|
],
|
|
"logging_steps": 1,
|
|
"max_steps": 200,
|
|
"num_input_tokens_seen": 57371901,
|
|
"num_train_epochs": 1,
|
|
"save_steps": 25,
|
|
"stateful_callbacks": {
|
|
"TrainerControl": {
|
|
"args": {
|
|
"should_epoch_stop": false,
|
|
"should_evaluate": false,
|
|
"should_log": false,
|
|
"should_save": true,
|
|
"should_training_stop": true
|
|
},
|
|
"attributes": {}
|
|
}
|
|
},
|
|
"total_flos": 0.0,
|
|
"train_batch_size": 4,
|
|
"trial_name": null,
|
|
"trial_params": null
|
|
}
|