8540 lines
329 KiB
JSON
8540 lines
329 KiB
JSON
|
{
|
||
|
|
"best_metric": null,
|
||
|
|
"best_model_checkpoint": null,
|
||
|
|
"epoch": 0.21333333333333335,
|
||
|
|
"eval_steps": 500,
|
||
|
|
"global_step": 200,
|
||
|
|
"is_hyper_param_search": false,
|
||
|
|
"is_local_process_zero": true,
|
||
|
|
"is_world_process_zero": true,
|
||
|
|
"log_history": [
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.015625,
|
||
|
|
"calib/avg_num_step_conf": 0.00390625,
|
||
|
|
"calib/final_conf_rate": 0.0,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.01953125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.00390625,
|
||
|
|
"calib/step_conf_rate": 0.00390625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.125,
|
||
|
|
"completions/max_length": 3022.0,
|
||
|
|
"completions/max_terminated_length": 3022.0,
|
||
|
|
"completions/mean_length": 584.16015625,
|
||
|
|
"completions/mean_terminated_length": 667.6116333007812,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.0010666666666666667,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 2.5000000000000004e-07,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 232809.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 1
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0234375,
|
||
|
|
"calib/avg_num_step_conf": 0.00390625,
|
||
|
|
"calib/ece": 0.9,
|
||
|
|
"calib/final_conf_rate": 0.00390625,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/mean_conf": 0.9,
|
||
|
|
"calib/mu_c": NaN,
|
||
|
|
"calib/mu_w": 0.9,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.00390625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.02734375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.00390625,
|
||
|
|
"calib/pce": 0.9,
|
||
|
|
"calib/std_conf": 0.0,
|
||
|
|
"calib/step_conf_rate": 0.00390625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3054.0,
|
||
|
|
"completions/max_terminated_length": 3054.0,
|
||
|
|
"completions/mean_length": 643.5234375,
|
||
|
|
"completions/mean_terminated_length": 725.7356567382812,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.0021333333333333334,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 5.000000000000001e-07,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 476519.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 2
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.015625,
|
||
|
|
"calib/avg_num_step_conf": 0.01953125,
|
||
|
|
"calib/final_conf_rate": 0.0,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.0234375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0078125,
|
||
|
|
"calib/step_conf_rate": 0.0078125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3072.0,
|
||
|
|
"completions/max_terminated_length": 3072.0,
|
||
|
|
"completions/mean_length": 736.22265625,
|
||
|
|
"completions/mean_terminated_length": 812.3836059570312,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 3.0,
|
||
|
|
"epoch": 0.0032,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 7.5e-07,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 745928.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 3
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0078125,
|
||
|
|
"calib/avg_num_step_conf": 0.0,
|
||
|
|
"calib/final_conf_rate": 0.0,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.0078125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0,
|
||
|
|
"calib/step_conf_rate": 0.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3022.0,
|
||
|
|
"completions/max_terminated_length": 3022.0,
|
||
|
|
"completions/mean_length": 764.01953125,
|
||
|
|
"completions/mean_terminated_length": 854.1004638671875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 4.0,
|
||
|
|
"epoch": 0.004266666666666667,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 1.0000000000000002e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 1023365.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 4
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0234375,
|
||
|
|
"calib/avg_num_step_conf": 0.046875,
|
||
|
|
"calib/final_conf_rate": 0.0,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.03125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.01171875,
|
||
|
|
"calib/step_conf_rate": 0.01171875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.15625,
|
||
|
|
"completions/max_length": 2980.0,
|
||
|
|
"completions/max_terminated_length": 2980.0,
|
||
|
|
"completions/mean_length": 677.6640625,
|
||
|
|
"completions/mean_terminated_length": 803.1574096679688,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 3.0,
|
||
|
|
"epoch": 0.005333333333333333,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 1.25e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 1279215.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 5
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.03515625,
|
||
|
|
"calib/auroc": 0.0,
|
||
|
|
"calib/avg_num_step_conf": 0.015625,
|
||
|
|
"calib/ece": 0.595,
|
||
|
|
"calib/final_conf_rate": 0.0078125,
|
||
|
|
"calib/format_rate": 0.0078125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.5,
|
||
|
|
"calib/gap": -0.18999999999999995,
|
||
|
|
"calib/mean_conf": 0.895,
|
||
|
|
"calib/mu_c": 0.8,
|
||
|
|
"calib/mu_w": 0.99,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0078125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.03515625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0078125,
|
||
|
|
"calib/pce": 0.495,
|
||
|
|
"calib/std_conf": 0.09499999999999997,
|
||
|
|
"calib/step_conf_rate": 0.0078125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.12109375,
|
||
|
|
"completions/max_length": 2959.0,
|
||
|
|
"completions/max_terminated_length": 2959.0,
|
||
|
|
"completions/mean_length": 629.1484375,
|
||
|
|
"completions/mean_terminated_length": 715.8311157226562,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.0064,
|
||
|
|
"grad_norm": 0.002231498947367072,
|
||
|
|
"learning_rate": 1.5e-06,
|
||
|
|
"loss": -0.002,
|
||
|
|
"num_tokens": 1521909.0,
|
||
|
|
"reward": 0.009758099913597107,
|
||
|
|
"reward_std": 0.02760007604956627,
|
||
|
|
"rewards/accuracy_reward_step": 0.00390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.0038277343846857548,
|
||
|
|
"rewards/format_reward_step": 0.0078125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.006376933306455612,
|
||
|
|
"step": 6
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.04296875,
|
||
|
|
"calib/auroc": 0.0,
|
||
|
|
"calib/avg_num_step_conf": 0.02734375,
|
||
|
|
"calib/ece": 0.595,
|
||
|
|
"calib/final_conf_rate": 0.0078125,
|
||
|
|
"calib/format_rate": 0.0078125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.5,
|
||
|
|
"calib/gap": -0.18999999999999995,
|
||
|
|
"calib/mean_conf": 0.895,
|
||
|
|
"calib/mu_c": 0.8,
|
||
|
|
"calib/mu_w": 0.99,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0078125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.046875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.015625,
|
||
|
|
"calib/pce": 0.495,
|
||
|
|
"calib/std_conf": 0.09499999999999997,
|
||
|
|
"calib/step_conf_rate": 0.015625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.125,
|
||
|
|
"completions/max_length": 2870.0,
|
||
|
|
"completions/max_terminated_length": 2870.0,
|
||
|
|
"completions/mean_length": 650.34375,
|
||
|
|
"completions/mean_terminated_length": 743.2500610351562,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.007466666666666667,
|
||
|
|
"grad_norm": 0.0029239056166261435,
|
||
|
|
"learning_rate": 1.75e-06,
|
||
|
|
"loss": 0.003,
|
||
|
|
"num_tokens": 1771501.0,
|
||
|
|
"reward": 0.0094246044754982,
|
||
|
|
"reward_std": 0.026656806468963623,
|
||
|
|
"rewards/accuracy_reward_step": 0.00390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.0038277343846857548,
|
||
|
|
"rewards/format_reward_step": 0.0078125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.005042951088398695,
|
||
|
|
"step": 7
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0234375,
|
||
|
|
"calib/avg_num_step_conf": 0.0078125,
|
||
|
|
"calib/ece": 0.30000000000000004,
|
||
|
|
"calib/final_conf_rate": 0.00390625,
|
||
|
|
"calib/format_rate": 0.00390625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/mean_conf": 0.7,
|
||
|
|
"calib/mu_c": 0.7,
|
||
|
|
"calib/mu_w": NaN,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.00390625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.02734375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0078125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.0,
|
||
|
|
"calib/step_conf_rate": 0.0078125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.09765625,
|
||
|
|
"completions/max_length": 3011.0,
|
||
|
|
"completions/max_terminated_length": 3011.0,
|
||
|
|
"completions/mean_length": 648.08203125,
|
||
|
|
"completions/mean_terminated_length": 718.2207641601562,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.008533333333333334,
|
||
|
|
"grad_norm": 0.0007931651780381799,
|
||
|
|
"learning_rate": 2.0000000000000003e-06,
|
||
|
|
"loss": 0.005,
|
||
|
|
"num_tokens": 2019602.0,
|
||
|
|
"reward": 0.007640509866178036,
|
||
|
|
"reward_std": 0.021610625088214874,
|
||
|
|
"rewards/accuracy_reward_step": 0.00390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.003554687602445483,
|
||
|
|
"rewards/format_reward_step": 0.00390625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0031401650048792362,
|
||
|
|
"step": 8
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.02734375,
|
||
|
|
"calib/avg_num_step_conf": 0.01171875,
|
||
|
|
"calib/final_conf_rate": 0.0,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.03125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.00390625,
|
||
|
|
"calib/step_conf_rate": 0.00390625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2965.0,
|
||
|
|
"completions/max_terminated_length": 2965.0,
|
||
|
|
"completions/mean_length": 677.796875,
|
||
|
|
"completions/mean_terminated_length": 747.913818359375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 3.0,
|
||
|
|
"epoch": 0.0096,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 2.25e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 2276334.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 9
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.01953125,
|
||
|
|
"calib/avg_num_step_conf": 0.00390625,
|
||
|
|
"calib/final_conf_rate": 0.0,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.0234375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.00390625,
|
||
|
|
"calib/step_conf_rate": 0.00390625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3049.0,
|
||
|
|
"completions/max_terminated_length": 3049.0,
|
||
|
|
"completions/mean_length": 776.8125,
|
||
|
|
"completions/mean_terminated_length": 872.2105102539062,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 5.0,
|
||
|
|
"epoch": 0.010666666666666666,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 2.5e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 2557678.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 10
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0234375,
|
||
|
|
"calib/avg_num_step_conf": 0.03515625,
|
||
|
|
"calib/ece": 0.55,
|
||
|
|
"calib/final_conf_rate": 0.0078125,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.5,
|
||
|
|
"calib/mean_conf": 0.55,
|
||
|
|
"calib/mu_c": NaN,
|
||
|
|
"calib/mu_w": 0.55,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0078125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.03515625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.01171875,
|
||
|
|
"calib/pce": 0.55,
|
||
|
|
"calib/std_conf": 0.45,
|
||
|
|
"calib/step_conf_rate": 0.01171875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3006.0,
|
||
|
|
"completions/max_terminated_length": 3006.0,
|
||
|
|
"completions/mean_length": 763.171875,
|
||
|
|
"completions/mean_terminated_length": 853.15283203125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 3.0,
|
||
|
|
"epoch": 0.011733333333333333,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 2.7500000000000004e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 2833210.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 11
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.02734375,
|
||
|
|
"calib/avg_num_step_conf": 0.0,
|
||
|
|
"calib/ece": 0.44375000000000003,
|
||
|
|
"calib/final_conf_rate": 0.0078125,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/mean_conf": 0.44375000000000003,
|
||
|
|
"calib/mu_c": NaN,
|
||
|
|
"calib/mu_w": 0.44375000000000003,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0078125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.02734375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0,
|
||
|
|
"calib/pce": 0.44375000000000003,
|
||
|
|
"calib/std_conf": 0.35625000000000007,
|
||
|
|
"calib/step_conf_rate": 0.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3043.0,
|
||
|
|
"completions/max_terminated_length": 3043.0,
|
||
|
|
"completions/mean_length": 722.77734375,
|
||
|
|
"completions/mean_terminated_length": 807.9956665039062,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.0128,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 3e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 3098097.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 12
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.02734375,
|
||
|
|
"calib/auroc": 0.0,
|
||
|
|
"calib/avg_num_step_conf": 0.03125,
|
||
|
|
"calib/ece": 0.6,
|
||
|
|
"calib/final_conf_rate": 0.0078125,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.5,
|
||
|
|
"calib/gap": -0.19999999999999996,
|
||
|
|
"calib/mean_conf": 0.9,
|
||
|
|
"calib/mu_c": 0.8,
|
||
|
|
"calib/mu_w": 1.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0078125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.03515625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.01171875,
|
||
|
|
"calib/pce": 0.5,
|
||
|
|
"calib/std_conf": 0.09999999999999998,
|
||
|
|
"calib/step_conf_rate": 0.01171875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3041.0,
|
||
|
|
"completions/max_terminated_length": 3041.0,
|
||
|
|
"completions/mean_length": 639.03515625,
|
||
|
|
"completions/mean_terminated_length": 702.1159057617188,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.013866666666666666,
|
||
|
|
"grad_norm": 0.001644036383368075,
|
||
|
|
"learning_rate": 3.2500000000000002e-06,
|
||
|
|
"loss": 0.0008,
|
||
|
|
"num_tokens": 3341962.0,
|
||
|
|
"reward": 0.00390625,
|
||
|
|
"reward_std": 0.011048543266952038,
|
||
|
|
"rewards/accuracy_reward_step": 0.00390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 13
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0234375,
|
||
|
|
"calib/avg_num_step_conf": 0.01171875,
|
||
|
|
"calib/ece": 0.19999999999999996,
|
||
|
|
"calib/final_conf_rate": 0.00390625,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/mean_conf": 0.8,
|
||
|
|
"calib/mu_c": 0.8,
|
||
|
|
"calib/mu_w": NaN,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.00390625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.03125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0078125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.0,
|
||
|
|
"calib/step_conf_rate": 0.0078125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0859375,
|
||
|
|
"completions/max_length": 3012.0,
|
||
|
|
"completions/max_terminated_length": 3012.0,
|
||
|
|
"completions/mean_length": 672.89453125,
|
||
|
|
"completions/mean_terminated_length": 736.1581420898438,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 3.0,
|
||
|
|
"epoch": 0.014933333333333333,
|
||
|
|
"grad_norm": 0.0014701585751026869,
|
||
|
|
"learning_rate": 3.5e-06,
|
||
|
|
"loss": 0.0094,
|
||
|
|
"num_tokens": 3595303.0,
|
||
|
|
"reward": 0.0078125,
|
||
|
|
"reward_std": 0.022097086533904076,
|
||
|
|
"rewards/accuracy_reward_step": 0.0078125,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 14
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0234375,
|
||
|
|
"calib/avg_num_step_conf": 0.0078125,
|
||
|
|
"calib/ece": 0.0,
|
||
|
|
"calib/final_conf_rate": 0.00390625,
|
||
|
|
"calib/format_rate": 0.00390625,
|
||
|
|
"calib/frac_conf_gt_0.9": 1.0,
|
||
|
|
"calib/mean_conf": 1.0,
|
||
|
|
"calib/mu_c": 1.0,
|
||
|
|
"calib/mu_w": NaN,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.00390625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.02734375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0078125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.0,
|
||
|
|
"calib/step_conf_rate": 0.0078125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2938.0,
|
||
|
|
"completions/max_terminated_length": 2938.0,
|
||
|
|
"completions/mean_length": 710.86328125,
|
||
|
|
"completions/mean_terminated_length": 771.10595703125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.016,
|
||
|
|
"grad_norm": 0.0019930792041122913,
|
||
|
|
"learning_rate": 3.7500000000000005e-06,
|
||
|
|
"loss": -0.0004,
|
||
|
|
"num_tokens": 3860844.0,
|
||
|
|
"reward": 0.007889438420534134,
|
||
|
|
"reward_std": 0.022314703091979027,
|
||
|
|
"rewards/accuracy_reward_step": 0.00390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.00390625,
|
||
|
|
"rewards/format_reward_step": 0.00390625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0034327523317188025,
|
||
|
|
"step": 15
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.02734375,
|
||
|
|
"calib/avg_num_step_conf": 0.05078125,
|
||
|
|
"calib/ece": 1.0,
|
||
|
|
"calib/final_conf_rate": 0.00390625,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 1.0,
|
||
|
|
"calib/mean_conf": 1.0,
|
||
|
|
"calib/mu_c": NaN,
|
||
|
|
"calib/mu_w": 1.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.00390625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.03515625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0078125,
|
||
|
|
"calib/pce": 1.0,
|
||
|
|
"calib/std_conf": 0.0,
|
||
|
|
"calib/step_conf_rate": 0.0078125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3006.0,
|
||
|
|
"completions/max_terminated_length": 3006.0,
|
||
|
|
"completions/mean_length": 760.90625,
|
||
|
|
"completions/mean_terminated_length": 854.3508911132812,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.017066666666666667,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 4.000000000000001e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 4140164.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 16
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.02734375,
|
||
|
|
"calib/avg_num_step_conf": 0.00390625,
|
||
|
|
"calib/final_conf_rate": 0.0,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.03125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.00390625,
|
||
|
|
"calib/step_conf_rate": 0.00390625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3033.0,
|
||
|
|
"completions/max_terminated_length": 3033.0,
|
||
|
|
"completions/mean_length": 655.0546875,
|
||
|
|
"completions/mean_terminated_length": 735.5,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 4.0,
|
||
|
|
"epoch": 0.018133333333333335,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 4.25e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 4387066.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 17
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.02734375,
|
||
|
|
"calib/avg_num_step_conf": 0.01953125,
|
||
|
|
"calib/final_conf_rate": 0.0,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.03125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.00390625,
|
||
|
|
"calib/step_conf_rate": 0.00390625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0859375,
|
||
|
|
"completions/max_length": 3066.0,
|
||
|
|
"completions/max_terminated_length": 3066.0,
|
||
|
|
"completions/mean_length": 736.14453125,
|
||
|
|
"completions/mean_terminated_length": 805.354736328125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.0192,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 4.5e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 4661919.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 18
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0390625,
|
||
|
|
"calib/avg_num_step_conf": 0.0,
|
||
|
|
"calib/ece": 0.0,
|
||
|
|
"calib/final_conf_rate": 0.00390625,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/mean_conf": 0.0,
|
||
|
|
"calib/mu_c": NaN,
|
||
|
|
"calib/mu_w": 0.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.00390625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.0390625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.0,
|
||
|
|
"calib/step_conf_rate": 0.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2964.0,
|
||
|
|
"completions/max_terminated_length": 2964.0,
|
||
|
|
"completions/mean_length": 628.3046875,
|
||
|
|
"completions/mean_terminated_length": 672.9957885742188,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.020266666666666665,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 4.75e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 4903205.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 19
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0625,
|
||
|
|
"calib/avg_num_step_conf": 0.01953125,
|
||
|
|
"calib/ece": 0.85,
|
||
|
|
"calib/final_conf_rate": 0.015625,
|
||
|
|
"calib/format_rate": 0.00390625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.75,
|
||
|
|
"calib/mean_conf": 0.85,
|
||
|
|
"calib/mu_c": NaN,
|
||
|
|
"calib/mu_w": 0.85,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.015625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.07421875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.01953125,
|
||
|
|
"calib/pce": 0.85,
|
||
|
|
"calib/std_conf": 0.203100960115899,
|
||
|
|
"calib/step_conf_rate": 0.01953125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2848.0,
|
||
|
|
"completions/max_terminated_length": 2848.0,
|
||
|
|
"completions/mean_length": 650.07421875,
|
||
|
|
"completions/mean_terminated_length": 717.3233032226562,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.021333333333333333,
|
||
|
|
"grad_norm": 0.0009754901984706521,
|
||
|
|
"learning_rate": 5e-06,
|
||
|
|
"loss": 0.0065,
|
||
|
|
"num_tokens": 5150176.0,
|
||
|
|
"reward": 0.003434200072661042,
|
||
|
|
"reward_std": 0.009713384322822094,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0029296875,
|
||
|
|
"rewards/format_reward_step": 0.00390625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0031899246387183666,
|
||
|
|
"step": 20
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.05078125,
|
||
|
|
"calib/auroc": 0.875,
|
||
|
|
"calib/avg_num_step_conf": 0.04296875,
|
||
|
|
"calib/ece": 0.5333333333333333,
|
||
|
|
"calib/final_conf_rate": 0.0234375,
|
||
|
|
"calib/format_rate": 0.015625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.5,
|
||
|
|
"calib/gap": 0.20000000000000007,
|
||
|
|
"calib/mean_conf": 0.8666666666666667,
|
||
|
|
"calib/mu_c": 1.0,
|
||
|
|
"calib/mu_w": 0.7999999999999999,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0234375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.0546875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0234375,
|
||
|
|
"calib/pce": 0.5333333333333333,
|
||
|
|
"calib/std_conf": 0.17950549357115014,
|
||
|
|
"calib/step_conf_rate": 0.0234375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2931.0,
|
||
|
|
"completions/max_terminated_length": 2931.0,
|
||
|
|
"completions/mean_length": 675.6796875,
|
||
|
|
"completions/mean_terminated_length": 720.7250366210938,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 3.0,
|
||
|
|
"epoch": 0.0224,
|
||
|
|
"grad_norm": 0.0026306321378797293,
|
||
|
|
"learning_rate": 4.9722222222222224e-06,
|
||
|
|
"loss": 0.0035,
|
||
|
|
"num_tokens": 5401790.0,
|
||
|
|
"reward": 0.021148759871721268,
|
||
|
|
"reward_std": 0.05981772020459175,
|
||
|
|
"rewards/accuracy_reward_step": 0.0078125,
|
||
|
|
"rewards/final_brier_reward_step": 0.011484375223517418,
|
||
|
|
"rewards/format_reward_step": 0.015625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.01162628922611475,
|
||
|
|
"step": 21
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0625,
|
||
|
|
"calib/avg_num_step_conf": 0.0078125,
|
||
|
|
"calib/ece": 0.6666666666666666,
|
||
|
|
"calib/final_conf_rate": 0.01171875,
|
||
|
|
"calib/format_rate": 0.00390625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.6666666666666666,
|
||
|
|
"calib/mean_conf": 0.6666666666666666,
|
||
|
|
"calib/mu_c": NaN,
|
||
|
|
"calib/mu_w": 0.6666666666666666,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.01171875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.06640625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0078125,
|
||
|
|
"calib/pce": 0.6666666666666666,
|
||
|
|
"calib/std_conf": 0.4714045207910317,
|
||
|
|
"calib/step_conf_rate": 0.0078125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.1171875,
|
||
|
|
"completions/max_length": 2723.0,
|
||
|
|
"completions/max_terminated_length": 2723.0,
|
||
|
|
"completions/mean_length": 566.3515625,
|
||
|
|
"completions/mean_terminated_length": 641.5309448242188,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 3.0,
|
||
|
|
"epoch": 0.023466666666666667,
|
||
|
|
"grad_norm": 0.0009022683370858431,
|
||
|
|
"learning_rate": 4.944444444444445e-06,
|
||
|
|
"loss": 0.0022,
|
||
|
|
"num_tokens": 5624272.0,
|
||
|
|
"reward": 0.004101060330867767,
|
||
|
|
"reward_std": 0.011599550023674965,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.00390625,
|
||
|
|
"rewards/format_reward_step": 0.00390625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0039042423013597727,
|
||
|
|
"step": 22
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.03515625,
|
||
|
|
"calib/avg_num_step_conf": 0.03515625,
|
||
|
|
"calib/final_conf_rate": 0.0,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.04296875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.01171875,
|
||
|
|
"calib/step_conf_rate": 0.01171875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.125,
|
||
|
|
"completions/max_length": 2889.0,
|
||
|
|
"completions/max_terminated_length": 2889.0,
|
||
|
|
"completions/mean_length": 531.07421875,
|
||
|
|
"completions/mean_terminated_length": 606.9420166015625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.024533333333333334,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 4.9166666666666665e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 5839843.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 23
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0625,
|
||
|
|
"calib/avg_num_step_conf": 0.07421875,
|
||
|
|
"calib/ece": 0.6125,
|
||
|
|
"calib/final_conf_rate": 0.015625,
|
||
|
|
"calib/format_rate": 0.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/mean_conf": 0.6125,
|
||
|
|
"calib/mu_c": NaN,
|
||
|
|
"calib/mu_w": 0.6125,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.015625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.08203125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0234375,
|
||
|
|
"calib/pce": 0.6125,
|
||
|
|
"calib/std_conf": 0.3542156828826189,
|
||
|
|
"calib/step_conf_rate": 0.0234375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2925.0,
|
||
|
|
"completions/max_terminated_length": 2925.0,
|
||
|
|
"completions/mean_length": 659.71875,
|
||
|
|
"completions/mean_terminated_length": 718.6723022460938,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.0256,
|
||
|
|
"grad_norm": 0.0,
|
||
|
|
"learning_rate": 4.888888888888889e-06,
|
||
|
|
"loss": 0.0,
|
||
|
|
"num_tokens": 6088923.0,
|
||
|
|
"reward": 0.0,
|
||
|
|
"reward_std": 0.0,
|
||
|
|
"rewards/accuracy_reward_step": 0.0,
|
||
|
|
"rewards/final_brier_reward_step": 0.0,
|
||
|
|
"rewards/format_reward_step": 0.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0,
|
||
|
|
"step": 24
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.08984375,
|
||
|
|
"calib/auroc": 0.08333333333333331,
|
||
|
|
"calib/avg_num_step_conf": 0.109375,
|
||
|
|
"calib/ece": 0.784375,
|
||
|
|
"calib/final_conf_rate": 0.03125,
|
||
|
|
"calib/format_rate": 0.0078125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.125,
|
||
|
|
"calib/gap": -0.6125,
|
||
|
|
"calib/mean_conf": 0.609375,
|
||
|
|
"calib/mu_c": 0.15,
|
||
|
|
"calib/mu_w": 0.7625000000000001,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.03125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.109375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.03515625,
|
||
|
|
"calib/pce": 0.571875,
|
||
|
|
"calib/std_conf": 0.35353957964420335,
|
||
|
|
"calib/step_conf_rate": 0.03515625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3065.0,
|
||
|
|
"completions/max_terminated_length": 3065.0,
|
||
|
|
"completions/mean_length": 618.3984375,
|
||
|
|
"completions/mean_terminated_length": 656.8880004882812,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 3.0,
|
||
|
|
"epoch": 0.02666666666666667,
|
||
|
|
"grad_norm": 0.0028888338711112738,
|
||
|
|
"learning_rate": 4.861111111111111e-06,
|
||
|
|
"loss": 0.0091,
|
||
|
|
"num_tokens": 6326137.0,
|
||
|
|
"reward": 0.015379859134554863,
|
||
|
|
"reward_std": 0.043500810861587524,
|
||
|
|
"rewards/accuracy_reward_step": 0.01171875,
|
||
|
|
"rewards/final_brier_reward_step": 0.0011230468517169356,
|
||
|
|
"rewards/format_reward_step": 0.0078125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.0030233401339501143,
|
||
|
|
"step": 25
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.07421875,
|
||
|
|
"calib/auroc": 0.5,
|
||
|
|
"calib/avg_num_step_conf": 0.0859375,
|
||
|
|
"calib/ece": 0.309,
|
||
|
|
"calib/final_conf_rate": 0.0390625,
|
||
|
|
"calib/format_rate": 0.0078125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.3,
|
||
|
|
"calib/gap": 0.2141666666666665,
|
||
|
|
"calib/mean_conf": 0.709,
|
||
|
|
"calib/mu_c": 0.8374999999999999,
|
||
|
|
"calib/mu_w": 0.6233333333333334,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.0390625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.08984375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.02734375,
|
||
|
|
"calib/pce": 0.309,
|
||
|
|
"calib/std_conf": 0.33738553614522365,
|
||
|
|
"calib/step_conf_rate": 0.02734375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3065.0,
|
||
|
|
"completions/max_terminated_length": 3065.0,
|
||
|
|
"completions/mean_length": 659.8671875,
|
||
|
|
"completions/mean_terminated_length": 706.8033447265625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 9.0,
|
||
|
|
"epoch": 0.027733333333333332,
|
||
|
|
"grad_norm": 0.0023705705534666777,
|
||
|
|
"learning_rate": 4.833333333333333e-06,
|
||
|
|
"loss": -0.0318,
|
||
|
|
"num_tokens": 6575983.0,
|
||
|
|
"reward": 0.02112744003534317,
|
||
|
|
"reward_std": 0.059757426381111145,
|
||
|
|
"rewards/accuracy_reward_step": 0.015625,
|
||
|
|
"rewards/final_brier_reward_step": 0.0038277343846857548,
|
||
|
|
"rewards/format_reward_step": 0.0078125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.004979296587407589,
|
||
|
|
"step": 26
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.08984375,
|
||
|
|
"calib/auroc": 0.8214285714285714,
|
||
|
|
"calib/avg_num_step_conf": 0.02734375,
|
||
|
|
"calib/ece": 0.5401111111111111,
|
||
|
|
"calib/final_conf_rate": 0.03515625,
|
||
|
|
"calib/format_rate": 0.0078125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.2222222222222222,
|
||
|
|
"calib/gap": 0.23871428571428566,
|
||
|
|
"calib/mean_conf": 0.7623333333333334,
|
||
|
|
"calib/mu_c": 0.948,
|
||
|
|
"calib/mu_w": 0.7092857142857143,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.03515625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.09765625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.01953125,
|
||
|
|
"calib/pce": 0.5401111111111111,
|
||
|
|
"calib/std_conf": 0.26555434681268375,
|
||
|
|
"calib/step_conf_rate": 0.01953125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3041.0,
|
||
|
|
"completions/max_terminated_length": 3041.0,
|
||
|
|
"completions/mean_length": 663.38671875,
|
||
|
|
"completions/mean_terminated_length": 687.5587158203125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 4.0,
|
||
|
|
"epoch": 0.0288,
|
||
|
|
"grad_norm": 0.0021662807557731867,
|
||
|
|
"learning_rate": 4.805555555555556e-06,
|
||
|
|
"loss": 0.0148,
|
||
|
|
"num_tokens": 6826706.0,
|
||
|
|
"reward": 0.01795351132750511,
|
||
|
|
"reward_std": 0.05078019201755524,
|
||
|
|
"rewards/accuracy_reward_step": 0.01171875,
|
||
|
|
"rewards/final_brier_reward_step": 0.005313574336469173,
|
||
|
|
"rewards/format_reward_step": 0.0078125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.004936892073601484,
|
||
|
|
"step": 27
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.0859375,
|
||
|
|
"calib/auroc": 0.9500000000000001,
|
||
|
|
"calib/avg_num_step_conf": 0.10546875,
|
||
|
|
"calib/ece": 0.5206666666666667,
|
||
|
|
"calib/final_conf_rate": 0.046875,
|
||
|
|
"calib/format_rate": 0.02734375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.4166666666666667,
|
||
|
|
"calib/gap": 0.36319999999999997,
|
||
|
|
"calib/mean_conf": 0.6873333333333332,
|
||
|
|
"calib/mu_c": 0.99,
|
||
|
|
"calib/mu_w": 0.6268,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.046875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.1015625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.04296875,
|
||
|
|
"calib/pce": 0.5206666666666667,
|
||
|
|
"calib/std_conf": 0.3450143314253611,
|
||
|
|
"calib/step_conf_rate": 0.04296875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2855.0,
|
||
|
|
"completions/max_terminated_length": 2855.0,
|
||
|
|
"completions/mean_length": 608.6796875,
|
||
|
|
"completions/mean_terminated_length": 663.0723266601562,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.029866666666666666,
|
||
|
|
"grad_norm": 0.003722765715792775,
|
||
|
|
"learning_rate": 4.777777777777778e-06,
|
||
|
|
"loss": 0.0091,
|
||
|
|
"num_tokens": 7065152.0,
|
||
|
|
"reward": 0.038684695959091187,
|
||
|
|
"reward_std": 0.09774065762758255,
|
||
|
|
"rewards/accuracy_reward_step": 0.01953125,
|
||
|
|
"rewards/final_brier_reward_step": 0.014613499864935875,
|
||
|
|
"rewards/format_reward_step": 0.02734375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.014574278146028519,
|
||
|
|
"step": 28
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.11328125,
|
||
|
|
"calib/auroc": 0.7916666666666667,
|
||
|
|
"calib/avg_num_step_conf": 0.16796875,
|
||
|
|
"calib/ece": 0.7155692307692308,
|
||
|
|
"calib/final_conf_rate": 0.05078125,
|
||
|
|
"calib/format_rate": 0.0234375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.38461538461538464,
|
||
|
|
"calib/gap": 0.1706333333333332,
|
||
|
|
"calib/mean_conf": 0.7924923076923077,
|
||
|
|
"calib/mu_c": 0.95,
|
||
|
|
"calib/mu_w": 0.7793666666666668,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.05078125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.1328125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.05859375,
|
||
|
|
"calib/pce": 0.7155692307692308,
|
||
|
|
"calib/std_conf": 0.234962682672208,
|
||
|
|
"calib/step_conf_rate": 0.05859375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.05078125,
|
||
|
|
"completions/max_length": 2535.0,
|
||
|
|
"completions/max_terminated_length": 2535.0,
|
||
|
|
"completions/mean_length": 609.9296875,
|
||
|
|
"completions/mean_terminated_length": 642.5596313476562,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.030933333333333334,
|
||
|
|
"grad_norm": 0.0025090023409575224,
|
||
|
|
"learning_rate": 4.75e-06,
|
||
|
|
"loss": 0.0229,
|
||
|
|
"num_tokens": 7304102.0,
|
||
|
|
"reward": 0.022272439673542976,
|
||
|
|
"reward_std": 0.05907098948955536,
|
||
|
|
"rewards/accuracy_reward_step": 0.0078125,
|
||
|
|
"rewards/final_brier_reward_step": 0.007529296912252903,
|
||
|
|
"rewards/format_reward_step": 0.0234375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.014656167477369308,
|
||
|
|
"step": 29
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.1171875,
|
||
|
|
"calib/auroc": 0.625,
|
||
|
|
"calib/avg_num_step_conf": 0.34375,
|
||
|
|
"calib/ece": 0.46705882352941175,
|
||
|
|
"calib/final_conf_rate": 0.06640625,
|
||
|
|
"calib/format_rate": 0.04296875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.47058823529411764,
|
||
|
|
"calib/gap": -0.04902777777777789,
|
||
|
|
"calib/mean_conf": 0.8552941176470588,
|
||
|
|
"calib/mu_c": 0.8322222222222222,
|
||
|
|
"calib/mu_w": 0.8812500000000001,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.06640625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.16796875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.09765625,
|
||
|
|
"calib/pce": 0.39647058823529413,
|
||
|
|
"calib/std_conf": 0.1809490551332977,
|
||
|
|
"calib/step_conf_rate": 0.09765625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3054.0,
|
||
|
|
"completions/max_terminated_length": 3054.0,
|
||
|
|
"completions/mean_length": 691.58203125,
|
||
|
|
"completions/mean_terminated_length": 750.190673828125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.032,
|
||
|
|
"grad_norm": 0.003610975109040737,
|
||
|
|
"learning_rate": 4.722222222222222e-06,
|
||
|
|
"loss": 0.0245,
|
||
|
|
"num_tokens": 7563811.0,
|
||
|
|
"reward": 0.07396630942821503,
|
||
|
|
"reward_std": 0.15866075456142426,
|
||
|
|
"rewards/accuracy_reward_step": 0.04296875,
|
||
|
|
"rewards/final_brier_reward_step": 0.022626563906669617,
|
||
|
|
"rewards/format_reward_step": 0.04296875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.02717459946870804,
|
||
|
|
"step": 30
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.17578125,
|
||
|
|
"calib/auroc": 0.615079365079365,
|
||
|
|
"calib/avg_num_step_conf": 0.4375,
|
||
|
|
"calib/ece": 0.5840185185185185,
|
||
|
|
"calib/final_conf_rate": 0.10546875,
|
||
|
|
"calib/format_rate": 0.0546875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.37037037037037035,
|
||
|
|
"calib/gap": 0.025309523809523893,
|
||
|
|
"calib/mean_conf": 0.7469814814814814,
|
||
|
|
"calib/mu_c": 0.7666666666666667,
|
||
|
|
"calib/mu_w": 0.7413571428571428,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.10546875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.21875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.125,
|
||
|
|
"calib/pce": 0.5543888888888888,
|
||
|
|
"calib/std_conf": 0.2969400264139109,
|
||
|
|
"calib/step_conf_rate": 0.125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2963.0,
|
||
|
|
"completions/max_terminated_length": 2963.0,
|
||
|
|
"completions/mean_length": 639.22265625,
|
||
|
|
"completions/mean_terminated_length": 665.207275390625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.03306666666666667,
|
||
|
|
"grad_norm": 0.003975396975874901,
|
||
|
|
"learning_rate": 4.694444444444445e-06,
|
||
|
|
"loss": 0.0077,
|
||
|
|
"num_tokens": 7809044.0,
|
||
|
|
"reward": 0.06872516125440598,
|
||
|
|
"reward_std": 0.15826590359210968,
|
||
|
|
"rewards/accuracy_reward_step": 0.03125,
|
||
|
|
"rewards/final_brier_reward_step": 0.026396874338388443,
|
||
|
|
"rewards/format_reward_step": 0.0546875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.03148187696933746,
|
||
|
|
"step": 31
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.15234375,
|
||
|
|
"calib/auroc": 0.3935185185185185,
|
||
|
|
"calib/avg_num_step_conf": 0.2109375,
|
||
|
|
"calib/ece": 0.6371861111111111,
|
||
|
|
"calib/final_conf_rate": 0.09375,
|
||
|
|
"calib/format_rate": 0.03515625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.25,
|
||
|
|
"calib/gap": -0.09652222222222218,
|
||
|
|
"calib/mean_conf": 0.7436472222222221,
|
||
|
|
"calib/mu_c": 0.6712555555555556,
|
||
|
|
"calib/mu_w": 0.7677777777777778,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.09375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.16796875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.0625,
|
||
|
|
"calib/pce": 0.5654166666666667,
|
||
|
|
"calib/std_conf": 0.23819251354902501,
|
||
|
|
"calib/step_conf_rate": 0.0625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2828.0,
|
||
|
|
"completions/max_terminated_length": 2828.0,
|
||
|
|
"completions/mean_length": 632.8515625,
|
||
|
|
"completions/mean_terminated_length": 669.4627685546875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.034133333333333335,
|
||
|
|
"grad_norm": 0.004196561872959137,
|
||
|
|
"learning_rate": 4.666666666666667e-06,
|
||
|
|
"loss": -0.015,
|
||
|
|
"num_tokens": 8053438.0,
|
||
|
|
"reward": 0.06246958300471306,
|
||
|
|
"reward_std": 0.14494431018829346,
|
||
|
|
"rewards/accuracy_reward_step": 0.03515625,
|
||
|
|
"rewards/final_brier_reward_step": 0.019197266548871994,
|
||
|
|
"rewards/format_reward_step": 0.03515625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.028671305626630783,
|
||
|
|
"step": 32
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.23828125,
|
||
|
|
"calib/auroc": 0.27567567567567564,
|
||
|
|
"calib/avg_num_step_conf": 0.33203125,
|
||
|
|
"calib/ece": 0.696779761904762,
|
||
|
|
"calib/final_conf_rate": 0.1640625,
|
||
|
|
"calib/format_rate": 0.0703125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.3333333333333333,
|
||
|
|
"calib/gap": -0.12012837837837842,
|
||
|
|
"calib/mean_conf": 0.8158273809523809,
|
||
|
|
"calib/mu_c": 0.71,
|
||
|
|
"calib/mu_w": 0.8301283783783784,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.1640625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.27734375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.12890625,
|
||
|
|
"calib/pce": 0.696779761904762,
|
||
|
|
"calib/std_conf": 0.16470455591263616,
|
||
|
|
"calib/step_conf_rate": 0.12890625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0703125,
|
||
|
|
"completions/max_length": 2874.0,
|
||
|
|
"completions/max_terminated_length": 2874.0,
|
||
|
|
"completions/mean_length": 615.3125,
|
||
|
|
"completions/mean_terminated_length": 661.8487548828125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 4.0,
|
||
|
|
"epoch": 0.0352,
|
||
|
|
"grad_norm": 0.005306170787662268,
|
||
|
|
"learning_rate": 4.638888888888889e-06,
|
||
|
|
"loss": 0.0001,
|
||
|
|
"num_tokens": 8293510.0,
|
||
|
|
"reward": 0.0785440132021904,
|
||
|
|
"reward_std": 0.17606551945209503,
|
||
|
|
"rewards/accuracy_reward_step": 0.03515625,
|
||
|
|
"rewards/final_brier_reward_step": 0.023517362773418427,
|
||
|
|
"rewards/format_reward_step": 0.0703125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.04214133322238922,
|
||
|
|
"step": 33
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.33984375,
|
||
|
|
"calib/auroc": 0.45524296675191817,
|
||
|
|
"calib/avg_num_step_conf": 0.83984375,
|
||
|
|
"calib/ece": 0.5635978835978837,
|
||
|
|
"calib/final_conf_rate": 0.24609375,
|
||
|
|
"calib/format_rate": 0.1484375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.4603174603174603,
|
||
|
|
"calib/gap": 0.010822676896845906,
|
||
|
|
"calib/mean_conf": 0.8205291005291003,
|
||
|
|
"calib/mu_c": 0.8284313725490197,
|
||
|
|
"calib/mu_w": 0.8176086956521738,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.24609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.4140625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.26171875,
|
||
|
|
"calib/pce": 0.5571428571428572,
|
||
|
|
"calib/std_conf": 0.22805689932454587,
|
||
|
|
"calib/step_conf_rate": 0.26171875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2979.0,
|
||
|
|
"completions/max_terminated_length": 2979.0,
|
||
|
|
"completions/mean_length": 594.96484375,
|
||
|
|
"completions/mean_terminated_length": 624.2254028320312,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.03626666666666667,
|
||
|
|
"grad_norm": 0.006227437872439623,
|
||
|
|
"learning_rate": 4.611111111111112e-06,
|
||
|
|
"loss": 0.0756,
|
||
|
|
"num_tokens": 8526613.0,
|
||
|
|
"reward": 0.18159806728363037,
|
||
|
|
"reward_std": 0.33083945512771606,
|
||
|
|
"rewards/accuracy_reward_step": 0.078125,
|
||
|
|
"rewards/final_brier_reward_step": 0.06646484136581421,
|
||
|
|
"rewards/format_reward_step": 0.1484375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.10283762216567993,
|
||
|
|
"step": 34
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.359375,
|
||
|
|
"calib/auroc": 0.3589409722222222,
|
||
|
|
"calib/avg_num_step_conf": 0.7578125,
|
||
|
|
"calib/ece": 0.578890243902439,
|
||
|
|
"calib/final_conf_rate": 0.3203125,
|
||
|
|
"calib/format_rate": 0.16015625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.2926829268292683,
|
||
|
|
"calib/gap": -0.06477604166666662,
|
||
|
|
"calib/mean_conf": 0.758890243902439,
|
||
|
|
"calib/mu_c": 0.7083333333333334,
|
||
|
|
"calib/mu_w": 0.773109375,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.3203125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.40625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.23046875,
|
||
|
|
"calib/pce": 0.5591341463414634,
|
||
|
|
"calib/std_conf": 0.2720138987705054,
|
||
|
|
"calib/step_conf_rate": 0.23046875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2935.0,
|
||
|
|
"completions/max_terminated_length": 2935.0,
|
||
|
|
"completions/mean_length": 694.01171875,
|
||
|
|
"completions/mean_terminated_length": 722.2235717773438,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 26.0,
|
||
|
|
"epoch": 0.037333333333333336,
|
||
|
|
"grad_norm": 0.004874664358794689,
|
||
|
|
"learning_rate": 4.583333333333333e-06,
|
||
|
|
"loss": 0.0596,
|
||
|
|
"num_tokens": 8789216.0,
|
||
|
|
"reward": 0.1838405728340149,
|
||
|
|
"reward_std": 0.3406270444393158,
|
||
|
|
"rewards/accuracy_reward_step": 0.078125,
|
||
|
|
"rewards/final_brier_reward_step": 0.06637898087501526,
|
||
|
|
"rewards/format_reward_step": 0.16015625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.09791683405637741,
|
||
|
|
"step": 35
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.59375,
|
||
|
|
"calib/auroc": 0.5722870879120879,
|
||
|
|
"calib/avg_num_step_conf": 1.77734375,
|
||
|
|
"calib/ece": 0.5842804878048782,
|
||
|
|
"calib/final_conf_rate": 0.48046875,
|
||
|
|
"calib/format_rate": 0.22265625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.3089430894308943,
|
||
|
|
"calib/gap": 0.012974416208791473,
|
||
|
|
"calib/mean_conf": 0.8021666666666667,
|
||
|
|
"calib/mu_c": 0.811765625,
|
||
|
|
"calib/mu_w": 0.7987912087912086,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.48046875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.6875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.3828125,
|
||
|
|
"calib/pce": 0.5631422764227643,
|
||
|
|
"calib/std_conf": 0.20696175824700988,
|
||
|
|
"calib/step_conf_rate": 0.3828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2872.0,
|
||
|
|
"completions/max_terminated_length": 2872.0,
|
||
|
|
"completions/mean_length": 560.46484375,
|
||
|
|
"completions/mean_terminated_length": 569.3611450195312,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 8.0,
|
||
|
|
"epoch": 0.0384,
|
||
|
|
"grad_norm": 0.00626989733427763,
|
||
|
|
"learning_rate": 4.555555555555556e-06,
|
||
|
|
"loss": 0.0187,
|
||
|
|
"num_tokens": 9011087.0,
|
||
|
|
"reward": 0.3228808045387268,
|
||
|
|
"reward_std": 0.4628177881240845,
|
||
|
|
"rewards/accuracy_reward_step": 0.1640625,
|
||
|
|
"rewards/final_brier_reward_step": 0.10468141734600067,
|
||
|
|
"rewards/format_reward_step": 0.22265625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.15872274339199066,
|
||
|
|
"step": 36
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.5390625,
|
||
|
|
"calib/auroc": 0.5586516853932585,
|
||
|
|
"calib/avg_num_step_conf": 1.17578125,
|
||
|
|
"calib/ece": 0.5739002923976608,
|
||
|
|
"calib/final_conf_rate": 0.4453125,
|
||
|
|
"calib/format_rate": 0.26953125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.30701754385964913,
|
||
|
|
"calib/gap": 0.05819513108614238,
|
||
|
|
"calib/mean_conf": 0.7905669590643274,
|
||
|
|
"calib/mu_c": 0.836,
|
||
|
|
"calib/mu_w": 0.7778048689138576,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.4453125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.64453125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.41796875,
|
||
|
|
"calib/pce": 0.5725845029239767,
|
||
|
|
"calib/std_conf": 0.21448465470792466,
|
||
|
|
"calib/step_conf_rate": 0.41796875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2718.0,
|
||
|
|
"completions/max_terminated_length": 2718.0,
|
||
|
|
"completions/mean_length": 530.2578125,
|
||
|
|
"completions/mean_terminated_length": 536.5454711914062,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 10.0,
|
||
|
|
"epoch": 0.039466666666666664,
|
||
|
|
"grad_norm": 0.007074254099279642,
|
||
|
|
"learning_rate": 4.527777777777778e-06,
|
||
|
|
"loss": 0.0148,
|
||
|
|
"num_tokens": 9229609.0,
|
||
|
|
"reward": 0.30819064378738403,
|
||
|
|
"reward_std": 0.42235076427459717,
|
||
|
|
"rewards/accuracy_reward_step": 0.1171875,
|
||
|
|
"rewards/final_brier_reward_step": 0.12816710770130157,
|
||
|
|
"rewards/format_reward_step": 0.26953125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.1842409372329712,
|
||
|
|
"step": 37
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.64453125,
|
||
|
|
"calib/auroc": 0.4983548766157461,
|
||
|
|
"calib/avg_num_step_conf": 1.26953125,
|
||
|
|
"calib/ece": 0.5826423245614034,
|
||
|
|
"calib/final_conf_rate": 0.59375,
|
||
|
|
"calib/format_rate": 0.34765625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.34210526315789475,
|
||
|
|
"calib/gap": 0.00913895025460254,
|
||
|
|
"calib/mean_conf": 0.8129054824561404,
|
||
|
|
"calib/mu_c": 0.8198198198198199,
|
||
|
|
"calib/mu_w": 0.8106808695652173,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.59375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.70703125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.43359375,
|
||
|
|
"calib/pce": 0.5760633771929824,
|
||
|
|
"calib/std_conf": 0.2098056093275935,
|
||
|
|
"calib/step_conf_rate": 0.43359375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3048.0,
|
||
|
|
"completions/max_terminated_length": 3048.0,
|
||
|
|
"completions/mean_length": 525.546875,
|
||
|
|
"completions/mean_terminated_length": 536.0159301757812,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.04053333333333333,
|
||
|
|
"grad_norm": 0.006136655807495117,
|
||
|
|
"learning_rate": 4.5e-06,
|
||
|
|
"loss": 0.0025,
|
||
|
|
"num_tokens": 9446717.0,
|
||
|
|
"reward": 0.4005172848701477,
|
||
|
|
"reward_std": 0.5092767477035522,
|
||
|
|
"rewards/accuracy_reward_step": 0.15625,
|
||
|
|
"rewards/final_brier_reward_step": 0.17309755086898804,
|
||
|
|
"rewards/format_reward_step": 0.34765625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.21368655562400818,
|
||
|
|
"step": 38
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.73046875,
|
||
|
|
"calib/auroc": 0.5160208643815202,
|
||
|
|
"calib/avg_num_step_conf": 1.578125,
|
||
|
|
"calib/ece": 0.5831737634408604,
|
||
|
|
"calib/final_conf_rate": 0.60546875,
|
||
|
|
"calib/format_rate": 0.31640625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.3548387096774194,
|
||
|
|
"calib/gap": 0.035850190428878936,
|
||
|
|
"calib/mean_conf": 0.781883440860215,
|
||
|
|
"calib/mu_c": 0.8101010101010101,
|
||
|
|
"calib/mu_w": 0.7742508196721312,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.60546875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.796875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.4375,
|
||
|
|
"calib/pce": 0.5760769892473119,
|
||
|
|
"calib/std_conf": 0.2403091633074413,
|
||
|
|
"calib/step_conf_rate": 0.4375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2433.0,
|
||
|
|
"completions/max_terminated_length": 2433.0,
|
||
|
|
"completions/mean_length": 528.0625,
|
||
|
|
"completions/mean_terminated_length": 530.1333618164062,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 43.0,
|
||
|
|
"epoch": 0.0416,
|
||
|
|
"grad_norm": 0.00601692171767354,
|
||
|
|
"learning_rate": 4.472222222222223e-06,
|
||
|
|
"loss": -0.02,
|
||
|
|
"num_tokens": 9663669.0,
|
||
|
|
"reward": 0.36516085267066956,
|
||
|
|
"reward_std": 0.4656815230846405,
|
||
|
|
"rewards/accuracy_reward_step": 0.14453125,
|
||
|
|
"rewards/final_brier_reward_step": 0.14888453483581543,
|
||
|
|
"rewards/format_reward_step": 0.31640625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.2050618827342987,
|
||
|
|
"step": 39
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.69140625,
|
||
|
|
"calib/auroc": 0.5190118152524168,
|
||
|
|
"calib/avg_num_step_conf": 2.08203125,
|
||
|
|
"calib/ece": 0.5816113095238096,
|
||
|
|
"calib/final_conf_rate": 0.65625,
|
||
|
|
"calib/format_rate": 0.41015625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.2916666666666667,
|
||
|
|
"calib/gap": 0.010224812030075303,
|
||
|
|
"calib/mean_conf": 0.7725434523809525,
|
||
|
|
"calib/mu_c": 0.7806380952380954,
|
||
|
|
"calib/mu_w": 0.77041328320802,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.65625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.80859375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.55859375,
|
||
|
|
"calib/pce": 0.5729107142857142,
|
||
|
|
"calib/std_conf": 0.2338081530174817,
|
||
|
|
"calib/step_conf_rate": 0.55859375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2571.0,
|
||
|
|
"completions/max_terminated_length": 2571.0,
|
||
|
|
"completions/mean_length": 515.50390625,
|
||
|
|
"completions/mean_terminated_length": 519.56298828125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 24.0,
|
||
|
|
"epoch": 0.042666666666666665,
|
||
|
|
"grad_norm": 0.005652237217873335,
|
||
|
|
"learning_rate": 4.444444444444444e-06,
|
||
|
|
"loss": 0.0625,
|
||
|
|
"num_tokens": 9878078.0,
|
||
|
|
"reward": 0.4190334677696228,
|
||
|
|
"reward_std": 0.45665931701660156,
|
||
|
|
"rewards/accuracy_reward_step": 0.1484375,
|
||
|
|
"rewards/final_brier_reward_step": 0.17388327419757843,
|
||
|
|
"rewards/format_reward_step": 0.41015625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.24242988228797913,
|
||
|
|
"step": 40
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.80078125,
|
||
|
|
"calib/auroc": 0.5799282018794214,
|
||
|
|
"calib/avg_num_step_conf": 1.85546875,
|
||
|
|
"calib/ece": 0.4081126666666667,
|
||
|
|
"calib/final_conf_rate": 0.78125,
|
||
|
|
"calib/format_rate": 0.5,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.225,
|
||
|
|
"calib/gap": 0.07252860310421272,
|
||
|
|
"calib/mean_conf": 0.7779706666666667,
|
||
|
|
"calib/mu_c": 0.8225757575757575,
|
||
|
|
"calib/mu_w": 0.7500471544715448,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.78125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.8671875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.6015625,
|
||
|
|
"calib/pce": 0.4005416666666667,
|
||
|
|
"calib/std_conf": 0.2304185956654637,
|
||
|
|
"calib/step_conf_rate": 0.6015625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3021.0,
|
||
|
|
"completions/max_terminated_length": 3021.0,
|
||
|
|
"completions/mean_length": 468.9921875,
|
||
|
|
"completions/mean_terminated_length": 468.9921875,
|
||
|
|
"completions/min_length": 7.0,
|
||
|
|
"completions/min_terminated_length": 7.0,
|
||
|
|
"epoch": 0.04373333333333333,
|
||
|
|
"grad_norm": 0.006907930597662926,
|
||
|
|
"learning_rate": 4.416666666666667e-06,
|
||
|
|
"loss": 0.071,
|
||
|
|
"num_tokens": 10081068.0,
|
||
|
|
"reward": 0.6845973134040833,
|
||
|
|
"reward_std": 0.6138066053390503,
|
||
|
|
"rewards/accuracy_reward_step": 0.30859375,
|
||
|
|
"rewards/final_brier_reward_step": 0.28266096115112305,
|
||
|
|
"rewards/format_reward_step": 0.5,
|
||
|
|
"rewards/stepwise_brier_reward": 0.3386923670768738,
|
||
|
|
"step": 41
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.875,
|
||
|
|
"calib/auroc": 0.5038810604251254,
|
||
|
|
"calib/avg_num_step_conf": 1.890625,
|
||
|
|
"calib/ece": 0.5712902053712481,
|
||
|
|
"calib/final_conf_rate": 0.82421875,
|
||
|
|
"calib/format_rate": 0.55859375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.2985781990521327,
|
||
|
|
"calib/gap": 0.0056131040522252595,
|
||
|
|
"calib/mean_conf": 0.8066930489731438,
|
||
|
|
"calib/mu_c": 0.8108962264150945,
|
||
|
|
"calib/mu_w": 0.8052831223628693,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.82421875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.91015625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.62890625,
|
||
|
|
"calib/pce": 0.5633992101105845,
|
||
|
|
"calib/std_conf": 0.20060389681325932,
|
||
|
|
"calib/step_conf_rate": 0.62890625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 2966.0,
|
||
|
|
"completions/max_terminated_length": 2966.0,
|
||
|
|
"completions/mean_length": 411.25,
|
||
|
|
"completions/mean_terminated_length": 416.1264953613281,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 59.0,
|
||
|
|
"epoch": 0.0448,
|
||
|
|
"grad_norm": 0.007180627901107073,
|
||
|
|
"learning_rate": 4.388888888888889e-06,
|
||
|
|
"loss": -0.0196,
|
||
|
|
"num_tokens": 10266396.0,
|
||
|
|
"reward": 0.6159687042236328,
|
||
|
|
"reward_std": 0.55170738697052,
|
||
|
|
"rewards/accuracy_reward_step": 0.23046875,
|
||
|
|
"rewards/final_brier_reward_step": 0.25868675112724304,
|
||
|
|
"rewards/format_reward_step": 0.55859375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.35431385040283203,
|
||
|
|
"step": 42
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.875,
|
||
|
|
"calib/auroc": 0.6032567049808429,
|
||
|
|
"calib/avg_num_step_conf": 2.55859375,
|
||
|
|
"calib/ece": 0.5645814307458142,
|
||
|
|
"calib/final_conf_rate": 0.85546875,
|
||
|
|
"calib/format_rate": 0.6171875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.228310502283105,
|
||
|
|
"calib/gap": 0.07952490421455949,
|
||
|
|
"calib/mean_conf": 0.7617047184170472,
|
||
|
|
"calib/mu_c": 0.824888888888889,
|
||
|
|
"calib/mu_w": 0.7453639846743295,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.85546875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.92578125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.7109375,
|
||
|
|
"calib/pce": 0.5604033485540334,
|
||
|
|
"calib/std_conf": 0.22752748859204217,
|
||
|
|
"calib/step_conf_rate": 0.7109375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3039.0,
|
||
|
|
"completions/max_terminated_length": 3039.0,
|
||
|
|
"completions/mean_length": 429.84765625,
|
||
|
|
"completions/mean_terminated_length": 429.84765625,
|
||
|
|
"completions/min_length": 54.0,
|
||
|
|
"completions/min_terminated_length": 54.0,
|
||
|
|
"epoch": 0.04586666666666667,
|
||
|
|
"grad_norm": 0.0070141577161848545,
|
||
|
|
"learning_rate": 4.361111111111112e-06,
|
||
|
|
"loss": 0.0093,
|
||
|
|
"num_tokens": 10457341.0,
|
||
|
|
"reward": 0.6214162707328796,
|
||
|
|
"reward_std": 0.550723671913147,
|
||
|
|
"rewards/accuracy_reward_step": 0.1796875,
|
||
|
|
"rewards/final_brier_reward_step": 0.31000176072120667,
|
||
|
|
"rewards/format_reward_step": 0.6171875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.4062865376472473,
|
||
|
|
"step": 43
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.90625,
|
||
|
|
"calib/auroc": 0.5537568306010929,
|
||
|
|
"calib/avg_num_step_conf": 2.48046875,
|
||
|
|
"calib/ece": 0.578165919282511,
|
||
|
|
"calib/final_conf_rate": 0.87109375,
|
||
|
|
"calib/format_rate": 0.6796875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.24663677130044842,
|
||
|
|
"calib/gap": 0.04479785974499084,
|
||
|
|
"calib/mean_conf": 0.7552959641255605,
|
||
|
|
"calib/mu_c": 0.7920583333333333,
|
||
|
|
"calib/mu_w": 0.7472604735883425,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.87109375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9453125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.7578125,
|
||
|
|
"calib/pce": 0.5770448430493272,
|
||
|
|
"calib/std_conf": 0.2368102410848513,
|
||
|
|
"calib/step_conf_rate": 0.7578125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2910.0,
|
||
|
|
"completions/max_terminated_length": 2910.0,
|
||
|
|
"completions/mean_length": 447.55078125,
|
||
|
|
"completions/mean_terminated_length": 449.305908203125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 53.0,
|
||
|
|
"epoch": 0.046933333333333334,
|
||
|
|
"grad_norm": 0.006897160783410072,
|
||
|
|
"learning_rate": 4.333333333333334e-06,
|
||
|
|
"loss": 0.0952,
|
||
|
|
"num_tokens": 10653914.0,
|
||
|
|
"reward": 0.6419980525970459,
|
||
|
|
"reward_std": 0.4891486167907715,
|
||
|
|
"rewards/accuracy_reward_step": 0.171875,
|
||
|
|
"rewards/final_brier_reward_step": 0.324765682220459,
|
||
|
|
"rewards/format_reward_step": 0.6796875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.41533565521240234,
|
||
|
|
"step": 44
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.8984375,
|
||
|
|
"calib/auroc": 0.5399687395333259,
|
||
|
|
"calib/avg_num_step_conf": 2.99609375,
|
||
|
|
"calib/ece": 0.5009142177177177,
|
||
|
|
"calib/final_conf_rate": 0.8671875,
|
||
|
|
"calib/format_rate": 0.77734375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.22972972972972974,
|
||
|
|
"calib/gap": 0.05840739604034073,
|
||
|
|
"calib/mean_conf": 0.7182419384384385,
|
||
|
|
"calib/mu_c": 0.7627052264150942,
|
||
|
|
"calib/mu_w": 0.7042978303747535,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.8671875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.94921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.859375,
|
||
|
|
"calib/pce": 0.4902087087087087,
|
||
|
|
"calib/std_conf": 0.27369767059658495,
|
||
|
|
"calib/step_conf_rate": 0.859375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3039.0,
|
||
|
|
"completions/max_terminated_length": 3039.0,
|
||
|
|
"completions/mean_length": 435.97265625,
|
||
|
|
"completions/mean_terminated_length": 435.97265625,
|
||
|
|
"completions/min_length": 8.0,
|
||
|
|
"completions/min_terminated_length": 8.0,
|
||
|
|
"epoch": 0.048,
|
||
|
|
"grad_norm": 0.006634102202951908,
|
||
|
|
"learning_rate": 4.305555555555556e-06,
|
||
|
|
"loss": 0.0477,
|
||
|
|
"num_tokens": 10846251.0,
|
||
|
|
"reward": 0.7824665904045105,
|
||
|
|
"reward_std": 0.5265895128250122,
|
||
|
|
"rewards/accuracy_reward_step": 0.2109375,
|
||
|
|
"rewards/final_brier_reward_step": 0.4158342778682709,
|
||
|
|
"rewards/format_reward_step": 0.77734375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.5216355323791504,
|
||
|
|
"step": 45
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9296875,
|
||
|
|
"calib/auroc": 0.4556926210955182,
|
||
|
|
"calib/avg_num_step_conf": 3.5,
|
||
|
|
"calib/ece": 0.5328596364235632,
|
||
|
|
"calib/final_conf_rate": 0.91796875,
|
||
|
|
"calib/format_rate": 0.7890625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.16170212765957448,
|
||
|
|
"calib/gap": -0.03243677957200686,
|
||
|
|
"calib/mean_conf": 0.7069770832320738,
|
||
|
|
"calib/mu_c": 0.6810276595744682,
|
||
|
|
"calib/mu_w": 0.7134644391464751,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.91796875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.85546875,
|
||
|
|
"calib/pce": 0.5199183598278184,
|
||
|
|
"calib/std_conf": 0.26160604105809415,
|
||
|
|
"calib/step_conf_rate": 0.85546875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2550.0,
|
||
|
|
"completions/max_terminated_length": 2550.0,
|
||
|
|
"completions/mean_length": 444.10546875,
|
||
|
|
"completions/mean_terminated_length": 444.10546875,
|
||
|
|
"completions/min_length": 44.0,
|
||
|
|
"completions/min_terminated_length": 44.0,
|
||
|
|
"epoch": 0.04906666666666667,
|
||
|
|
"grad_norm": 0.006138576194643974,
|
||
|
|
"learning_rate": 4.277777777777778e-06,
|
||
|
|
"loss": 0.0768,
|
||
|
|
"num_tokens": 11040390.0,
|
||
|
|
"reward": 0.7579419016838074,
|
||
|
|
"reward_std": 0.4797509014606476,
|
||
|
|
"rewards/accuracy_reward_step": 0.19140625,
|
||
|
|
"rewards/final_brier_reward_step": 0.39381933212280273,
|
||
|
|
"rewards/format_reward_step": 0.7890625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.5316290259361267,
|
||
|
|
"step": 46
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9453125,
|
||
|
|
"calib/auroc": 0.534437146506112,
|
||
|
|
"calib/avg_num_step_conf": 3.515625,
|
||
|
|
"calib/ece": 0.516381856538678,
|
||
|
|
"calib/final_conf_rate": 0.92578125,
|
||
|
|
"calib/format_rate": 0.8359375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.22784810126582278,
|
||
|
|
"calib/gap": 0.010183588755883877,
|
||
|
|
"calib/mean_conf": 0.7540049226427566,
|
||
|
|
"calib/mu_c": 0.7614814814761903,
|
||
|
|
"calib/mu_w": 0.7512978927203064,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.92578125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.984375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.90625,
|
||
|
|
"calib/pce": 0.502281997185654,
|
||
|
|
"calib/std_conf": 0.22681493785340529,
|
||
|
|
"calib/step_conf_rate": 0.90625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2132.0,
|
||
|
|
"completions/max_terminated_length": 2132.0,
|
||
|
|
"completions/mean_length": 427.46484375,
|
||
|
|
"completions/mean_terminated_length": 429.1412048339844,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 58.0,
|
||
|
|
"epoch": 0.050133333333333335,
|
||
|
|
"grad_norm": 0.006385274697095156,
|
||
|
|
"learning_rate": 4.25e-06,
|
||
|
|
"loss": 0.044,
|
||
|
|
"num_tokens": 11231477.0,
|
||
|
|
"reward": 0.8602737188339233,
|
||
|
|
"reward_std": 0.5443920493125916,
|
||
|
|
"rewards/accuracy_reward_step": 0.24609375,
|
||
|
|
"rewards/final_brier_reward_step": 0.4434400200843811,
|
||
|
|
"rewards/format_reward_step": 0.8359375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.5667149424552917,
|
||
|
|
"step": 47
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.953125,
|
||
|
|
"calib/auroc": 0.5800638913846461,
|
||
|
|
"calib/avg_num_step_conf": 3.41015625,
|
||
|
|
"calib/ece": 0.4818471074380165,
|
||
|
|
"calib/final_conf_rate": 0.9453125,
|
||
|
|
"calib/format_rate": 0.85546875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.1446280991735537,
|
||
|
|
"calib/gap": 0.08433752620545076,
|
||
|
|
"calib/mean_conf": 0.7003595041322315,
|
||
|
|
"calib/mu_c": 0.7662264150943398,
|
||
|
|
"calib/mu_w": 0.681888888888889,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9453125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9296875,
|
||
|
|
"calib/pce": 0.48159917355371895,
|
||
|
|
"calib/std_conf": 0.25700948404487006,
|
||
|
|
"calib/step_conf_rate": 0.9296875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1335.0,
|
||
|
|
"completions/max_terminated_length": 1335.0,
|
||
|
|
"completions/mean_length": 395.3203125,
|
||
|
|
"completions/mean_terminated_length": 396.87060546875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 42.0,
|
||
|
|
"epoch": 0.0512,
|
||
|
|
"grad_norm": 0.007040057796984911,
|
||
|
|
"learning_rate": 4.222222222222223e-06,
|
||
|
|
"loss": 0.0314,
|
||
|
|
"num_tokens": 11412047.0,
|
||
|
|
"reward": 0.8526144027709961,
|
||
|
|
"reward_std": 0.5278269648551941,
|
||
|
|
"rewards/accuracy_reward_step": 0.2109375,
|
||
|
|
"rewards/final_brier_reward_step": 0.47880131006240845,
|
||
|
|
"rewards/format_reward_step": 0.85546875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.5825426578521729,
|
||
|
|
"step": 48
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.94921875,
|
||
|
|
"calib/auroc": 0.567470664928292,
|
||
|
|
"calib/avg_num_step_conf": 3.8203125,
|
||
|
|
"calib/ece": 0.4378973886583678,
|
||
|
|
"calib/final_conf_rate": 0.94140625,
|
||
|
|
"calib/format_rate": 0.84375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0995850622406639,
|
||
|
|
"calib/gap": 0.05385923586018515,
|
||
|
|
"calib/mean_conf": 0.6501413388658368,
|
||
|
|
"calib/mu_c": 0.6908151186440679,
|
||
|
|
"calib/mu_w": 0.6369558827838827,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.94140625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.984375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9296875,
|
||
|
|
"calib/pce": 0.42161272475795286,
|
||
|
|
"calib/std_conf": 0.26861273252570295,
|
||
|
|
"calib/step_conf_rate": 0.9296875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2198.0,
|
||
|
|
"completions/max_terminated_length": 2198.0,
|
||
|
|
"completions/mean_length": 411.94921875,
|
||
|
|
"completions/mean_terminated_length": 411.94921875,
|
||
|
|
"completions/min_length": 2.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.05226666666666667,
|
||
|
|
"grad_norm": 0.006596356630325317,
|
||
|
|
"learning_rate": 4.194444444444445e-06,
|
||
|
|
"loss": 0.0017,
|
||
|
|
"num_tokens": 11597722.0,
|
||
|
|
"reward": 0.8926715850830078,
|
||
|
|
"reward_std": 0.5276837348937988,
|
||
|
|
"rewards/accuracy_reward_step": 0.2421875,
|
||
|
|
"rewards/final_brier_reward_step": 0.5090051293373108,
|
||
|
|
"rewards/format_reward_step": 0.84375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.5714260339736938,
|
||
|
|
"step": 49
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.953125,
|
||
|
|
"calib/auroc": 0.5872340425531916,
|
||
|
|
"calib/avg_num_step_conf": 3.265625,
|
||
|
|
"calib/ece": 0.4700798898071625,
|
||
|
|
"calib/final_conf_rate": 0.9453125,
|
||
|
|
"calib/format_rate": 0.89453125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.10330578512396695,
|
||
|
|
"calib/gap": 0.07569354428077846,
|
||
|
|
"calib/mean_conf": 0.6575179063360881,
|
||
|
|
"calib/mu_c": 0.7185106382978724,
|
||
|
|
"calib/mu_w": 0.642817094017094,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9453125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.984375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.953125,
|
||
|
|
"calib/pce": 0.4666914600550964,
|
||
|
|
"calib/std_conf": 0.26108166766519925,
|
||
|
|
"calib/step_conf_rate": 0.953125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1155.0,
|
||
|
|
"completions/max_terminated_length": 1155.0,
|
||
|
|
"completions/mean_length": 442.37109375,
|
||
|
|
"completions/mean_terminated_length": 444.10589599609375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 45.0,
|
||
|
|
"epoch": 0.05333333333333334,
|
||
|
|
"grad_norm": 0.006352792028337717,
|
||
|
|
"learning_rate": 4.166666666666667e-06,
|
||
|
|
"loss": -0.049,
|
||
|
|
"num_tokens": 11792009.0,
|
||
|
|
"reward": 0.8802676200866699,
|
||
|
|
"reward_std": 0.5019233822822571,
|
||
|
|
"rewards/accuracy_reward_step": 0.1875,
|
||
|
|
"rewards/final_brier_reward_step": 0.5297186970710754,
|
||
|
|
"rewards/format_reward_step": 0.89453125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.6381955146789551,
|
||
|
|
"step": 50
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.96875,
|
||
|
|
"calib/auroc": 0.5470850169044282,
|
||
|
|
"calib/avg_num_step_conf": 3.0625,
|
||
|
|
"calib/ece": 0.37765926075268813,
|
||
|
|
"calib/final_conf_rate": 0.96875,
|
||
|
|
"calib/format_rate": 0.91796875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.08064516129032258,
|
||
|
|
"calib/gap": 0.05392048514334402,
|
||
|
|
"calib/mean_conf": 0.6316915188172042,
|
||
|
|
"calib/mu_c": 0.6710447761194029,
|
||
|
|
"calib/mu_w": 0.6171242909760589,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.984375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.953125,
|
||
|
|
"calib/pce": 0.3695947446236559,
|
||
|
|
"calib/std_conf": 0.26770725785027366,
|
||
|
|
"calib/step_conf_rate": 0.953125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2372.0,
|
||
|
|
"completions/max_terminated_length": 2372.0,
|
||
|
|
"completions/mean_length": 450.21875,
|
||
|
|
"completions/mean_terminated_length": 450.21875,
|
||
|
|
"completions/min_length": 55.0,
|
||
|
|
"completions/min_terminated_length": 55.0,
|
||
|
|
"epoch": 0.0544,
|
||
|
|
"grad_norm": 0.005892525427043438,
|
||
|
|
"learning_rate": 4.138888888888889e-06,
|
||
|
|
"loss": 0.018,
|
||
|
|
"num_tokens": 11992241.0,
|
||
|
|
"reward": 0.9921143054962158,
|
||
|
|
"reward_std": 0.4867195188999176,
|
||
|
|
"rewards/accuracy_reward_step": 0.26953125,
|
||
|
|
"rewards/final_brier_reward_step": 0.5696330666542053,
|
||
|
|
"rewards/format_reward_step": 0.91796875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.6495035886764526,
|
||
|
|
"step": 51
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9609375,
|
||
|
|
"calib/auroc": 0.5020485345099275,
|
||
|
|
"calib/avg_num_step_conf": 2.91015625,
|
||
|
|
"calib/ece": 0.3590770941426612,
|
||
|
|
"calib/final_conf_rate": 0.94921875,
|
||
|
|
"calib/format_rate": 0.89453125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.09876543209876543,
|
||
|
|
"calib/gap": 0.001355779917007971,
|
||
|
|
"calib/mean_conf": 0.5949893026474623,
|
||
|
|
"calib/mu_c": 0.5959210526315789,
|
||
|
|
"calib/mu_w": 0.5945652727145709,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.94921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9453125,
|
||
|
|
"calib/pce": 0.3206545975720165,
|
||
|
|
"calib/std_conf": 0.2748466045476414,
|
||
|
|
"calib/step_conf_rate": 0.9453125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2351.0,
|
||
|
|
"completions/max_terminated_length": 2351.0,
|
||
|
|
"completions/mean_length": 460.1640625,
|
||
|
|
"completions/mean_terminated_length": 461.9686584472656,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 103.0,
|
||
|
|
"epoch": 0.055466666666666664,
|
||
|
|
"grad_norm": 0.0061750151216983795,
|
||
|
|
"learning_rate": 4.111111111111111e-06,
|
||
|
|
"loss": 0.0662,
|
||
|
|
"num_tokens": 12193675.0,
|
||
|
|
"reward": 1.0142697095870972,
|
||
|
|
"reward_std": 0.5381784439086914,
|
||
|
|
"rewards/accuracy_reward_step": 0.30078125,
|
||
|
|
"rewards/final_brier_reward_step": 0.5638716220855713,
|
||
|
|
"rewards/format_reward_step": 0.89453125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.6527732610702515,
|
||
|
|
"step": 52
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.5710921401236385,
|
||
|
|
"calib/avg_num_step_conf": 2.70703125,
|
||
|
|
"calib/ece": 0.295949535192563,
|
||
|
|
"calib/final_conf_rate": 0.98046875,
|
||
|
|
"calib/format_rate": 0.953125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.04780876494023904,
|
||
|
|
"calib/gap": 0.06633402021391421,
|
||
|
|
"calib/mean_conf": 0.5518857901726427,
|
||
|
|
"calib/mu_c": 0.5973417721518987,
|
||
|
|
"calib/mu_w": 0.5310077519379844,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.266547144754316,
|
||
|
|
"calib/std_conf": 0.2802968818600421,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1343.0,
|
||
|
|
"completions/max_terminated_length": 1343.0,
|
||
|
|
"completions/mean_length": 441.8515625,
|
||
|
|
"completions/mean_terminated_length": 443.5843505859375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 104.0,
|
||
|
|
"epoch": 0.05653333333333333,
|
||
|
|
"grad_norm": 0.0061105904169380665,
|
||
|
|
"learning_rate": 4.083333333333334e-06,
|
||
|
|
"loss": -0.0193,
|
||
|
|
"num_tokens": 12388293.0,
|
||
|
|
"reward": 1.1016957759857178,
|
||
|
|
"reward_std": 0.4669555127620697,
|
||
|
|
"rewards/accuracy_reward_step": 0.3125,
|
||
|
|
"rewards/final_brier_reward_step": 0.648909330368042,
|
||
|
|
"rewards/format_reward_step": 0.953125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.7152143120765686,
|
||
|
|
"step": 53
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98046875,
|
||
|
|
"calib/auroc": 0.46703296703296704,
|
||
|
|
"calib/avg_num_step_conf": 2.9453125,
|
||
|
|
"calib/ece": 0.3181579301075268,
|
||
|
|
"calib/final_conf_rate": 0.96875,
|
||
|
|
"calib/format_rate": 0.921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.03225806451612903,
|
||
|
|
"calib/gap": -0.03299165908401591,
|
||
|
|
"calib/mean_conf": 0.5291001344086022,
|
||
|
|
"calib/mu_c": 0.5082142857142856,
|
||
|
|
"calib/mu_w": 0.5412059447983015,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.96484375,
|
||
|
|
"calib/pce": 0.2401612903225806,
|
||
|
|
"calib/std_conf": 0.28768419210339424,
|
||
|
|
"calib/step_conf_rate": 0.96484375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1200.0,
|
||
|
|
"completions/max_terminated_length": 1200.0,
|
||
|
|
"completions/mean_length": 398.16015625,
|
||
|
|
"completions/mean_terminated_length": 399.7215881347656,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 2.0,
|
||
|
|
"epoch": 0.0576,
|
||
|
|
"grad_norm": 0.007216614205390215,
|
||
|
|
"learning_rate": 4.055555555555556e-06,
|
||
|
|
"loss": 0.0037,
|
||
|
|
"num_tokens": 12572134.0,
|
||
|
|
"reward": 1.1137436628341675,
|
||
|
|
"reward_std": 0.5061899423599243,
|
||
|
|
"rewards/accuracy_reward_step": 0.36328125,
|
||
|
|
"rewards/final_brier_reward_step": 0.5936059951782227,
|
||
|
|
"rewards/format_reward_step": 0.921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.7083876132965088,
|
||
|
|
"step": 54
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9765625,
|
||
|
|
"calib/auroc": 0.5846567411083541,
|
||
|
|
"calib/avg_num_step_conf": 2.75,
|
||
|
|
"calib/ece": 0.267946015936255,
|
||
|
|
"calib/final_conf_rate": 0.98046875,
|
||
|
|
"calib/format_rate": 0.953125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.01593625498007968,
|
||
|
|
"calib/gap": 0.07323633581472294,
|
||
|
|
"calib/mean_conf": 0.4770838645418327,
|
||
|
|
"calib/mu_c": 0.5313546153846155,
|
||
|
|
"calib/mu_w": 0.45811827956989254,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9765625,
|
||
|
|
"calib/pce": 0.24303286852589645,
|
||
|
|
"calib/std_conf": 0.2663581623790452,
|
||
|
|
"calib/step_conf_rate": 0.9765625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2425.0,
|
||
|
|
"completions/max_terminated_length": 2425.0,
|
||
|
|
"completions/mean_length": 447.69921875,
|
||
|
|
"completions/mean_terminated_length": 447.69921875,
|
||
|
|
"completions/min_length": 54.0,
|
||
|
|
"completions/min_terminated_length": 54.0,
|
||
|
|
"epoch": 0.058666666666666666,
|
||
|
|
"grad_norm": 0.005838231183588505,
|
||
|
|
"learning_rate": 4.027777777777779e-06,
|
||
|
|
"loss": 0.0461,
|
||
|
|
"num_tokens": 12770249.0,
|
||
|
|
"reward": 1.063176155090332,
|
||
|
|
"reward_std": 0.416034460067749,
|
||
|
|
"rewards/accuracy_reward_step": 0.25390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6880024671554565,
|
||
|
|
"rewards/format_reward_step": 0.953125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.7173251509666443,
|
||
|
|
"step": 55
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.984375,
|
||
|
|
"calib/auroc": 0.4855025773195877,
|
||
|
|
"calib/avg_num_step_conf": 2.75,
|
||
|
|
"calib/ece": 0.27800292456666675,
|
||
|
|
"calib/final_conf_rate": 0.9765625,
|
||
|
|
"calib/format_rate": 0.953125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.016,
|
||
|
|
"calib/gap": -0.0017181261720667362,
|
||
|
|
"calib/mean_conf": 0.4046904087666667,
|
||
|
|
"calib/mu_c": 0.40335714285714286,
|
||
|
|
"calib/mu_w": 0.4050752690292096,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98046875,
|
||
|
|
"calib/pce": 0.22934666666666673,
|
||
|
|
"calib/std_conf": 0.2744626342017327,
|
||
|
|
"calib/step_conf_rate": 0.98046875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2200.0,
|
||
|
|
"completions/max_terminated_length": 2200.0,
|
||
|
|
"completions/mean_length": 463.6484375,
|
||
|
|
"completions/mean_terminated_length": 463.6484375,
|
||
|
|
"completions/min_length": 71.0,
|
||
|
|
"completions/min_terminated_length": 71.0,
|
||
|
|
"epoch": 0.05973333333333333,
|
||
|
|
"grad_norm": 0.0065548000857234,
|
||
|
|
"learning_rate": 4.000000000000001e-06,
|
||
|
|
"loss": 0.0268,
|
||
|
|
"num_tokens": 12971463.0,
|
||
|
|
"reward": 1.0270369052886963,
|
||
|
|
"reward_std": 0.3856312036514282,
|
||
|
|
"rewards/accuracy_reward_step": 0.21875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6830077767372131,
|
||
|
|
"rewards/format_reward_step": 0.953125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.7233819961547852,
|
||
|
|
"step": 56
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.96484375,
|
||
|
|
"calib/auroc": 0.5052573529411765,
|
||
|
|
"calib/avg_num_step_conf": 2.5703125,
|
||
|
|
"calib/ece": 0.23630473130666665,
|
||
|
|
"calib/final_conf_rate": 0.9765625,
|
||
|
|
"calib/format_rate": 0.9453125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.012,
|
||
|
|
"calib/gap": 0.002304751921568593,
|
||
|
|
"calib/mean_conf": 0.36149526869333337,
|
||
|
|
"calib/mu_c": 0.36306249999999995,
|
||
|
|
"calib/mu_w": 0.36075774807843136,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.984375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9765625,
|
||
|
|
"calib/pce": 0.13890000000000002,
|
||
|
|
"calib/std_conf": 0.2615626427706827,
|
||
|
|
"calib/step_conf_rate": 0.9765625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2112.0,
|
||
|
|
"completions/max_terminated_length": 2112.0,
|
||
|
|
"completions/mean_length": 439.7578125,
|
||
|
|
"completions/mean_terminated_length": 439.7578125,
|
||
|
|
"completions/min_length": 110.0,
|
||
|
|
"completions/min_terminated_length": 110.0,
|
||
|
|
"epoch": 0.0608,
|
||
|
|
"grad_norm": 0.0059964158572256565,
|
||
|
|
"learning_rate": 3.972222222222223e-06,
|
||
|
|
"loss": 0.0257,
|
||
|
|
"num_tokens": 13166513.0,
|
||
|
|
"reward": 1.1208088397979736,
|
||
|
|
"reward_std": 0.4135684370994568,
|
||
|
|
"rewards/accuracy_reward_step": 0.3125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6733556389808655,
|
||
|
|
"rewards/format_reward_step": 0.9453125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.7521493434906006,
|
||
|
|
"step": 57
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.5710806697108066,
|
||
|
|
"calib/avg_num_step_conf": 2.44921875,
|
||
|
|
"calib/ece": 0.16502314624505926,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.98046875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.015810276679841896,
|
||
|
|
"calib/gap": 0.06396107214611874,
|
||
|
|
"calib/mean_conf": 0.3664393043478261,
|
||
|
|
"calib/mu_c": 0.41194520547945207,
|
||
|
|
"calib/mu_w": 0.34798413333333333,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.12146245059288535,
|
||
|
|
"calib/std_conf": 0.24582607562520056,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2526.0,
|
||
|
|
"completions/max_terminated_length": 2526.0,
|
||
|
|
"completions/mean_length": 470.84375,
|
||
|
|
"completions/mean_terminated_length": 470.84375,
|
||
|
|
"completions/min_length": 97.0,
|
||
|
|
"completions/min_terminated_length": 97.0,
|
||
|
|
"epoch": 0.06186666666666667,
|
||
|
|
"grad_norm": 0.00580226257443428,
|
||
|
|
"learning_rate": 3.944444444444445e-06,
|
||
|
|
"loss": 0.0212,
|
||
|
|
"num_tokens": 13369049.0,
|
||
|
|
"reward": 1.1401081085205078,
|
||
|
|
"reward_std": 0.36763840913772583,
|
||
|
|
"rewards/accuracy_reward_step": 0.28515625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7382141351699829,
|
||
|
|
"rewards/format_reward_step": 0.98046875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.7668165564537048,
|
||
|
|
"step": 58
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.49960278053624635,
|
||
|
|
"calib/avg_num_step_conf": 2.328125,
|
||
|
|
"calib/ece": 0.24439829396325458,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98046875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.007874015748031496,
|
||
|
|
"calib/gap": 0.014792794880282567,
|
||
|
|
"calib/mean_conf": 0.38057152230971125,
|
||
|
|
"calib/mu_c": 0.3898315789473686,
|
||
|
|
"calib/mu_w": 0.375038784067086,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.1254770341207349,
|
||
|
|
"calib/std_conf": 0.2619453349670135,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 1851.0,
|
||
|
|
"completions/max_terminated_length": 1851.0,
|
||
|
|
"completions/mean_length": 454.171875,
|
||
|
|
"completions/mean_terminated_length": 454.171875,
|
||
|
|
"completions/min_length": 111.0,
|
||
|
|
"completions/min_terminated_length": 111.0,
|
||
|
|
"epoch": 0.06293333333333333,
|
||
|
|
"grad_norm": 0.00570636847987771,
|
||
|
|
"learning_rate": 3.916666666666667e-06,
|
||
|
|
"loss": 0.0191,
|
||
|
|
"num_tokens": 13567245.0,
|
||
|
|
"reward": 1.2028120756149292,
|
||
|
|
"reward_std": 0.38464105129241943,
|
||
|
|
"rewards/accuracy_reward_step": 0.37109375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6876335740089417,
|
||
|
|
"rewards/format_reward_step": 0.98046875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.7750437259674072,
|
||
|
|
"step": 59
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.984375,
|
||
|
|
"calib/auroc": 0.5375184516975562,
|
||
|
|
"calib/avg_num_step_conf": 2.109375,
|
||
|
|
"calib/ece": 0.16830111807228917,
|
||
|
|
"calib/final_conf_rate": 0.97265625,
|
||
|
|
"calib/format_rate": 0.96484375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.004016064257028112,
|
||
|
|
"calib/gap": 0.03220580226340819,
|
||
|
|
"calib/mean_conf": 0.3078032995983936,
|
||
|
|
"calib/mu_c": 0.33134328358208953,
|
||
|
|
"calib/mu_w": 0.29913748131868134,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.97265625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.984375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.97265625,
|
||
|
|
"calib/pce": 0.1035140562248996,
|
||
|
|
"calib/std_conf": 0.22954419704442436,
|
||
|
|
"calib/step_conf_rate": 0.97265625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2593.0,
|
||
|
|
"completions/max_terminated_length": 2593.0,
|
||
|
|
"completions/mean_length": 472.59765625,
|
||
|
|
"completions/mean_terminated_length": 474.4510192871094,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 117.0,
|
||
|
|
"epoch": 0.064,
|
||
|
|
"grad_norm": 0.005720623768866062,
|
||
|
|
"learning_rate": 3.88888888888889e-06,
|
||
|
|
"loss": 0.0277,
|
||
|
|
"num_tokens": 13772766.0,
|
||
|
|
"reward": 1.1164627075195312,
|
||
|
|
"reward_std": 0.35981249809265137,
|
||
|
|
"rewards/accuracy_reward_step": 0.26171875,
|
||
|
|
"rewards/final_brier_reward_step": 0.7337855100631714,
|
||
|
|
"rewards/format_reward_step": 0.96484375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.7935928106307983,
|
||
|
|
"step": 60
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.551148761391349,
|
||
|
|
"calib/avg_num_step_conf": 2.51953125,
|
||
|
|
"calib/ece": 0.2384769433465085,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.003952569169960474,
|
||
|
|
"calib/gap": 0.01993505326658962,
|
||
|
|
"calib/mean_conf": 0.3057127799736496,
|
||
|
|
"calib/mu_c": 0.3172955974842767,
|
||
|
|
"calib/mu_w": 0.2973605442176871,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.06260869565217392,
|
||
|
|
"calib/std_conf": 0.23264621599619584,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1869.0,
|
||
|
|
"completions/max_terminated_length": 1869.0,
|
||
|
|
"completions/mean_length": 399.1875,
|
||
|
|
"completions/mean_terminated_length": 400.7529602050781,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 88.0,
|
||
|
|
"epoch": 0.06506666666666666,
|
||
|
|
"grad_norm": 0.006268184166401625,
|
||
|
|
"learning_rate": 3.861111111111112e-06,
|
||
|
|
"loss": 0.0118,
|
||
|
|
"num_tokens": 13954702.0,
|
||
|
|
"reward": 1.254476547241211,
|
||
|
|
"reward_std": 0.3372269570827484,
|
||
|
|
"rewards/accuracy_reward_step": 0.4140625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6911241412162781,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.7934705018997192,
|
||
|
|
"step": 61
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.984375,
|
||
|
|
"calib/auroc": 0.5436766623207302,
|
||
|
|
"calib/avg_num_step_conf": 1.890625,
|
||
|
|
"calib/ece": 0.18494039215686273,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98046875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.0006246849196001603,
|
||
|
|
"calib/mean_conf": 0.25341254901960786,
|
||
|
|
"calib/mu_c": 0.25384615384615383,
|
||
|
|
"calib/mu_w": 0.25322146892655367,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.06623529411764706,
|
||
|
|
"calib/std_conf": 0.2003308169870002,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1528.0,
|
||
|
|
"completions/max_terminated_length": 1528.0,
|
||
|
|
"completions/mean_length": 439.51171875,
|
||
|
|
"completions/mean_terminated_length": 441.2353210449219,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 148.0,
|
||
|
|
"epoch": 0.06613333333333334,
|
||
|
|
"grad_norm": 0.006458355579525232,
|
||
|
|
"learning_rate": 3.833333333333334e-06,
|
||
|
|
"loss": 0.0026,
|
||
|
|
"num_tokens": 14149977.0,
|
||
|
|
"reward": 1.1732728481292725,
|
||
|
|
"reward_std": 0.33111098408699036,
|
||
|
|
"rewards/accuracy_reward_step": 0.3046875,
|
||
|
|
"rewards/final_brier_reward_step": 0.7326979041099548,
|
||
|
|
"rewards/format_reward_step": 0.98046875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8323830962181091,
|
||
|
|
"step": 62
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.49906191369606007,
|
||
|
|
"calib/avg_num_step_conf": 1.890625,
|
||
|
|
"calib/ece": 0.18821015936254984,
|
||
|
|
"calib/final_conf_rate": 0.98046875,
|
||
|
|
"calib/format_rate": 0.9765625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": -0.001516434550440121,
|
||
|
|
"calib/mean_conf": 0.2428655378486056,
|
||
|
|
"calib/mu_c": 0.241844512195122,
|
||
|
|
"calib/mu_w": 0.24336094674556213,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.05219123505976098,
|
||
|
|
"calib/std_conf": 0.1834396838939196,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2843.0,
|
||
|
|
"completions/max_terminated_length": 2843.0,
|
||
|
|
"completions/mean_length": 458.81640625,
|
||
|
|
"completions/mean_terminated_length": 458.81640625,
|
||
|
|
"completions/min_length": 114.0,
|
||
|
|
"completions/min_terminated_length": 114.0,
|
||
|
|
"epoch": 0.0672,
|
||
|
|
"grad_norm": 0.006227673031389713,
|
||
|
|
"learning_rate": 3.8055555555555556e-06,
|
||
|
|
"loss": 0.04,
|
||
|
|
"num_tokens": 14351754.0,
|
||
|
|
"reward": 1.1826722621917725,
|
||
|
|
"reward_std": 0.35445255041122437,
|
||
|
|
"rewards/accuracy_reward_step": 0.3203125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7207086682319641,
|
||
|
|
"rewards/format_reward_step": 0.9765625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8361467719078064,
|
||
|
|
"step": 63
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98046875,
|
||
|
|
"calib/auroc": 0.5357548999933106,
|
||
|
|
"calib/avg_num_step_conf": 1.9453125,
|
||
|
|
"calib/ece": 0.21857719999999997,
|
||
|
|
"calib/final_conf_rate": 0.9765625,
|
||
|
|
"calib/format_rate": 0.95703125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.026542624924744185,
|
||
|
|
"calib/mean_conf": 0.1894228,
|
||
|
|
"calib/mu_c": 0.20545454545454553,
|
||
|
|
"calib/mu_w": 0.17891192052980134,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9765625,
|
||
|
|
"calib/pce": 0.006000000000000001,
|
||
|
|
"calib/std_conf": 0.15477011339454397,
|
||
|
|
"calib/step_conf_rate": 0.9765625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2297.0,
|
||
|
|
"completions/max_terminated_length": 2297.0,
|
||
|
|
"completions/mean_length": 452.79296875,
|
||
|
|
"completions/mean_terminated_length": 454.56866455078125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 148.0,
|
||
|
|
"epoch": 0.06826666666666667,
|
||
|
|
"grad_norm": 0.006078002508729696,
|
||
|
|
"learning_rate": 3.777777777777778e-06,
|
||
|
|
"loss": 0.0262,
|
||
|
|
"num_tokens": 14547125.0,
|
||
|
|
"reward": 1.216360330581665,
|
||
|
|
"reward_std": 0.344184935092926,
|
||
|
|
"rewards/accuracy_reward_step": 0.390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.672036349773407,
|
||
|
|
"rewards/format_reward_step": 0.95703125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8104315400123596,
|
||
|
|
"step": 64
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.5056730769230768,
|
||
|
|
"calib/avg_num_step_conf": 2.03125,
|
||
|
|
"calib/ece": 0.274,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": -0.015825641025641074,
|
||
|
|
"calib/mean_conf": 0.19084375000000003,
|
||
|
|
"calib/mu_c": 0.1812,
|
||
|
|
"calib/mu_w": 0.19702564102564107,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.037109375,
|
||
|
|
"calib/std_conf": 0.150726933221762,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1287.0,
|
||
|
|
"completions/max_terminated_length": 1287.0,
|
||
|
|
"completions/mean_length": 382.0625,
|
||
|
|
"completions/mean_terminated_length": 383.5608215332031,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 152.0,
|
||
|
|
"epoch": 0.06933333333333333,
|
||
|
|
"grad_norm": 0.006557927932590246,
|
||
|
|
"learning_rate": 3.7500000000000005e-06,
|
||
|
|
"loss": 0.0233,
|
||
|
|
"num_tokens": 14725637.0,
|
||
|
|
"reward": 1.2358492612838745,
|
||
|
|
"reward_std": 0.24881812930107117,
|
||
|
|
"rewards/accuracy_reward_step": 0.390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6803131699562073,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.834333062171936,
|
||
|
|
"step": 65
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.581155741376736,
|
||
|
|
"calib/avg_num_step_conf": 1.6953125,
|
||
|
|
"calib/ece": 0.15649803921568628,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03742593698671051,
|
||
|
|
"calib/mean_conf": 0.16428627450980393,
|
||
|
|
"calib/mu_c": 0.1908513513513514,
|
||
|
|
"calib/mu_w": 0.15342541436464088,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.015294117647058828,
|
||
|
|
"calib/std_conf": 0.1362584640574566,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1797.0,
|
||
|
|
"completions/max_terminated_length": 1797.0,
|
||
|
|
"completions/mean_length": 459.1484375,
|
||
|
|
"completions/mean_terminated_length": 460.94903564453125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 141.0,
|
||
|
|
"epoch": 0.0704,
|
||
|
|
"grad_norm": 0.006001452449709177,
|
||
|
|
"learning_rate": 3.7222222222222225e-06,
|
||
|
|
"loss": 0.023,
|
||
|
|
"num_tokens": 14925211.0,
|
||
|
|
"reward": 1.181628942489624,
|
||
|
|
"reward_std": 0.22775799036026,
|
||
|
|
"rewards/accuracy_reward_step": 0.2890625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7677407264709473,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8488467931747437,
|
||
|
|
"step": 66
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5434293429342935,
|
||
|
|
"calib/avg_num_step_conf": 1.84375,
|
||
|
|
"calib/ece": 0.26070901960784315,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.026889533239038255,
|
||
|
|
"calib/mean_conf": 0.16203607843137255,
|
||
|
|
"calib/mu_c": 0.17827524752475254,
|
||
|
|
"calib/mu_w": 0.1513857142857143,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.013333333333333332,
|
||
|
|
"calib/std_conf": 0.1439163709784321,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1576.0,
|
||
|
|
"completions/max_terminated_length": 1576.0,
|
||
|
|
"completions/mean_length": 422.50390625,
|
||
|
|
"completions/mean_terminated_length": 424.1607971191406,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 160.0,
|
||
|
|
"epoch": 0.07146666666666666,
|
||
|
|
"grad_norm": 0.005936103407293558,
|
||
|
|
"learning_rate": 3.694444444444445e-06,
|
||
|
|
"loss": 0.0276,
|
||
|
|
"num_tokens": 15114060.0,
|
||
|
|
"reward": 1.252760410308838,
|
||
|
|
"reward_std": 0.26562878489494324,
|
||
|
|
"rewards/accuracy_reward_step": 0.39453125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6947064399719238,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.852878749370575,
|
||
|
|
"step": 67
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.583870101986044,
|
||
|
|
"calib/avg_num_step_conf": 1.57421875,
|
||
|
|
"calib/ece": 0.22806299212598427,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98046875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.02360815888352119,
|
||
|
|
"calib/mean_conf": 0.15146456692913388,
|
||
|
|
"calib/mu_c": 0.16652173913043478,
|
||
|
|
"calib/mu_w": 0.14291358024691359,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.008661417322834646,
|
||
|
|
"calib/std_conf": 0.11778331869080748,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1164.0,
|
||
|
|
"completions/max_terminated_length": 1164.0,
|
||
|
|
"completions/mean_length": 406.69140625,
|
||
|
|
"completions/mean_terminated_length": 408.2862854003906,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 130.0,
|
||
|
|
"epoch": 0.07253333333333334,
|
||
|
|
"grad_norm": 0.006685419473797083,
|
||
|
|
"learning_rate": 3.6666666666666666e-06,
|
||
|
|
"loss": 0.0119,
|
||
|
|
"num_tokens": 15297941.0,
|
||
|
|
"reward": 1.2206246852874756,
|
||
|
|
"reward_std": 0.2626305818557739,
|
||
|
|
"rewards/accuracy_reward_step": 0.359375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7107387781143188,
|
||
|
|
"rewards/format_reward_step": 0.98046875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8469581007957458,
|
||
|
|
"step": 68
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.5575284090909091,
|
||
|
|
"calib/avg_num_step_conf": 1.46875,
|
||
|
|
"calib/ece": 0.17670634920634926,
|
||
|
|
"calib/final_conf_rate": 0.984375,
|
||
|
|
"calib/format_rate": 0.96875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.021387559808612455,
|
||
|
|
"calib/mean_conf": 0.13440476190476192,
|
||
|
|
"calib/mu_c": 0.14934210526315791,
|
||
|
|
"calib/mu_w": 0.12795454545454546,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.004761904761904762,
|
||
|
|
"calib/std_conf": 0.08806136736258817,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2975.0,
|
||
|
|
"completions/max_terminated_length": 2975.0,
|
||
|
|
"completions/mean_length": 441.32421875,
|
||
|
|
"completions/mean_terminated_length": 443.054931640625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 113.0,
|
||
|
|
"epoch": 0.0736,
|
||
|
|
"grad_norm": 0.005892947781831026,
|
||
|
|
"learning_rate": 3.638888888888889e-06,
|
||
|
|
"loss": 0.0477,
|
||
|
|
"num_tokens": 15491096.0,
|
||
|
|
"reward": 1.16652512550354,
|
||
|
|
"reward_std": 0.2544803023338318,
|
||
|
|
"rewards/accuracy_reward_step": 0.296875,
|
||
|
|
"rewards/final_brier_reward_step": 0.7387734651565552,
|
||
|
|
"rewards/format_reward_step": 0.96875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8385534882545471,
|
||
|
|
"step": 69
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5252142857142857,
|
||
|
|
"calib/avg_num_step_conf": 1.4140625,
|
||
|
|
"calib/ece": 0.19878039215686275,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.01270500000000005,
|
||
|
|
"calib/mean_conf": 0.13690588235294118,
|
||
|
|
"calib/mu_c": 0.14562500000000003,
|
||
|
|
"calib/mu_w": 0.13291999999999998,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.010980392156862747,
|
||
|
|
"calib/std_conf": 0.10751707098704472,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1628.0,
|
||
|
|
"completions/max_terminated_length": 1628.0,
|
||
|
|
"completions/mean_length": 419.92578125,
|
||
|
|
"completions/mean_terminated_length": 421.57257080078125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 115.0,
|
||
|
|
"epoch": 0.07466666666666667,
|
||
|
|
"grad_norm": 0.006532968487590551,
|
||
|
|
"learning_rate": 3.6111111111111115e-06,
|
||
|
|
"loss": 0.0464,
|
||
|
|
"num_tokens": 15681269.0,
|
||
|
|
"reward": 1.207397699356079,
|
||
|
|
"reward_std": 0.19588616490364075,
|
||
|
|
"rewards/accuracy_reward_step": 0.31640625,
|
||
|
|
"rewards/final_brier_reward_step": 0.74442458152771,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8798038959503174,
|
||
|
|
"step": 70
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6059593428014481,
|
||
|
|
"calib/avg_num_step_conf": 1.43359375,
|
||
|
|
"calib/ece": 0.21764705882352942,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.025062656641603995,
|
||
|
|
"calib/mean_conf": 0.11176470588235295,
|
||
|
|
"calib/mu_c": 0.12857142857142856,
|
||
|
|
"calib/mu_w": 0.10350877192982456,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.05600790848754119,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2506.0,
|
||
|
|
"completions/max_terminated_length": 2506.0,
|
||
|
|
"completions/mean_length": 443.25390625,
|
||
|
|
"completions/mean_terminated_length": 443.25390625,
|
||
|
|
"completions/min_length": 132.0,
|
||
|
|
"completions/min_terminated_length": 132.0,
|
||
|
|
"epoch": 0.07573333333333333,
|
||
|
|
"grad_norm": 0.006109386682510376,
|
||
|
|
"learning_rate": 3.5833333333333335e-06,
|
||
|
|
"loss": 0.0309,
|
||
|
|
"num_tokens": 15874830.0,
|
||
|
|
"reward": 1.207634687423706,
|
||
|
|
"reward_std": 0.24222128093242645,
|
||
|
|
"rewards/accuracy_reward_step": 0.328125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7290421724319458,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8740170001983643,
|
||
|
|
"step": 71
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5144973544973545,
|
||
|
|
"calib/avg_num_step_conf": 1.3046875,
|
||
|
|
"calib/ece": 0.21828125,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": -0.0014476190476190726,
|
||
|
|
"calib/mean_conf": 0.11765625000000002,
|
||
|
|
"calib/mu_c": 0.11666666666666665,
|
||
|
|
"calib/mu_w": 0.11811428571428573,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.009765624999999997,
|
||
|
|
"calib/std_conf": 0.07070498098392715,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1094.0,
|
||
|
|
"completions/max_terminated_length": 1094.0,
|
||
|
|
"completions/mean_length": 368.89453125,
|
||
|
|
"completions/mean_terminated_length": 370.3411865234375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 134.0,
|
||
|
|
"epoch": 0.0768,
|
||
|
|
"grad_norm": 0.006662310566753149,
|
||
|
|
"learning_rate": 3.555555555555556e-06,
|
||
|
|
"loss": -0.0097,
|
||
|
|
"num_tokens": 16049355.0,
|
||
|
|
"reward": 1.205643892288208,
|
||
|
|
"reward_std": 0.22846034169197083,
|
||
|
|
"rewards/accuracy_reward_step": 0.31640625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7347124814987183,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8922134637832642,
|
||
|
|
"step": 72
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5062370062370062,
|
||
|
|
"calib/avg_num_step_conf": 1.1796875,
|
||
|
|
"calib/ece": 0.32673228346456684,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": -0.0012770112770112696,
|
||
|
|
"calib/mean_conf": 0.1165748031496063,
|
||
|
|
"calib/mu_c": 0.11585585585585587,
|
||
|
|
"calib/mu_w": 0.11713286713286714,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0031496062992125966,
|
||
|
|
"calib/std_conf": 0.06897220936240427,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 945.0,
|
||
|
|
"completions/max_terminated_length": 945.0,
|
||
|
|
"completions/mean_length": 378.60546875,
|
||
|
|
"completions/mean_terminated_length": 380.0902099609375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 117.0,
|
||
|
|
"epoch": 0.07786666666666667,
|
||
|
|
"grad_norm": 0.006181009113788605,
|
||
|
|
"learning_rate": 3.5277777777777784e-06,
|
||
|
|
"loss": -0.0322,
|
||
|
|
"num_tokens": 16228990.0,
|
||
|
|
"reward": 1.2618048191070557,
|
||
|
|
"reward_std": 0.19632968306541443,
|
||
|
|
"rewards/accuracy_reward_step": 0.43359375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6369625329971313,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8529819250106812,
|
||
|
|
"step": 73
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5107733175914994,
|
||
|
|
"calib/avg_num_step_conf": 1.2890625,
|
||
|
|
"calib/ece": 0.1924466403162055,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.003162337662337658,
|
||
|
|
"calib/mean_conf": 0.11664426877470357,
|
||
|
|
"calib/mu_c": 0.11884415584415584,
|
||
|
|
"calib/mu_w": 0.11568181818181818,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0023715415019762843,
|
||
|
|
"calib/std_conf": 0.06966356145946241,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1589.0,
|
||
|
|
"completions/max_terminated_length": 1589.0,
|
||
|
|
"completions/mean_length": 369.80859375,
|
||
|
|
"completions/mean_terminated_length": 371.25885009765625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 86.0,
|
||
|
|
"epoch": 0.07893333333333333,
|
||
|
|
"grad_norm": 0.006986550521105528,
|
||
|
|
"learning_rate": 3.5e-06,
|
||
|
|
"loss": -0.0384,
|
||
|
|
"num_tokens": 16403269.0,
|
||
|
|
"reward": 1.1860867738723755,
|
||
|
|
"reward_std": 0.19439929723739624,
|
||
|
|
"rewards/accuracy_reward_step": 0.3046875,
|
||
|
|
"rewards/final_brier_reward_step": 0.7368823885917664,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8705822229385376,
|
||
|
|
"step": 74
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5315934065934066,
|
||
|
|
"calib/avg_num_step_conf": 1.39453125,
|
||
|
|
"calib/ece": 0.39097656249999996,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.0032564102564102693,
|
||
|
|
"calib/mean_conf": 0.10667968750000001,
|
||
|
|
"calib/mu_c": 0.10833333333333335,
|
||
|
|
"calib/mu_w": 0.10507692307692308,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.002734375,
|
||
|
|
"calib/std_conf": 0.05525783790470221,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1048.0,
|
||
|
|
"completions/max_terminated_length": 1048.0,
|
||
|
|
"completions/mean_length": 355.7109375,
|
||
|
|
"completions/mean_terminated_length": 357.10589599609375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 119.0,
|
||
|
|
"epoch": 0.08,
|
||
|
|
"grad_norm": 0.006790259387344122,
|
||
|
|
"learning_rate": 3.4722222222222224e-06,
|
||
|
|
"loss": 0.0141,
|
||
|
|
"num_tokens": 16574763.0,
|
||
|
|
"reward": 1.2971038818359375,
|
||
|
|
"reward_std": 0.21821537613868713,
|
||
|
|
"rewards/accuracy_reward_step": 0.4921875,
|
||
|
|
"rewards/final_brier_reward_step": 0.600019097328186,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8196269869804382,
|
||
|
|
"step": 75
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.490164039471998,
|
||
|
|
"calib/avg_num_step_conf": 1.28125,
|
||
|
|
"calib/ece": 0.29662745098039217,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": -0.0012091503267973869,
|
||
|
|
"calib/mean_conf": 0.10337254901960784,
|
||
|
|
"calib/mu_c": 0.1026470588235294,
|
||
|
|
"calib/mu_w": 0.10385620915032678,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.04954299411674186,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2006.0,
|
||
|
|
"completions/max_terminated_length": 2006.0,
|
||
|
|
"completions/mean_length": 362.609375,
|
||
|
|
"completions/mean_terminated_length": 362.609375,
|
||
|
|
"completions/min_length": 149.0,
|
||
|
|
"completions/min_terminated_length": 149.0,
|
||
|
|
"epoch": 0.08106666666666666,
|
||
|
|
"grad_norm": 0.007501825224608183,
|
||
|
|
"learning_rate": 3.444444444444445e-06,
|
||
|
|
"loss": 0.0288,
|
||
|
|
"num_tokens": 16746327.0,
|
||
|
|
"reward": 1.2401504516601562,
|
||
|
|
"reward_std": 0.22100690007209778,
|
||
|
|
"rewards/accuracy_reward_step": 0.3984375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6580047011375427,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8695923089981079,
|
||
|
|
"step": 76
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.5179235179235179,
|
||
|
|
"calib/avg_num_step_conf": 1.2109375,
|
||
|
|
"calib/ece": 0.3316929133858268,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.0013286713286713436,
|
||
|
|
"calib/mean_conf": 0.10925196850393701,
|
||
|
|
"calib/mu_c": 0.11000000000000001,
|
||
|
|
"calib/mu_w": 0.10867132867132867,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.001968503937007874,
|
||
|
|
"calib/std_conf": 0.06096849173396016,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2688.0,
|
||
|
|
"completions/max_terminated_length": 2688.0,
|
||
|
|
"completions/mean_length": 406.18359375,
|
||
|
|
"completions/mean_terminated_length": 406.18359375,
|
||
|
|
"completions/min_length": 76.0,
|
||
|
|
"completions/min_terminated_length": 76.0,
|
||
|
|
"epoch": 0.08213333333333334,
|
||
|
|
"grad_norm": 0.006342964246869087,
|
||
|
|
"learning_rate": 3.416666666666667e-06,
|
||
|
|
"loss": 0.098,
|
||
|
|
"num_tokens": 16930654.0,
|
||
|
|
"reward": 1.2683703899383545,
|
||
|
|
"reward_std": 0.24026376008987427,
|
||
|
|
"rewards/accuracy_reward_step": 0.43359375,
|
||
|
|
"rewards/final_brier_reward_step": 0.638453483581543,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8715747594833374,
|
||
|
|
"step": 77
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5432713945862534,
|
||
|
|
"calib/avg_num_step_conf": 1.375,
|
||
|
|
"calib/ece": 0.3044094488188976,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.009740886002700455,
|
||
|
|
"calib/mean_conf": 0.1011023622047244,
|
||
|
|
"calib/mu_c": 0.10689320388349513,
|
||
|
|
"calib/mu_w": 0.09715231788079468,
|
||
|
|
"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": 0.046528770990141785,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1416.0,
|
||
|
|
"completions/max_terminated_length": 1416.0,
|
||
|
|
"completions/mean_length": 412.46484375,
|
||
|
|
"completions/mean_terminated_length": 414.0823669433594,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 166.0,
|
||
|
|
"epoch": 0.0832,
|
||
|
|
"grad_norm": 0.007117137312889099,
|
||
|
|
"learning_rate": 3.3888888888888893e-06,
|
||
|
|
"loss": 0.0219,
|
||
|
|
"num_tokens": 17119949.0,
|
||
|
|
"reward": 1.2550793886184692,
|
||
|
|
"reward_std": 0.21007688343524933,
|
||
|
|
"rewards/accuracy_reward_step": 0.40625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6635695695877075,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8775535821914673,
|
||
|
|
"step": 78
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.525512127686314,
|
||
|
|
"calib/avg_num_step_conf": 1.26953125,
|
||
|
|
"calib/ece": 0.3317254901960784,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.0020114364710318022,
|
||
|
|
"calib/mean_conf": 0.10435294117647059,
|
||
|
|
"calib/mu_c": 0.10550458715596331,
|
||
|
|
"calib/mu_w": 0.1034931506849315,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.004313725490196078,
|
||
|
|
"calib/std_conf": 0.04398164658636054,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2057.0,
|
||
|
|
"completions/max_terminated_length": 2057.0,
|
||
|
|
"completions/mean_length": 450.1015625,
|
||
|
|
"completions/mean_terminated_length": 450.1015625,
|
||
|
|
"completions/min_length": 119.0,
|
||
|
|
"completions/min_terminated_length": 119.0,
|
||
|
|
"epoch": 0.08426666666666667,
|
||
|
|
"grad_norm": 0.006018451880663633,
|
||
|
|
"learning_rate": 3.3611111111111117e-06,
|
||
|
|
"loss": 0.0178,
|
||
|
|
"num_tokens": 17317231.0,
|
||
|
|
"reward": 1.2497444152832031,
|
||
|
|
"reward_std": 0.20812439918518066,
|
||
|
|
"rewards/accuracy_reward_step": 0.42578125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6396187543869019,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8306777477264404,
|
||
|
|
"step": 79
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5166295057599405,
|
||
|
|
"calib/avg_num_step_conf": 1.50390625,
|
||
|
|
"calib/ece": 0.35180392156862744,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.006296915644741746,
|
||
|
|
"calib/mean_conf": 0.10701960784313726,
|
||
|
|
"calib/mu_c": 0.11042735042735043,
|
||
|
|
"calib/mu_w": 0.10413043478260868,
|
||
|
|
"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": 0.052121084588386525,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1906.0,
|
||
|
|
"completions/max_terminated_length": 1906.0,
|
||
|
|
"completions/mean_length": 383.921875,
|
||
|
|
"completions/mean_terminated_length": 385.4274597167969,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 113.0,
|
||
|
|
"epoch": 0.08533333333333333,
|
||
|
|
"grad_norm": 0.008160736411809921,
|
||
|
|
"learning_rate": 3.3333333333333333e-06,
|
||
|
|
"loss": -0.0023,
|
||
|
|
"num_tokens": 17493355.0,
|
||
|
|
"reward": 1.2862913608551025,
|
||
|
|
"reward_std": 0.21927902102470398,
|
||
|
|
"rewards/accuracy_reward_step": 0.4609375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6258856058120728,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8543317317962646,
|
||
|
|
"step": 80
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.560911150264284,
|
||
|
|
"calib/avg_num_step_conf": 1.3359375,
|
||
|
|
"calib/ece": 0.34845454545454535,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.012496476214447536,
|
||
|
|
"calib/mean_conf": 0.11004347826086958,
|
||
|
|
"calib/mu_c": 0.11681034482758622,
|
||
|
|
"calib/mu_w": 0.10431386861313868,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.046031460359654164,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2509.0,
|
||
|
|
"completions/max_terminated_length": 2509.0,
|
||
|
|
"completions/mean_length": 449.65234375,
|
||
|
|
"completions/mean_terminated_length": 449.65234375,
|
||
|
|
"completions/min_length": 149.0,
|
||
|
|
"completions/min_terminated_length": 149.0,
|
||
|
|
"epoch": 0.0864,
|
||
|
|
"grad_norm": 0.006808146368712187,
|
||
|
|
"learning_rate": 3.3055555555555558e-06,
|
||
|
|
"loss": 0.0815,
|
||
|
|
"num_tokens": 17690394.0,
|
||
|
|
"reward": 1.2709615230560303,
|
||
|
|
"reward_std": 0.2466251254081726,
|
||
|
|
"rewards/accuracy_reward_step": 0.453125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6269539594650269,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8315008878707886,
|
||
|
|
"step": 81
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5553890384144848,
|
||
|
|
"calib/avg_num_step_conf": 1.23046875,
|
||
|
|
"calib/ece": 0.36363281250000007,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.013037680450208006,
|
||
|
|
"calib/mean_conf": 0.11292968750000004,
|
||
|
|
"calib/mu_c": 0.11975409836065574,
|
||
|
|
"calib/mu_w": 0.10671641791044774,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.043763512840634074,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1228.0,
|
||
|
|
"completions/max_terminated_length": 1228.0,
|
||
|
|
"completions/mean_length": 380.171875,
|
||
|
|
"completions/mean_terminated_length": 381.66278076171875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 154.0,
|
||
|
|
"epoch": 0.08746666666666666,
|
||
|
|
"grad_norm": 0.007882577367126942,
|
||
|
|
"learning_rate": 3.277777777777778e-06,
|
||
|
|
"loss": 0.0379,
|
||
|
|
"num_tokens": 17868950.0,
|
||
|
|
"reward": 1.2923544645309448,
|
||
|
|
"reward_std": 0.18476907908916473,
|
||
|
|
"rewards/accuracy_reward_step": 0.4765625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6229097843170166,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8173484206199646,
|
||
|
|
"step": 82
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.549424005945745,
|
||
|
|
"calib/avg_num_step_conf": 1.2265625,
|
||
|
|
"calib/ece": 0.3451960784313725,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.007748049052396899,
|
||
|
|
"calib/mean_conf": 0.11362745098039216,
|
||
|
|
"calib/mu_c": 0.11782051282051284,
|
||
|
|
"calib/mu_w": 0.11007246376811594,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.04847354677181931,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1152.0,
|
||
|
|
"completions/max_terminated_length": 1152.0,
|
||
|
|
"completions/mean_length": 441.8515625,
|
||
|
|
"completions/mean_terminated_length": 443.5843505859375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 156.0,
|
||
|
|
"epoch": 0.08853333333333334,
|
||
|
|
"grad_norm": 0.00784524530172348,
|
||
|
|
"learning_rate": 3.2500000000000002e-06,
|
||
|
|
"loss": 0.0501,
|
||
|
|
"num_tokens": 18065008.0,
|
||
|
|
"reward": 1.2880065441131592,
|
||
|
|
"reward_std": 0.18693551421165466,
|
||
|
|
"rewards/accuracy_reward_step": 0.4609375,
|
||
|
|
"rewards/final_brier_reward_step": 0.627806544303894,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8620378971099854,
|
||
|
|
"step": 83
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5292688925782452,
|
||
|
|
"calib/avg_num_step_conf": 1.23046875,
|
||
|
|
"calib/ece": 0.33453125000000006,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.008844616614400821,
|
||
|
|
"calib/mean_conf": 0.12562500000000001,
|
||
|
|
"calib/mu_c": 0.13042735042735046,
|
||
|
|
"calib/mu_w": 0.12158273381294964,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0015625,
|
||
|
|
"calib/std_conf": 0.05904513633653496,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1096.0,
|
||
|
|
"completions/max_terminated_length": 1096.0,
|
||
|
|
"completions/mean_length": 392.98046875,
|
||
|
|
"completions/mean_terminated_length": 394.5216064453125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 124.0,
|
||
|
|
"epoch": 0.0896,
|
||
|
|
"grad_norm": 0.00771920895203948,
|
||
|
|
"learning_rate": 3.2222222222222227e-06,
|
||
|
|
"loss": 0.0447,
|
||
|
|
"num_tokens": 18247211.0,
|
||
|
|
"reward": 1.292407512664795,
|
||
|
|
"reward_std": 0.20232713222503662,
|
||
|
|
"rewards/accuracy_reward_step": 0.45703125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6421773433685303,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8618377447128296,
|
||
|
|
"step": 84
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5650152827305146,
|
||
|
|
"calib/avg_num_step_conf": 1.30859375,
|
||
|
|
"calib/ece": 0.2864901960784314,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.016713576158940466,
|
||
|
|
"calib/mean_conf": 0.1213529411764706,
|
||
|
|
"calib/mu_c": 0.13125000000000003,
|
||
|
|
"calib/mu_w": 0.11453642384105957,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.051280016950156174,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2860.0,
|
||
|
|
"completions/max_terminated_length": 2860.0,
|
||
|
|
"completions/mean_length": 433.58984375,
|
||
|
|
"completions/mean_terminated_length": 433.58984375,
|
||
|
|
"completions/min_length": 118.0,
|
||
|
|
"completions/min_terminated_length": 118.0,
|
||
|
|
"epoch": 0.09066666666666667,
|
||
|
|
"grad_norm": 0.006601530127227306,
|
||
|
|
"learning_rate": 3.1944444444444443e-06,
|
||
|
|
"loss": 0.0578,
|
||
|
|
"num_tokens": 18441714.0,
|
||
|
|
"reward": 1.26658296585083,
|
||
|
|
"reward_std": 0.20831358432769775,
|
||
|
|
"rewards/accuracy_reward_step": 0.40625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6791960000991821,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8876274824142456,
|
||
|
|
"step": 85
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5501410971161126,
|
||
|
|
"calib/avg_num_step_conf": 1.11328125,
|
||
|
|
"calib/ece": 0.21464566929133858,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.014422190102553495,
|
||
|
|
"calib/mean_conf": 0.12787401574803148,
|
||
|
|
"calib/mu_c": 0.13735632183908045,
|
||
|
|
"calib/mu_w": 0.12293413173652695,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.05324331197155695,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2849.0,
|
||
|
|
"completions/max_terminated_length": 2849.0,
|
||
|
|
"completions/mean_length": 426.640625,
|
||
|
|
"completions/mean_terminated_length": 426.640625,
|
||
|
|
"completions/min_length": 102.0,
|
||
|
|
"completions/min_terminated_length": 102.0,
|
||
|
|
"epoch": 0.09173333333333333,
|
||
|
|
"grad_norm": 0.007412768434733152,
|
||
|
|
"learning_rate": 3.1666666666666667e-06,
|
||
|
|
"loss": 0.0749,
|
||
|
|
"num_tokens": 18632126.0,
|
||
|
|
"reward": 1.212226390838623,
|
||
|
|
"reward_std": 0.19229844212532043,
|
||
|
|
"rewards/accuracy_reward_step": 0.33984375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7252601385116577,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8530725240707397,
|
||
|
|
"step": 86
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5480605487228003,
|
||
|
|
"calib/avg_num_step_conf": 1.13671875,
|
||
|
|
"calib/ece": 0.4560546875,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.012075055187637934,
|
||
|
|
"calib/mean_conf": 0.13378906250000003,
|
||
|
|
"calib/mu_c": 0.1387417218543046,
|
||
|
|
"calib/mu_w": 0.12666666666666668,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.05555911158730937,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1123.0,
|
||
|
|
"completions/max_terminated_length": 1123.0,
|
||
|
|
"completions/mean_length": 374.06640625,
|
||
|
|
"completions/mean_terminated_length": 375.5333557128906,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 102.0,
|
||
|
|
"epoch": 0.0928,
|
||
|
|
"grad_norm": 0.01204943098127842,
|
||
|
|
"learning_rate": 3.138888888888889e-06,
|
||
|
|
"loss": 0.0105,
|
||
|
|
"num_tokens": 18809063.0,
|
||
|
|
"reward": 1.3697707653045654,
|
||
|
|
"reward_std": 0.20046617090702057,
|
||
|
|
"rewards/accuracy_reward_step": 0.58984375,
|
||
|
|
"rewards/final_brier_reward_step": 0.5483496189117432,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8323837518692017,
|
||
|
|
"step": 87
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5213235294117646,
|
||
|
|
"calib/avg_num_step_conf": 1.19921875,
|
||
|
|
"calib/ece": 0.3510546875000001,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.005906862745098079,
|
||
|
|
"calib/mean_conf": 0.11769531250000004,
|
||
|
|
"calib/mu_c": 0.12083333333333336,
|
||
|
|
"calib/mu_w": 0.11492647058823528,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.04484087466282681,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2034.0,
|
||
|
|
"completions/max_terminated_length": 2034.0,
|
||
|
|
"completions/mean_length": 450.77734375,
|
||
|
|
"completions/mean_terminated_length": 452.5451354980469,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 150.0,
|
||
|
|
"epoch": 0.09386666666666667,
|
||
|
|
"grad_norm": 0.008023022674024105,
|
||
|
|
"learning_rate": 3.1111111111111116e-06,
|
||
|
|
"loss": 0.0397,
|
||
|
|
"num_tokens": 19009990.0,
|
||
|
|
"reward": 1.2934889793395996,
|
||
|
|
"reward_std": 0.22242587804794312,
|
||
|
|
"rewards/accuracy_reward_step": 0.46875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6248011589050293,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8540412187576294,
|
||
|
|
"step": 88
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5638919995023017,
|
||
|
|
"calib/avg_num_step_conf": 1.046875,
|
||
|
|
"calib/ece": 0.3166274509803922,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.012545726017170605,
|
||
|
|
"calib/mean_conf": 0.13043137254901963,
|
||
|
|
"calib/mu_c": 0.13736842105263158,
|
||
|
|
"calib/mu_w": 0.12482269503546098,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.05720457094788664,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1840.0,
|
||
|
|
"completions/max_terminated_length": 1840.0,
|
||
|
|
"completions/mean_length": 425.8515625,
|
||
|
|
"completions/mean_terminated_length": 427.5216064453125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 127.0,
|
||
|
|
"epoch": 0.09493333333333333,
|
||
|
|
"grad_norm": 0.006762202363461256,
|
||
|
|
"learning_rate": 3.0833333333333336e-06,
|
||
|
|
"loss": -0.0273,
|
||
|
|
"num_tokens": 19203576.0,
|
||
|
|
"reward": 1.286118984222412,
|
||
|
|
"reward_std": 0.17611607909202576,
|
||
|
|
"rewards/accuracy_reward_step": 0.4453125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6529195308685303,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8620744347572327,
|
||
|
|
"step": 89
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5667138287401574,
|
||
|
|
"calib/avg_num_step_conf": 1.1015625,
|
||
|
|
"calib/ece": 0.37058823529411766,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.01433316929133864,
|
||
|
|
"calib/mean_conf": 0.12745098039215685,
|
||
|
|
"calib/mu_c": 0.13464566929133867,
|
||
|
|
"calib/mu_w": 0.12031250000000003,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.05469498802198522,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1697.0,
|
||
|
|
"completions/max_terminated_length": 1697.0,
|
||
|
|
"completions/mean_length": 405.1875,
|
||
|
|
"completions/mean_terminated_length": 406.7764892578125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 123.0,
|
||
|
|
"epoch": 0.096,
|
||
|
|
"grad_norm": 0.008788934908807278,
|
||
|
|
"learning_rate": 3.055555555555556e-06,
|
||
|
|
"loss": 0.0517,
|
||
|
|
"num_tokens": 19386304.0,
|
||
|
|
"reward": 1.3127435445785522,
|
||
|
|
"reward_std": 0.2274431735277176,
|
||
|
|
"rewards/accuracy_reward_step": 0.49609375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6106835603713989,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8546072244644165,
|
||
|
|
"step": 90
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.536080586080586,
|
||
|
|
"calib/avg_num_step_conf": 1.12109375,
|
||
|
|
"calib/ece": 0.364451171875,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.00711459096459105,
|
||
|
|
"calib/mean_conf": 0.12773632812500002,
|
||
|
|
"calib/mu_c": 0.1313492063492064,
|
||
|
|
"calib/mu_w": 0.12423461538461536,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.05600496811627371,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1367.0,
|
||
|
|
"completions/max_terminated_length": 1367.0,
|
||
|
|
"completions/mean_length": 422.421875,
|
||
|
|
"completions/mean_terminated_length": 424.0784606933594,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 167.0,
|
||
|
|
"epoch": 0.09706666666666666,
|
||
|
|
"grad_norm": 0.006352297496050596,
|
||
|
|
"learning_rate": 3.0277777777777776e-06,
|
||
|
|
"loss": -0.0096,
|
||
|
|
"num_tokens": 19577836.0,
|
||
|
|
"reward": 1.3052550554275513,
|
||
|
|
"reward_std": 0.21970926225185394,
|
||
|
|
"rewards/accuracy_reward_step": 0.4921875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6176562309265137,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8169575929641724,
|
||
|
|
"step": 91
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5578652377432864,
|
||
|
|
"calib/avg_num_step_conf": 1.06640625,
|
||
|
|
"calib/ece": 0.35862745098039217,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.013063562453806396,
|
||
|
|
"calib/mean_conf": 0.12372549019607845,
|
||
|
|
"calib/mu_c": 0.1304878048780488,
|
||
|
|
"calib/mu_w": 0.11742424242424242,
|
||
|
|
"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": 0.051978909158510896,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1062.0,
|
||
|
|
"completions/max_terminated_length": 1062.0,
|
||
|
|
"completions/mean_length": 390.27734375,
|
||
|
|
"completions/mean_terminated_length": 391.807861328125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 157.0,
|
||
|
|
"epoch": 0.09813333333333334,
|
||
|
|
"grad_norm": 0.007925019599497318,
|
||
|
|
"learning_rate": 3e-06,
|
||
|
|
"loss": 0.0486,
|
||
|
|
"num_tokens": 19760147.0,
|
||
|
|
"reward": 1.3043065071105957,
|
||
|
|
"reward_std": 0.22197169065475464,
|
||
|
|
"rewards/accuracy_reward_step": 0.48046875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6230762004852295,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8538861870765686,
|
||
|
|
"step": 92
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5310823311748382,
|
||
|
|
"calib/avg_num_step_conf": 1.05859375,
|
||
|
|
"calib/ece": 0.3203125,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.005932778291705243,
|
||
|
|
"calib/mean_conf": 0.12890625000000003,
|
||
|
|
"calib/mu_c": 0.13217391304347828,
|
||
|
|
"calib/mu_w": 0.12624113475177304,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.051390575117792756,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1162.0,
|
||
|
|
"completions/max_terminated_length": 1162.0,
|
||
|
|
"completions/mean_length": 398.0625,
|
||
|
|
"completions/mean_terminated_length": 399.6235656738281,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 112.0,
|
||
|
|
"epoch": 0.0992,
|
||
|
|
"grad_norm": 0.013263871893286705,
|
||
|
|
"learning_rate": 2.9722222222222225e-06,
|
||
|
|
"loss": 0.0339,
|
||
|
|
"num_tokens": 19943507.0,
|
||
|
|
"reward": 1.2797186374664307,
|
||
|
|
"reward_std": 0.2574073076248169,
|
||
|
|
"rewards/accuracy_reward_step": 0.44921875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6425390243530273,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8462965488433838,
|
||
|
|
"step": 93
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5782787893700787,
|
||
|
|
"calib/avg_num_step_conf": 1.05859375,
|
||
|
|
"calib/ece": 0.3829411764705883,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.02063545767716539,
|
||
|
|
"calib/mean_conf": 0.11901960784313724,
|
||
|
|
"calib/mu_c": 0.12929687500000003,
|
||
|
|
"calib/mu_w": 0.10866141732283464,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.05314629834049774,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1234.0,
|
||
|
|
"completions/max_terminated_length": 1234.0,
|
||
|
|
"completions/mean_length": 383.265625,
|
||
|
|
"completions/mean_terminated_length": 384.7686462402344,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 145.0,
|
||
|
|
"epoch": 0.10026666666666667,
|
||
|
|
"grad_norm": 0.010008939541876316,
|
||
|
|
"learning_rate": 2.944444444444445e-06,
|
||
|
|
"loss": -0.0112,
|
||
|
|
"num_tokens": 20125983.0,
|
||
|
|
"reward": 1.3118412494659424,
|
||
|
|
"reward_std": 0.1737186312675476,
|
||
|
|
"rewards/accuracy_reward_step": 0.5,
|
||
|
|
"rewards/final_brier_reward_step": 0.6084668636322021,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8351192474365234,
|
||
|
|
"step": 94
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5236263736263738,
|
||
|
|
"calib/avg_num_step_conf": 1.07421875,
|
||
|
|
"calib/ece": 0.384375,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.005518925518925508,
|
||
|
|
"calib/mean_conf": 0.12343750000000002,
|
||
|
|
"calib/mu_c": 0.12615384615384614,
|
||
|
|
"calib/mu_w": 0.12063492063492064,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.050751193027849105,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1607.0,
|
||
|
|
"completions/max_terminated_length": 1607.0,
|
||
|
|
"completions/mean_length": 394.0546875,
|
||
|
|
"completions/mean_terminated_length": 395.60003662109375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 147.0,
|
||
|
|
"epoch": 0.10133333333333333,
|
||
|
|
"grad_norm": 0.009900784119963646,
|
||
|
|
"learning_rate": 2.916666666666667e-06,
|
||
|
|
"loss": 0.0047,
|
||
|
|
"num_tokens": 20308669.0,
|
||
|
|
"reward": 1.3128812313079834,
|
||
|
|
"reward_std": 0.20290324091911316,
|
||
|
|
"rewards/accuracy_reward_step": 0.5078125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6024999618530273,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8152750730514526,
|
||
|
|
"step": 95
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5637755102040818,
|
||
|
|
"calib/avg_num_step_conf": 1.0390625,
|
||
|
|
"calib/ece": 0.4892578125000001,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.017973391888400927,
|
||
|
|
"calib/mean_conf": 0.1279296875,
|
||
|
|
"calib/mu_c": 0.13481012658227848,
|
||
|
|
"calib/mu_w": 0.11683673469387755,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.056173654244248195,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 965.0,
|
||
|
|
"completions/max_terminated_length": 965.0,
|
||
|
|
"completions/mean_length": 363.50390625,
|
||
|
|
"completions/mean_terminated_length": 364.929443359375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 105.0,
|
||
|
|
"epoch": 0.1024,
|
||
|
|
"grad_norm": 0.01069644931703806,
|
||
|
|
"learning_rate": 2.888888888888889e-06,
|
||
|
|
"loss": 0.0379,
|
||
|
|
"num_tokens": 20483222.0,
|
||
|
|
"reward": 1.3792392015457153,
|
||
|
|
"reward_std": 0.1949031949043274,
|
||
|
|
"rewards/accuracy_reward_step": 0.6171875,
|
||
|
|
"rewards/final_brier_reward_step": 0.5296972990036011,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.7888127565383911,
|
||
|
|
"step": 96
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6126050420168067,
|
||
|
|
"calib/avg_num_step_conf": 1.046875,
|
||
|
|
"calib/ece": 0.40688976377952757,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.02732337379396202,
|
||
|
|
"calib/mean_conf": 0.12460629921259843,
|
||
|
|
"calib/mu_c": 0.1374074074074074,
|
||
|
|
"calib/mu_w": 0.11008403361344538,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.05797469682655482,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1152.0,
|
||
|
|
"completions/max_terminated_length": 1152.0,
|
||
|
|
"completions/mean_length": 379.9140625,
|
||
|
|
"completions/mean_terminated_length": 381.4039306640625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 155.0,
|
||
|
|
"epoch": 0.10346666666666667,
|
||
|
|
"grad_norm": 0.011838745325803757,
|
||
|
|
"learning_rate": 2.861111111111111e-06,
|
||
|
|
"loss": 0.0095,
|
||
|
|
"num_tokens": 20661232.0,
|
||
|
|
"reward": 1.335360050201416,
|
||
|
|
"reward_std": 0.21025191247463226,
|
||
|
|
"rewards/accuracy_reward_step": 0.53125,
|
||
|
|
"rewards/final_brier_reward_step": 0.5910254120826721,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8437644243240356,
|
||
|
|
"step": 97
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.500244140625,
|
||
|
|
"calib/avg_num_step_conf": 1.03515625,
|
||
|
|
"calib/ece": 0.36464843750000003,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": -0.0011718750000000167,
|
||
|
|
"calib/mean_conf": 0.13535156250000002,
|
||
|
|
"calib/mu_c": 0.13476562500000003,
|
||
|
|
"calib/mu_w": 0.13593750000000004,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.05986861785450366,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1440.0,
|
||
|
|
"completions/max_terminated_length": 1440.0,
|
||
|
|
"completions/mean_length": 402.171875,
|
||
|
|
"completions/mean_terminated_length": 403.7490539550781,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 105.0,
|
||
|
|
"epoch": 0.10453333333333334,
|
||
|
|
"grad_norm": 0.013612802140414715,
|
||
|
|
"learning_rate": 2.8333333333333335e-06,
|
||
|
|
"loss": 0.0005,
|
||
|
|
"num_tokens": 20846052.0,
|
||
|
|
"reward": 1.315796136856079,
|
||
|
|
"reward_std": 0.21005511283874512,
|
||
|
|
"rewards/accuracy_reward_step": 0.5,
|
||
|
|
"rewards/final_brier_reward_step": 0.6128613352775574,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8374618887901306,
|
||
|
|
"step": 98
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5737812911725955,
|
||
|
|
"calib/avg_num_step_conf": 1.08203125,
|
||
|
|
"calib/ece": 0.13203125000000002,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.016724792683871947,
|
||
|
|
"calib/mean_conf": 0.14140625000000004,
|
||
|
|
"calib/mu_c": 0.15362318840579708,
|
||
|
|
"calib/mu_w": 0.13689839572192514,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.001953125,
|
||
|
|
"calib/std_conf": 0.06676374361086637,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1511.0,
|
||
|
|
"completions/max_terminated_length": 1511.0,
|
||
|
|
"completions/mean_length": 460.77734375,
|
||
|
|
"completions/mean_terminated_length": 462.5843505859375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 128.0,
|
||
|
|
"epoch": 0.1056,
|
||
|
|
"grad_norm": 0.007334718015044928,
|
||
|
|
"learning_rate": 2.805555555555556e-06,
|
||
|
|
"loss": -0.0092,
|
||
|
|
"num_tokens": 21045491.0,
|
||
|
|
"reward": 1.188649296760559,
|
||
|
|
"reward_std": 0.18395203351974487,
|
||
|
|
"rewards/accuracy_reward_step": 0.26953125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7888281345367432,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8988158702850342,
|
||
|
|
"step": 99
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.571071071071071,
|
||
|
|
"calib/avg_num_step_conf": 1.0546875,
|
||
|
|
"calib/ece": 0.29254901960784313,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.017098348348348447,
|
||
|
|
"calib/mean_conf": 0.14980392156862743,
|
||
|
|
"calib/mu_c": 0.15945945945945955,
|
||
|
|
"calib/mu_w": 0.1423611111111111,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0035294117647058825,
|
||
|
|
"calib/std_conf": 0.07468863987111943,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1628.0,
|
||
|
|
"completions/max_terminated_length": 1628.0,
|
||
|
|
"completions/mean_length": 432.60546875,
|
||
|
|
"completions/mean_terminated_length": 434.3019714355469,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 150.0,
|
||
|
|
"epoch": 0.10666666666666667,
|
||
|
|
"grad_norm": 0.011174692772328854,
|
||
|
|
"learning_rate": 2.7777777777777783e-06,
|
||
|
|
"loss": -0.0189,
|
||
|
|
"num_tokens": 21239326.0,
|
||
|
|
"reward": 1.2878634929656982,
|
||
|
|
"reward_std": 0.20618480443954468,
|
||
|
|
"rewards/accuracy_reward_step": 0.4375,
|
||
|
|
"rewards/final_brier_reward_step": 0.669003963470459,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8728216886520386,
|
||
|
|
"step": 100
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6354033506021664,
|
||
|
|
"calib/avg_num_step_conf": 1.0625,
|
||
|
|
"calib/ece": 0.20195312500000004,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03624436520217991,
|
||
|
|
"calib/mean_conf": 0.14882812500000003,
|
||
|
|
"calib/mu_c": 0.17247191011235954,
|
||
|
|
"calib/mu_w": 0.13622754491017963,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0015625,
|
||
|
|
"calib/std_conf": 0.07420538530985722,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1819.0,
|
||
|
|
"completions/max_terminated_length": 1819.0,
|
||
|
|
"completions/mean_length": 445.125,
|
||
|
|
"completions/mean_terminated_length": 446.87060546875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 100.0,
|
||
|
|
"epoch": 0.10773333333333333,
|
||
|
|
"grad_norm": 0.008462260477244854,
|
||
|
|
"learning_rate": 2.7500000000000004e-06,
|
||
|
|
"loss": -0.0132,
|
||
|
|
"num_tokens": 21435950.0,
|
||
|
|
"reward": 1.2457029819488525,
|
||
|
|
"reward_std": 0.18622824549674988,
|
||
|
|
"rewards/accuracy_reward_step": 0.34765625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7408593893051147,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9151560068130493,
|
||
|
|
"step": 101
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6699993570372276,
|
||
|
|
"calib/avg_num_step_conf": 1.03515625,
|
||
|
|
"calib/ece": 0.4326771653543307,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04813862277374137,
|
||
|
|
"calib/mean_conf": 0.1625984251968504,
|
||
|
|
"calib/mu_c": 0.1821192052980132,
|
||
|
|
"calib/mu_w": 0.13398058252427184,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0003937007874015748,
|
||
|
|
"calib/std_conf": 0.07343337650497607,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1068.0,
|
||
|
|
"completions/max_terminated_length": 1068.0,
|
||
|
|
"completions/mean_length": 357.359375,
|
||
|
|
"completions/mean_terminated_length": 358.76080322265625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 143.0,
|
||
|
|
"epoch": 0.1088,
|
||
|
|
"grad_norm": 0.009046705439686775,
|
||
|
|
"learning_rate": 2.7222222222222224e-06,
|
||
|
|
"loss": 0.0208,
|
||
|
|
"num_tokens": 21609810.0,
|
||
|
|
"reward": 1.3842028379440308,
|
||
|
|
"reward_std": 0.21628153324127197,
|
||
|
|
"rewards/accuracy_reward_step": 0.58984375,
|
||
|
|
"rewards/final_brier_reward_step": 0.581113338470459,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8339598178863525,
|
||
|
|
"step": 102
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6123626373626374,
|
||
|
|
"calib/avg_num_step_conf": 1.03515625,
|
||
|
|
"calib/ece": 0.3538671875,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.031056166056165985,
|
||
|
|
"calib/mean_conf": 0.15394531250000004,
|
||
|
|
"calib/mu_c": 0.16923076923076924,
|
||
|
|
"calib/mu_w": 0.13817460317460326,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07148304088577476,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1770.0,
|
||
|
|
"completions/max_terminated_length": 1770.0,
|
||
|
|
"completions/mean_length": 467.49609375,
|
||
|
|
"completions/mean_terminated_length": 469.3294372558594,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 137.0,
|
||
|
|
"epoch": 0.10986666666666667,
|
||
|
|
"grad_norm": 0.016644254326820374,
|
||
|
|
"learning_rate": 2.6944444444444444e-06,
|
||
|
|
"loss": 0.0239,
|
||
|
|
"num_tokens": 21809721.0,
|
||
|
|
"reward": 1.3374496698379517,
|
||
|
|
"reward_std": 0.20862965285778046,
|
||
|
|
"rewards/accuracy_reward_step": 0.5078125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6345113515853882,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8542134761810303,
|
||
|
|
"step": 103
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6197704321645863,
|
||
|
|
"calib/avg_num_step_conf": 1.15625,
|
||
|
|
"calib/ece": 0.2625,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03395847707457819,
|
||
|
|
"calib/mean_conf": 0.15546875,
|
||
|
|
"calib/mu_c": 0.17523364485981308,
|
||
|
|
"calib/mu_w": 0.14127516778523488,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07294967630796931,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1040.0,
|
||
|
|
"completions/max_terminated_length": 1040.0,
|
||
|
|
"completions/mean_length": 410.25390625,
|
||
|
|
"completions/mean_terminated_length": 411.8627624511719,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 154.0,
|
||
|
|
"epoch": 0.11093333333333333,
|
||
|
|
"grad_norm": 0.010387678630650043,
|
||
|
|
"learning_rate": 2.666666666666667e-06,
|
||
|
|
"loss": 0.0168,
|
||
|
|
"num_tokens": 21997106.0,
|
||
|
|
"reward": 1.29205322265625,
|
||
|
|
"reward_std": 0.17050066590309143,
|
||
|
|
"rewards/accuracy_reward_step": 0.41796875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6990234851837158,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.898290753364563,
|
||
|
|
"step": 104
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6384274640088594,
|
||
|
|
"calib/avg_num_step_conf": 1.16015625,
|
||
|
|
"calib/ece": 0.3392156862745098,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.039147286821705346,
|
||
|
|
"calib/mean_conf": 0.16980392156862745,
|
||
|
|
"calib/mu_c": 0.18914728682170542,
|
||
|
|
"calib/mu_w": 0.15000000000000008,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0015686274509803923,
|
||
|
|
"calib/std_conf": 0.07865262909180258,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2895.0,
|
||
|
|
"completions/max_terminated_length": 2895.0,
|
||
|
|
"completions/mean_length": 446.64453125,
|
||
|
|
"completions/mean_terminated_length": 446.64453125,
|
||
|
|
"completions/min_length": 146.0,
|
||
|
|
"completions/min_terminated_length": 146.0,
|
||
|
|
"epoch": 0.112,
|
||
|
|
"grad_norm": 0.018622733652591705,
|
||
|
|
"learning_rate": 2.6388888888888893e-06,
|
||
|
|
"loss": 0.0452,
|
||
|
|
"num_tokens": 22192887.0,
|
||
|
|
"reward": 1.342691421508789,
|
||
|
|
"reward_std": 0.25465911626815796,
|
||
|
|
"rewards/accuracy_reward_step": 0.50390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6479296684265137,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8639689683914185,
|
||
|
|
"step": 105
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6150293613897725,
|
||
|
|
"calib/avg_num_step_conf": 1.109375,
|
||
|
|
"calib/ece": 0.29121093750000004,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03357597259603623,
|
||
|
|
"calib/mean_conf": 0.1865234375,
|
||
|
|
"calib/mu_c": 0.20409836065573772,
|
||
|
|
"calib/mu_w": 0.1705223880597015,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0005859375000000001,
|
||
|
|
"calib/std_conf": 0.08006359433714923,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1013.0,
|
||
|
|
"completions/max_terminated_length": 1013.0,
|
||
|
|
"completions/mean_length": 400.0078125,
|
||
|
|
"completions/mean_terminated_length": 401.5765075683594,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 165.0,
|
||
|
|
"epoch": 0.11306666666666666,
|
||
|
|
"grad_norm": 0.011190079152584076,
|
||
|
|
"learning_rate": 2.6111111111111113e-06,
|
||
|
|
"loss": -0.0159,
|
||
|
|
"num_tokens": 22375553.0,
|
||
|
|
"reward": 1.330780029296875,
|
||
|
|
"reward_std": 0.21178483963012695,
|
||
|
|
"rewards/accuracy_reward_step": 0.4765625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6729004383087158,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8757568597793579,
|
||
|
|
"step": 106
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5785692260252853,
|
||
|
|
"calib/avg_num_step_conf": 1.125,
|
||
|
|
"calib/ece": 0.3541015625,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.023179155103299348,
|
||
|
|
"calib/mean_conf": 0.19667968750000003,
|
||
|
|
"calib/mu_c": 0.20709219858156028,
|
||
|
|
"calib/mu_w": 0.18391304347826093,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07954160012158631,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1371.0,
|
||
|
|
"completions/max_terminated_length": 1371.0,
|
||
|
|
"completions/mean_length": 409.85546875,
|
||
|
|
"completions/mean_terminated_length": 411.4627685546875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 130.0,
|
||
|
|
"epoch": 0.11413333333333334,
|
||
|
|
"grad_norm": 0.01202862337231636,
|
||
|
|
"learning_rate": 2.5833333333333337e-06,
|
||
|
|
"loss": 0.016,
|
||
|
|
"num_tokens": 22560772.0,
|
||
|
|
"reward": 1.383345365524292,
|
||
|
|
"reward_std": 0.2616827189922333,
|
||
|
|
"rewards/accuracy_reward_step": 0.55078125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6323339939117432,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8655886650085449,
|
||
|
|
"step": 107
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5877176558696874,
|
||
|
|
"calib/avg_num_step_conf": 1.10546875,
|
||
|
|
"calib/ece": 0.3658203125,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.027051738126443153,
|
||
|
|
"calib/mean_conf": 0.2095703125,
|
||
|
|
"calib/mu_c": 0.22108843537414966,
|
||
|
|
"calib/mu_w": 0.1940366972477065,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0005859375000000001,
|
||
|
|
"calib/std_conf": 0.08038123844313635,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1279.0,
|
||
|
|
"completions/max_terminated_length": 1279.0,
|
||
|
|
"completions/mean_length": 417.55859375,
|
||
|
|
"completions/mean_terminated_length": 419.19610595703125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 104.0,
|
||
|
|
"epoch": 0.1152,
|
||
|
|
"grad_norm": 0.0160062238574028,
|
||
|
|
"learning_rate": 2.5555555555555557e-06,
|
||
|
|
"loss": 0.0162,
|
||
|
|
"num_tokens": 22746579.0,
|
||
|
|
"reward": 1.4096243381500244,
|
||
|
|
"reward_std": 0.22388026118278503,
|
||
|
|
"rewards/accuracy_reward_step": 0.57421875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6293066143989563,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8830093741416931,
|
||
|
|
"step": 108
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6354875283446711,
|
||
|
|
"calib/avg_num_step_conf": 1.09375,
|
||
|
|
"calib/ece": 0.20647058823529413,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04014550264550262,
|
||
|
|
"calib/mean_conf": 0.22176470588235292,
|
||
|
|
"calib/mu_c": 0.2449074074074074,
|
||
|
|
"calib/mu_w": 0.20476190476190478,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0023529411764705885,
|
||
|
|
"calib/std_conf": 0.08658455874547602,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1906.0,
|
||
|
|
"completions/max_terminated_length": 1906.0,
|
||
|
|
"completions/mean_length": 421.09375,
|
||
|
|
"completions/mean_terminated_length": 422.7451171875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 146.0,
|
||
|
|
"epoch": 0.11626666666666667,
|
||
|
|
"grad_norm": 0.011149784550070763,
|
||
|
|
"learning_rate": 2.5277777777777778e-06,
|
||
|
|
"loss": -0.0142,
|
||
|
|
"num_tokens": 22934659.0,
|
||
|
|
"reward": 1.3087148666381836,
|
||
|
|
"reward_std": 0.18745285272598267,
|
||
|
|
"rewards/accuracy_reward_step": 0.421875,
|
||
|
|
"rewards/final_brier_reward_step": 0.7244043350219727,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9032385945320129,
|
||
|
|
"step": 109
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5625038544557508,
|
||
|
|
"calib/avg_num_step_conf": 1.203125,
|
||
|
|
"calib/ece": 0.21113281250000007,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.01761640456367558,
|
||
|
|
"calib/mean_conf": 0.24160156249999998,
|
||
|
|
"calib/mu_c": 0.25130434782608696,
|
||
|
|
"calib/mu_w": 0.23368794326241138,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0017578125,
|
||
|
|
"calib/std_conf": 0.07124570599382529,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1013.0,
|
||
|
|
"completions/max_terminated_length": 1013.0,
|
||
|
|
"completions/mean_length": 386.8984375,
|
||
|
|
"completions/mean_terminated_length": 388.41571044921875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 63.0,
|
||
|
|
"epoch": 0.11733333333333333,
|
||
|
|
"grad_norm": 0.019615396857261658,
|
||
|
|
"learning_rate": 2.5e-06,
|
||
|
|
"loss": 0.0011,
|
||
|
|
"num_tokens": 23114305.0,
|
||
|
|
"reward": 1.3225723505020142,
|
||
|
|
"reward_std": 0.24186158180236816,
|
||
|
|
"rewards/accuracy_reward_step": 0.44921875,
|
||
|
|
"rewards/final_brier_reward_step": 0.7060058116912842,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8907778263092041,
|
||
|
|
"step": 110
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6133149678604224,
|
||
|
|
"calib/avg_num_step_conf": 1.078125,
|
||
|
|
"calib/ece": 0.22578125,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.031074380165289184,
|
||
|
|
"calib/mean_conf": 0.25468749999999996,
|
||
|
|
"calib/mu_c": 0.27107438016528923,
|
||
|
|
"calib/mu_w": 0.24000000000000005,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.00390625,
|
||
|
|
"calib/std_conf": 0.07588990277863057,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1702.0,
|
||
|
|
"completions/max_terminated_length": 1702.0,
|
||
|
|
"completions/mean_length": 398.65234375,
|
||
|
|
"completions/mean_terminated_length": 400.2156982421875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 104.0,
|
||
|
|
"epoch": 0.1184,
|
||
|
|
"grad_norm": 0.020358692854642868,
|
||
|
|
"learning_rate": 2.4722222222222226e-06,
|
||
|
|
"loss": -0.0176,
|
||
|
|
"num_tokens": 23299448.0,
|
||
|
|
"reward": 1.3487699031829834,
|
||
|
|
"reward_std": 0.26086199283599854,
|
||
|
|
"rewards/accuracy_reward_step": 0.47265625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7109765410423279,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.887189507484436,
|
||
|
|
"step": 111
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.6372567482736975,
|
||
|
|
"calib/avg_num_step_conf": 1.07421875,
|
||
|
|
"calib/ece": 0.21581027667984193,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03700564971751413,
|
||
|
|
"calib/mean_conf": 0.25059288537549407,
|
||
|
|
"calib/mu_c": 0.27033898305084747,
|
||
|
|
"calib/mu_w": 0.23333333333333334,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.070357951734733,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.01171875,
|
||
|
|
"completions/max_length": 2581.0,
|
||
|
|
"completions/max_terminated_length": 2581.0,
|
||
|
|
"completions/mean_length": 428.26171875,
|
||
|
|
"completions/mean_terminated_length": 433.3399353027344,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 140.0,
|
||
|
|
"epoch": 0.11946666666666667,
|
||
|
|
"grad_norm": 0.013701686635613441,
|
||
|
|
"learning_rate": 2.4444444444444447e-06,
|
||
|
|
"loss": -0.0403,
|
||
|
|
"num_tokens": 23492683.0,
|
||
|
|
"reward": 1.3275620937347412,
|
||
|
|
"reward_std": 0.25652819871902466,
|
||
|
|
"rewards/accuracy_reward_step": 0.4609375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7076171636581421,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8700134754180908,
|
||
|
|
"step": 112
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6255586164677074,
|
||
|
|
"calib/avg_num_step_conf": 1.16015625,
|
||
|
|
"calib/ece": 0.26171875000000006,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.030413223140495882,
|
||
|
|
"calib/mean_conf": 0.26562499999999994,
|
||
|
|
"calib/mu_c": 0.28,
|
||
|
|
"calib/mu_w": 0.24958677685950414,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.06304450313072504,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1103.0,
|
||
|
|
"completions/max_terminated_length": 1103.0,
|
||
|
|
"completions/mean_length": 395.12890625,
|
||
|
|
"completions/mean_terminated_length": 396.678466796875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 153.0,
|
||
|
|
"epoch": 0.12053333333333334,
|
||
|
|
"grad_norm": 0.0159810371696949,
|
||
|
|
"learning_rate": 2.4166666666666667e-06,
|
||
|
|
"loss": 0.0352,
|
||
|
|
"num_tokens": 23674716.0,
|
||
|
|
"reward": 1.3917934894561768,
|
||
|
|
"reward_std": 0.254723459482193,
|
||
|
|
"rewards/accuracy_reward_step": 0.52734375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6875781416893005,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8920177817344666,
|
||
|
|
"step": 113
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.610899014778325,
|
||
|
|
"calib/avg_num_step_conf": 1.08203125,
|
||
|
|
"calib/ece": 0.27968750000000003,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.027536945812807967,
|
||
|
|
"calib/mean_conf": 0.27109374999999997,
|
||
|
|
"calib/mu_c": 0.2835714285714286,
|
||
|
|
"calib/mu_w": 0.2560344827586206,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.001953125,
|
||
|
|
"calib/std_conf": 0.05949309801092476,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1902.0,
|
||
|
|
"completions/max_terminated_length": 1902.0,
|
||
|
|
"completions/mean_length": 368.24609375,
|
||
|
|
"completions/mean_terminated_length": 369.6902160644531,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 122.0,
|
||
|
|
"epoch": 0.1216,
|
||
|
|
"grad_norm": 0.016689948737621307,
|
||
|
|
"learning_rate": 2.388888888888889e-06,
|
||
|
|
"loss": -0.0054,
|
||
|
|
"num_tokens": 23849691.0,
|
||
|
|
"reward": 1.4119011163711548,
|
||
|
|
"reward_std": 0.20957492291927338,
|
||
|
|
"rewards/accuracy_reward_step": 0.546875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6862499713897705,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8876047134399414,
|
||
|
|
"step": 114
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6031255848774096,
|
||
|
|
"calib/avg_num_step_conf": 1.05859375,
|
||
|
|
"calib/ece": 0.19330708661417323,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.02889138436583677,
|
||
|
|
"calib/mean_conf": 0.2673228346456693,
|
||
|
|
"calib/mu_c": 0.2829059829059828,
|
||
|
|
"calib/mu_w": 0.254014598540146,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.06927008442819448,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2621.0,
|
||
|
|
"completions/max_terminated_length": 2621.0,
|
||
|
|
"completions/mean_length": 434.8828125,
|
||
|
|
"completions/mean_terminated_length": 434.8828125,
|
||
|
|
"completions/min_length": 160.0,
|
||
|
|
"completions/min_terminated_length": 160.0,
|
||
|
|
"epoch": 0.12266666666666666,
|
||
|
|
"grad_norm": 0.023221217095851898,
|
||
|
|
"learning_rate": 2.361111111111111e-06,
|
||
|
|
"loss": 0.0444,
|
||
|
|
"num_tokens": 24041965.0,
|
||
|
|
"reward": 1.3284200429916382,
|
||
|
|
"reward_std": 0.3041614294052124,
|
||
|
|
"rewards/accuracy_reward_step": 0.45703125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7142187356948853,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8711804747581482,
|
||
|
|
"step": 115
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6542277825711821,
|
||
|
|
"calib/avg_num_step_conf": 1.09375,
|
||
|
|
"calib/ece": 0.2168627450980392,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.042253173918402526,
|
||
|
|
"calib/mean_conf": 0.26156862745098036,
|
||
|
|
"calib/mu_c": 0.28360655737704915,
|
||
|
|
"calib/mu_w": 0.24135338345864663,
|
||
|
|
"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": 0.0663730558673844,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 987.0,
|
||
|
|
"completions/max_terminated_length": 987.0,
|
||
|
|
"completions/mean_length": 425.69140625,
|
||
|
|
"completions/mean_terminated_length": 427.3608093261719,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 105.0,
|
||
|
|
"epoch": 0.12373333333333333,
|
||
|
|
"grad_norm": 0.012497956864535809,
|
||
|
|
"learning_rate": 2.3333333333333336e-06,
|
||
|
|
"loss": -0.0198,
|
||
|
|
"num_tokens": 24231142.0,
|
||
|
|
"reward": 1.3626313209533691,
|
||
|
|
"reward_std": 0.2106991708278656,
|
||
|
|
"rewards/accuracy_reward_step": 0.48046875,
|
||
|
|
"rewards/final_brier_reward_step": 0.7173047065734863,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8987287282943726,
|
||
|
|
"step": 116
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6378205128205129,
|
||
|
|
"calib/avg_num_step_conf": 1.078125,
|
||
|
|
"calib/ece": 0.22695312500000006,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.032515262515262405,
|
||
|
|
"calib/mean_conf": 0.265234375,
|
||
|
|
"calib/mu_c": 0.28174603174603163,
|
||
|
|
"calib/mu_w": 0.24923076923076923,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.06189033703543207,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1812.0,
|
||
|
|
"completions/max_terminated_length": 1812.0,
|
||
|
|
"completions/mean_length": 401.19140625,
|
||
|
|
"completions/mean_terminated_length": 402.7647399902344,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 145.0,
|
||
|
|
"epoch": 0.1248,
|
||
|
|
"grad_norm": 0.0186613742262125,
|
||
|
|
"learning_rate": 2.305555555555556e-06,
|
||
|
|
"loss": 0.0198,
|
||
|
|
"num_tokens": 24416127.0,
|
||
|
|
"reward": 1.368299961090088,
|
||
|
|
"reward_std": 0.23996347188949585,
|
||
|
|
"rewards/accuracy_reward_step": 0.4921875,
|
||
|
|
"rewards/final_brier_reward_step": 0.7074218988418579,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8942936658859253,
|
||
|
|
"step": 117
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6276030534351145,
|
||
|
|
"calib/avg_num_step_conf": 1.1171875,
|
||
|
|
"calib/ece": 0.24062500000000003,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.030876335877862537,
|
||
|
|
"calib/mean_conf": 0.27499999999999997,
|
||
|
|
"calib/mu_c": 0.2900763358778626,
|
||
|
|
"calib/mu_w": 0.25920000000000004,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.001953125,
|
||
|
|
"calib/std_conf": 0.059292706128157104,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1060.0,
|
||
|
|
"completions/max_terminated_length": 1060.0,
|
||
|
|
"completions/mean_length": 385.32421875,
|
||
|
|
"completions/mean_terminated_length": 386.8353271484375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 115.0,
|
||
|
|
"epoch": 0.12586666666666665,
|
||
|
|
"grad_norm": 0.013285689987242222,
|
||
|
|
"learning_rate": 2.277777777777778e-06,
|
||
|
|
"loss": -0.0066,
|
||
|
|
"num_tokens": 24594458.0,
|
||
|
|
"reward": 1.377181053161621,
|
||
|
|
"reward_std": 0.2066250741481781,
|
||
|
|
"rewards/accuracy_reward_step": 0.51171875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6987109184265137,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8738024234771729,
|
||
|
|
"step": 118
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.599326495304004,
|
||
|
|
"calib/avg_num_step_conf": 1.11328125,
|
||
|
|
"calib/ece": 0.27686274509803926,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.024999999999999967,
|
||
|
|
"calib/mean_conf": 0.26039215686274514,
|
||
|
|
"calib/mu_c": 0.27205882352941174,
|
||
|
|
"calib/mu_w": 0.24705882352941178,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.00196078431372549,
|
||
|
|
"calib/std_conf": 0.06230729615519736,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2266.0,
|
||
|
|
"completions/max_terminated_length": 2266.0,
|
||
|
|
"completions/mean_length": 425.3671875,
|
||
|
|
"completions/mean_terminated_length": 427.0353088378906,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 131.0,
|
||
|
|
"epoch": 0.12693333333333334,
|
||
|
|
"grad_norm": 0.021426772698760033,
|
||
|
|
"learning_rate": 2.25e-06,
|
||
|
|
"loss": -0.0409,
|
||
|
|
"num_tokens": 24784096.0,
|
||
|
|
"reward": 1.3954346179962158,
|
||
|
|
"reward_std": 0.24831436574459076,
|
||
|
|
"rewards/accuracy_reward_step": 0.53125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6825000047683716,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8964260220527649,
|
||
|
|
"step": 119
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.606035334815913,
|
||
|
|
"calib/avg_num_step_conf": 1.11328125,
|
||
|
|
"calib/ece": 0.312109375,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.023881887818136938,
|
||
|
|
"calib/mean_conf": 0.245703125,
|
||
|
|
"calib/mu_c": 0.25633802816901413,
|
||
|
|
"calib/mu_w": 0.2324561403508772,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0015625,
|
||
|
|
"calib/std_conf": 0.06540957013491508,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1347.0,
|
||
|
|
"completions/max_terminated_length": 1347.0,
|
||
|
|
"completions/mean_length": 385.70703125,
|
||
|
|
"completions/mean_terminated_length": 387.2196350097656,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 171.0,
|
||
|
|
"epoch": 0.128,
|
||
|
|
"grad_norm": 0.013562988489866257,
|
||
|
|
"learning_rate": 2.222222222222222e-06,
|
||
|
|
"loss": -0.0012,
|
||
|
|
"num_tokens": 24965205.0,
|
||
|
|
"reward": 1.4110522270202637,
|
||
|
|
"reward_std": 0.2052556872367859,
|
||
|
|
"rewards/accuracy_reward_step": 0.5546875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6650390625,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8953807353973389,
|
||
|
|
"step": 120
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6352926587301587,
|
||
|
|
"calib/avg_num_step_conf": 1.0390625,
|
||
|
|
"calib/ece": 0.2633858267716537,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03164682539682537,
|
||
|
|
"calib/mean_conf": 0.24055118110236223,
|
||
|
|
"calib/mu_c": 0.25625,
|
||
|
|
"calib/mu_w": 0.2246031746031746,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.060585166015491625,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 1416.0,
|
||
|
|
"completions/max_terminated_length": 1416.0,
|
||
|
|
"completions/mean_length": 445.78125,
|
||
|
|
"completions/mean_terminated_length": 449.2913513183594,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 133.0,
|
||
|
|
"epoch": 0.12906666666666666,
|
||
|
|
"grad_norm": 0.017739994451403618,
|
||
|
|
"learning_rate": 2.1944444444444445e-06,
|
||
|
|
"loss": 0.0258,
|
||
|
|
"num_tokens": 25160061.0,
|
||
|
|
"reward": 1.3621540069580078,
|
||
|
|
"reward_std": 0.2875977158546448,
|
||
|
|
"rewards/accuracy_reward_step": 0.5,
|
||
|
|
"rewards/final_brier_reward_step": 0.6838281154632568,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8950220346450806,
|
||
|
|
"step": 121
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6609586881110059,
|
||
|
|
"calib/avg_num_step_conf": 1.0859375,
|
||
|
|
"calib/ece": 0.34316406250000003,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04185745821507403,
|
||
|
|
"calib/mean_conf": 0.24707031249999997,
|
||
|
|
"calib/mu_c": 0.26423841059602643,
|
||
|
|
"calib/mu_w": 0.2223809523809524,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0001953125,
|
||
|
|
"calib/std_conf": 0.06481050305430704,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1336.0,
|
||
|
|
"completions/max_terminated_length": 1336.0,
|
||
|
|
"completions/mean_length": 403.1484375,
|
||
|
|
"completions/mean_terminated_length": 404.72943115234375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 146.0,
|
||
|
|
"epoch": 0.13013333333333332,
|
||
|
|
"grad_norm": 0.018619483336806297,
|
||
|
|
"learning_rate": 2.166666666666667e-06,
|
||
|
|
"loss": -0.0262,
|
||
|
|
"num_tokens": 25346291.0,
|
||
|
|
"reward": 1.436205506324768,
|
||
|
|
"reward_std": 0.1859695315361023,
|
||
|
|
"rewards/accuracy_reward_step": 0.58984375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6566308736801147,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8721853494644165,
|
||
|
|
"step": 122
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6331501831501831,
|
||
|
|
"calib/avg_num_step_conf": 1.08203125,
|
||
|
|
"calib/ece": 0.2750000000000001,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03323565323565328,
|
||
|
|
"calib/mean_conf": 0.23671874999999998,
|
||
|
|
"calib/mu_c": 0.25307692307692314,
|
||
|
|
"calib/mu_w": 0.21984126984126987,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.001953125,
|
||
|
|
"calib/std_conf": 0.06945220225764982,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1452.0,
|
||
|
|
"completions/max_terminated_length": 1452.0,
|
||
|
|
"completions/mean_length": 452.7421875,
|
||
|
|
"completions/mean_terminated_length": 454.5176696777344,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 137.0,
|
||
|
|
"epoch": 0.1312,
|
||
|
|
"grad_norm": 0.011628609150648117,
|
||
|
|
"learning_rate": 2.138888888888889e-06,
|
||
|
|
"loss": 0.0318,
|
||
|
|
"num_tokens": 25543161.0,
|
||
|
|
"reward": 1.3704475164413452,
|
||
|
|
"reward_std": 0.25859522819519043,
|
||
|
|
"rewards/accuracy_reward_step": 0.5078125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6883593797683716,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8738211989402771,
|
||
|
|
"step": 123
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6389239706996717,
|
||
|
|
"calib/avg_num_step_conf": 1.03125,
|
||
|
|
"calib/ece": 0.337843137254902,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03546981561000259,
|
||
|
|
"calib/mean_conf": 0.24254901960784314,
|
||
|
|
"calib/mu_c": 0.25743243243243247,
|
||
|
|
"calib/mu_w": 0.22196261682242988,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.06540501681468346,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2833.0,
|
||
|
|
"completions/max_terminated_length": 2833.0,
|
||
|
|
"completions/mean_length": 413.26171875,
|
||
|
|
"completions/mean_terminated_length": 413.26171875,
|
||
|
|
"completions/min_length": 133.0,
|
||
|
|
"completions/min_terminated_length": 133.0,
|
||
|
|
"epoch": 0.13226666666666667,
|
||
|
|
"grad_norm": 0.012678918428719044,
|
||
|
|
"learning_rate": 2.1111111111111114e-06,
|
||
|
|
"loss": 0.0195,
|
||
|
|
"num_tokens": 25731452.0,
|
||
|
|
"reward": 1.4276506900787354,
|
||
|
|
"reward_std": 0.23279699683189392,
|
||
|
|
"rewards/accuracy_reward_step": 0.578125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6527636647224426,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8972632884979248,
|
||
|
|
"step": 124
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6089549413101097,
|
||
|
|
"calib/avg_num_step_conf": 1.10546875,
|
||
|
|
"calib/ece": 0.30355731225296445,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.02757162690899917,
|
||
|
|
"calib/mean_conf": 0.2458498023715415,
|
||
|
|
"calib/mu_c": 0.2582733812949641,
|
||
|
|
"calib/mu_w": 0.2307017543859649,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07190634808706294,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2338.0,
|
||
|
|
"completions/max_terminated_length": 2338.0,
|
||
|
|
"completions/mean_length": 436.14453125,
|
||
|
|
"completions/mean_terminated_length": 437.85491943359375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 139.0,
|
||
|
|
"epoch": 0.13333333333333333,
|
||
|
|
"grad_norm": 0.010722563602030277,
|
||
|
|
"learning_rate": 2.0833333333333334e-06,
|
||
|
|
"loss": 0.0323,
|
||
|
|
"num_tokens": 25923593.0,
|
||
|
|
"reward": 1.3947930335998535,
|
||
|
|
"reward_std": 0.2935622036457062,
|
||
|
|
"rewards/accuracy_reward_step": 0.546875,
|
||
|
|
"rewards/final_brier_reward_step": 0.658007800579071,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8944070339202881,
|
||
|
|
"step": 125
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6487941429801896,
|
||
|
|
"calib/avg_num_step_conf": 1.12890625,
|
||
|
|
"calib/ece": 0.257843137254902,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04203580657069031,
|
||
|
|
"calib/mean_conf": 0.23627450980392156,
|
||
|
|
"calib/mu_c": 0.25753968253968257,
|
||
|
|
"calib/mu_w": 0.21550387596899226,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07321632691360917,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1216.0,
|
||
|
|
"completions/max_terminated_length": 1216.0,
|
||
|
|
"completions/mean_length": 409.92578125,
|
||
|
|
"completions/mean_terminated_length": 411.5333557128906,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 115.0,
|
||
|
|
"epoch": 0.1344,
|
||
|
|
"grad_norm": 0.012465385720133781,
|
||
|
|
"learning_rate": 2.0555555555555555e-06,
|
||
|
|
"loss": 0.0187,
|
||
|
|
"num_tokens": 26109678.0,
|
||
|
|
"reward": 1.3625491857528687,
|
||
|
|
"reward_std": 0.18904824554920197,
|
||
|
|
"rewards/accuracy_reward_step": 0.49609375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6927245855331421,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8897475004196167,
|
||
|
|
"step": 126
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.712732824427481,
|
||
|
|
"calib/avg_num_step_conf": 1.05859375,
|
||
|
|
"calib/ece": 0.244140625,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.054442748091603016,
|
||
|
|
"calib/mean_conf": 0.244140625,
|
||
|
|
"calib/mu_c": 0.27199999999999996,
|
||
|
|
"calib/mu_w": 0.21755725190839695,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07046749409911901,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1611.0,
|
||
|
|
"completions/max_terminated_length": 1611.0,
|
||
|
|
"completions/mean_length": 408.80859375,
|
||
|
|
"completions/mean_terminated_length": 410.41180419921875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 130.0,
|
||
|
|
"epoch": 0.13546666666666668,
|
||
|
|
"grad_norm": 0.021145053207874298,
|
||
|
|
"learning_rate": 2.027777777777778e-06,
|
||
|
|
"loss": 0.0046,
|
||
|
|
"num_tokens": 26293685.0,
|
||
|
|
"reward": 1.371145486831665,
|
||
|
|
"reward_std": 0.19424743950366974,
|
||
|
|
"rewards/accuracy_reward_step": 0.48828125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7127734422683716,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9059106111526489,
|
||
|
|
"step": 127
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6013173429441373,
|
||
|
|
"calib/avg_num_step_conf": 1.046875,
|
||
|
|
"calib/ece": 0.22322834645669293,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.02987012987012977,
|
||
|
|
"calib/mean_conf": 0.2570866141732283,
|
||
|
|
"calib/mu_c": 0.2727272727272726,
|
||
|
|
"calib/mu_w": 0.24285714285714283,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.001968503937007874,
|
||
|
|
"calib/std_conf": 0.07049443197111299,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 3051.0,
|
||
|
|
"completions/max_terminated_length": 3051.0,
|
||
|
|
"completions/mean_length": 470.16796875,
|
||
|
|
"completions/mean_terminated_length": 470.16796875,
|
||
|
|
"completions/min_length": 125.0,
|
||
|
|
"completions/min_terminated_length": 125.0,
|
||
|
|
"epoch": 0.13653333333333334,
|
||
|
|
"grad_norm": 0.021324368193745613,
|
||
|
|
"learning_rate": 2.0000000000000003e-06,
|
||
|
|
"loss": 0.0349,
|
||
|
|
"num_tokens": 26496392.0,
|
||
|
|
"reward": 1.352662205696106,
|
||
|
|
"reward_std": 0.2628103792667389,
|
||
|
|
"rewards/accuracy_reward_step": 0.4765625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7068359851837158,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9001019597053528,
|
||
|
|
"step": 128
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5622184684684683,
|
||
|
|
"calib/avg_num_step_conf": 1.09765625,
|
||
|
|
"calib/ece": 0.30784313725490203,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.017736486486486513,
|
||
|
|
"calib/mean_conf": 0.2568627450980392,
|
||
|
|
"calib/mu_c": 0.26458333333333334,
|
||
|
|
"calib/mu_w": 0.24684684684684682,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.0710009719921987,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 871.0,
|
||
|
|
"completions/max_terminated_length": 871.0,
|
||
|
|
"completions/mean_length": 368.359375,
|
||
|
|
"completions/mean_terminated_length": 369.803955078125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 155.0,
|
||
|
|
"epoch": 0.1376,
|
||
|
|
"grad_norm": 0.014148689806461334,
|
||
|
|
"learning_rate": 1.9722222222222224e-06,
|
||
|
|
"loss": 0.0068,
|
||
|
|
"num_tokens": 26668756.0,
|
||
|
|
"reward": 1.4037660360336304,
|
||
|
|
"reward_std": 0.2250811755657196,
|
||
|
|
"rewards/accuracy_reward_step": 0.5625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6555860042572021,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8679548501968384,
|
||
|
|
"step": 129
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6164713541666667,
|
||
|
|
"calib/avg_num_step_conf": 1.09765625,
|
||
|
|
"calib/ece": 0.37246093750000003,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03072916666666667,
|
||
|
|
"calib/mean_conf": 0.2525390624999999,
|
||
|
|
"calib/mu_c": 0.2640625,
|
||
|
|
"calib/mu_w": 0.2333333333333333,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.06572461515004172,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1189.0,
|
||
|
|
"completions/max_terminated_length": 1189.0,
|
||
|
|
"completions/mean_length": 375.1640625,
|
||
|
|
"completions/mean_terminated_length": 376.63531494140625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 145.0,
|
||
|
|
"epoch": 0.13866666666666666,
|
||
|
|
"grad_norm": 0.017018822953104973,
|
||
|
|
"learning_rate": 1.944444444444445e-06,
|
||
|
|
"loss": 0.0049,
|
||
|
|
"num_tokens": 26845766.0,
|
||
|
|
"reward": 1.4653195142745972,
|
||
|
|
"reward_std": 0.13975851237773895,
|
||
|
|
"rewards/accuracy_reward_step": 0.625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6369824409484863,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.887313187122345,
|
||
|
|
"step": 130
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.7051282051282051,
|
||
|
|
"calib/avg_num_step_conf": 1.13671875,
|
||
|
|
"calib/ece": 0.12109375000000003,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05919413919413921,
|
||
|
|
"calib/mean_conf": 0.234375,
|
||
|
|
"calib/mu_c": 0.27252747252747256,
|
||
|
|
"calib/mu_w": 0.21333333333333335,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07749747979773276,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1217.0,
|
||
|
|
"completions/max_terminated_length": 1217.0,
|
||
|
|
"completions/mean_length": 405.03515625,
|
||
|
|
"completions/mean_terminated_length": 406.6235656738281,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 72.0,
|
||
|
|
"epoch": 0.13973333333333332,
|
||
|
|
"grad_norm": 0.014553806744515896,
|
||
|
|
"learning_rate": 1.916666666666667e-06,
|
||
|
|
"loss": -0.0039,
|
||
|
|
"num_tokens": 27031343.0,
|
||
|
|
"reward": 1.2662906646728516,
|
||
|
|
"reward_std": 0.1472969353199005,
|
||
|
|
"rewards/accuracy_reward_step": 0.35546875,
|
||
|
|
"rewards/final_brier_reward_step": 0.77734375,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.88860023021698,
|
||
|
|
"step": 131
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6583458458458459,
|
||
|
|
"calib/avg_num_step_conf": 1.078125,
|
||
|
|
"calib/ece": 0.33019607843137255,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.046753003003003,
|
||
|
|
"calib/mean_conf": 0.23450980392156862,
|
||
|
|
"calib/mu_c": 0.25486111111111115,
|
||
|
|
"calib/mu_w": 0.20810810810810815,
|
||
|
|
"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": 0.07608853783690175,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1798.0,
|
||
|
|
"completions/max_terminated_length": 1798.0,
|
||
|
|
"completions/mean_length": 403.84765625,
|
||
|
|
"completions/mean_terminated_length": 405.431396484375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 104.0,
|
||
|
|
"epoch": 0.1408,
|
||
|
|
"grad_norm": 0.014779605902731419,
|
||
|
|
"learning_rate": 1.888888888888889e-06,
|
||
|
|
"loss": 0.0146,
|
||
|
|
"num_tokens": 27216000.0,
|
||
|
|
"reward": 1.4141907691955566,
|
||
|
|
"reward_std": 0.27150774002075195,
|
||
|
|
"rewards/accuracy_reward_step": 0.5625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6597656011581421,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8919193744659424,
|
||
|
|
"step": 132
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.6344942935852027,
|
||
|
|
"calib/avg_num_step_conf": 1.04296875,
|
||
|
|
"calib/ece": 0.18023715415019764,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.039898989898989934,
|
||
|
|
"calib/mean_conf": 0.21106719367588936,
|
||
|
|
"calib/mu_c": 0.23535353535353537,
|
||
|
|
"calib/mu_w": 0.19545454545454544,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07974791948932293,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.01171875,
|
||
|
|
"completions/max_length": 2233.0,
|
||
|
|
"completions/max_terminated_length": 2233.0,
|
||
|
|
"completions/mean_length": 474.875,
|
||
|
|
"completions/mean_terminated_length": 480.5059509277344,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 156.0,
|
||
|
|
"epoch": 0.14186666666666667,
|
||
|
|
"grad_norm": 0.013779650442302227,
|
||
|
|
"learning_rate": 1.8611111111111113e-06,
|
||
|
|
"loss": -0.017,
|
||
|
|
"num_tokens": 27419592.0,
|
||
|
|
"reward": 1.2738828659057617,
|
||
|
|
"reward_std": 0.2782604396343231,
|
||
|
|
"rewards/accuracy_reward_step": 0.38671875,
|
||
|
|
"rewards/final_brier_reward_step": 0.7297265529632568,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.907953143119812,
|
||
|
|
"step": 133
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.6899810366624526,
|
||
|
|
"calib/avg_num_step_conf": 1.06640625,
|
||
|
|
"calib/ece": 0.2434782608695652,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05345764854614407,
|
||
|
|
"calib/mean_conf": 0.20316205533596837,
|
||
|
|
"calib/mu_c": 0.23274336283185837,
|
||
|
|
"calib/mu_w": 0.1792857142857143,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07641860884361723,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 1774.0,
|
||
|
|
"completions/max_terminated_length": 1774.0,
|
||
|
|
"completions/mean_length": 459.703125,
|
||
|
|
"completions/mean_terminated_length": 463.3228454589844,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 150.0,
|
||
|
|
"epoch": 0.14293333333333333,
|
||
|
|
"grad_norm": 0.017993899062275887,
|
||
|
|
"learning_rate": 1.8333333333333333e-06,
|
||
|
|
"loss": 0.001,
|
||
|
|
"num_tokens": 27621908.0,
|
||
|
|
"reward": 1.319052815437317,
|
||
|
|
"reward_std": 0.2571974992752075,
|
||
|
|
"rewards/accuracy_reward_step": 0.44140625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7057812809944153,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9130858778953552,
|
||
|
|
"step": 134
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6095503875968993,
|
||
|
|
"calib/avg_num_step_conf": 1.12109375,
|
||
|
|
"calib/ece": 0.2751968503937008,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.031249612403100757,
|
||
|
|
"calib/mean_conf": 0.2169291338582677,
|
||
|
|
"calib/mu_c": 0.23279999999999998,
|
||
|
|
"calib/mu_w": 0.20155038759689922,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.08028504151693379,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2629.0,
|
||
|
|
"completions/max_terminated_length": 2629.0,
|
||
|
|
"completions/mean_length": 436.69921875,
|
||
|
|
"completions/mean_terminated_length": 438.41180419921875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 150.0,
|
||
|
|
"epoch": 0.144,
|
||
|
|
"grad_norm": 0.021255752071738243,
|
||
|
|
"learning_rate": 1.8055555555555557e-06,
|
||
|
|
"loss": 0.0371,
|
||
|
|
"num_tokens": 27815263.0,
|
||
|
|
"reward": 1.3500044345855713,
|
||
|
|
"reward_std": 0.25803160667419434,
|
||
|
|
"rewards/accuracy_reward_step": 0.48828125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6781641244888306,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8999398946762085,
|
||
|
|
"step": 135
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6774089382588478,
|
||
|
|
"calib/avg_num_step_conf": 1.19140625,
|
||
|
|
"calib/ece": 0.16972656250000004,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.049438129682252585,
|
||
|
|
"calib/mean_conf": 0.2205078125,
|
||
|
|
"calib/mu_c": 0.25102040816326526,
|
||
|
|
"calib/mu_w": 0.20158227848101268,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0037109375,
|
||
|
|
"calib/std_conf": 0.07793311395975938,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1072.0,
|
||
|
|
"completions/max_terminated_length": 1072.0,
|
||
|
|
"completions/mean_length": 429.92578125,
|
||
|
|
"completions/mean_terminated_length": 431.6117858886719,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 152.0,
|
||
|
|
"epoch": 0.14506666666666668,
|
||
|
|
"grad_norm": 0.025372860953211784,
|
||
|
|
"learning_rate": 1.777777777777778e-06,
|
||
|
|
"loss": -0.0017,
|
||
|
|
"num_tokens": 28009492.0,
|
||
|
|
"reward": 1.2842981815338135,
|
||
|
|
"reward_std": 0.2165205031633377,
|
||
|
|
"rewards/accuracy_reward_step": 0.3828125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7442480325698853,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9315086603164673,
|
||
|
|
"step": 136
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5685860993948284,
|
||
|
|
"calib/avg_num_step_conf": 1.12890625,
|
||
|
|
"calib/ece": 0.319921875,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.016315178189375834,
|
||
|
|
"calib/mean_conf": 0.221484375,
|
||
|
|
"calib/mu_c": 0.22932330827067668,
|
||
|
|
"calib/mu_w": 0.21300813008130084,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0109375,
|
||
|
|
"calib/std_conf": 0.08731757629973116,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1803.0,
|
||
|
|
"completions/max_terminated_length": 1803.0,
|
||
|
|
"completions/mean_length": 425.12890625,
|
||
|
|
"completions/mean_terminated_length": 426.7961120605469,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 159.0,
|
||
|
|
"epoch": 0.14613333333333334,
|
||
|
|
"grad_norm": 0.017735032364726067,
|
||
|
|
"learning_rate": 1.75e-06,
|
||
|
|
"loss": 0.001,
|
||
|
|
"num_tokens": 28200989.0,
|
||
|
|
"reward": 1.3787713050842285,
|
||
|
|
"reward_std": 0.18992286920547485,
|
||
|
|
"rewards/accuracy_reward_step": 0.51953125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6583203077316284,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9250069856643677,
|
||
|
|
"step": 137
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6439030796506665,
|
||
|
|
"calib/avg_num_step_conf": 1.0625,
|
||
|
|
"calib/ece": 0.41299212598425206,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04133560969203495,
|
||
|
|
"calib/mean_conf": 0.20905511811023622,
|
||
|
|
"calib/mu_c": 0.22484076433121022,
|
||
|
|
"calib/mu_w": 0.18350515463917527,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.001968503937007874,
|
||
|
|
"calib/std_conf": 0.08250920321720014,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1449.0,
|
||
|
|
"completions/max_terminated_length": 1449.0,
|
||
|
|
"completions/mean_length": 437.875,
|
||
|
|
"completions/mean_terminated_length": 439.5921936035156,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 78.0,
|
||
|
|
"epoch": 0.1472,
|
||
|
|
"grad_norm": 0.014441309496760368,
|
||
|
|
"learning_rate": 1.7222222222222224e-06,
|
||
|
|
"loss": -0.0072,
|
||
|
|
"num_tokens": 28393101.0,
|
||
|
|
"reward": 1.4396114349365234,
|
||
|
|
"reward_std": 0.23829461634159088,
|
||
|
|
"rewards/accuracy_reward_step": 0.61328125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6007031202316284,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.917976975440979,
|
||
|
|
"step": 138
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6668275107765553,
|
||
|
|
"calib/avg_num_step_conf": 1.09375,
|
||
|
|
"calib/ece": 0.3953125000000001,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05235797465096825,
|
||
|
|
"calib/mean_conf": 0.21796875000000002,
|
||
|
|
"calib/mu_c": 0.23821656050955417,
|
||
|
|
"calib/mu_w": 0.18585858585858592,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.08332316018633414,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1193.0,
|
||
|
|
"completions/max_terminated_length": 1193.0,
|
||
|
|
"completions/mean_length": 411.86328125,
|
||
|
|
"completions/mean_terminated_length": 413.47845458984375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 119.0,
|
||
|
|
"epoch": 0.14826666666666666,
|
||
|
|
"grad_norm": 0.0161239393055439,
|
||
|
|
"learning_rate": 1.6944444444444446e-06,
|
||
|
|
"loss": 0.0096,
|
||
|
|
"num_tokens": 28577314.0,
|
||
|
|
"reward": 1.4570918083190918,
|
||
|
|
"reward_std": 0.17924131453037262,
|
||
|
|
"rewards/accuracy_reward_step": 0.61328125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6244531273841858,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9263359904289246,
|
||
|
|
"step": 139
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.620045045045045,
|
||
|
|
"calib/avg_num_step_conf": 1.11328125,
|
||
|
|
"calib/ece": 0.3697628458498024,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03578185328185324,
|
||
|
|
"calib/mean_conf": 0.21521739130434783,
|
||
|
|
"calib/mu_c": 0.23006756756756755,
|
||
|
|
"calib/mu_w": 0.1942857142857143,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.0815829809491323,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 925.0,
|
||
|
|
"completions/max_terminated_length": 925.0,
|
||
|
|
"completions/mean_length": 420.44921875,
|
||
|
|
"completions/mean_terminated_length": 422.0980529785156,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 111.0,
|
||
|
|
"epoch": 0.14933333333333335,
|
||
|
|
"grad_norm": 0.015183106064796448,
|
||
|
|
"learning_rate": 1.6666666666666667e-06,
|
||
|
|
"loss": 0.0102,
|
||
|
|
"num_tokens": 28765645.0,
|
||
|
|
"reward": 1.4175596237182617,
|
||
|
|
"reward_std": 0.22534862160682678,
|
||
|
|
"rewards/accuracy_reward_step": 0.58203125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6238183975219727,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9085395336151123,
|
||
|
|
"step": 140
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.984375,
|
||
|
|
"calib/auroc": 0.7487771739130435,
|
||
|
|
"calib/avg_num_step_conf": 1.04296875,
|
||
|
|
"calib/ece": 0.41130952380952385,
|
||
|
|
"calib/final_conf_rate": 0.984375,
|
||
|
|
"calib/format_rate": 0.9765625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.0859782608695652,
|
||
|
|
"calib/mean_conf": 0.22361111111111115,
|
||
|
|
"calib/mu_c": 0.255,
|
||
|
|
"calib/mu_w": 0.1690217391304348,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.984375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98046875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.08823700697753031,
|
||
|
|
"calib/step_conf_rate": 0.98046875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2652.0,
|
||
|
|
"completions/max_terminated_length": 2652.0,
|
||
|
|
"completions/mean_length": 508.20703125,
|
||
|
|
"completions/mean_terminated_length": 508.20703125,
|
||
|
|
"completions/min_length": 174.0,
|
||
|
|
"completions/min_terminated_length": 174.0,
|
||
|
|
"epoch": 0.1504,
|
||
|
|
"grad_norm": 0.011680899187922478,
|
||
|
|
"learning_rate": 1.638888888888889e-06,
|
||
|
|
"loss": 0.0494,
|
||
|
|
"num_tokens": 28978522.0,
|
||
|
|
"reward": 1.4483433961868286,
|
||
|
|
"reward_std": 0.22178013622760773,
|
||
|
|
"rewards/accuracy_reward_step": 0.625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6166308522224426,
|
||
|
|
"rewards/format_reward_step": 0.9765625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8882371187210083,
|
||
|
|
"step": 141
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6945656238314845,
|
||
|
|
"calib/avg_num_step_conf": 1.1875,
|
||
|
|
"calib/ece": 0.33125490196078433,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.0649295774647887,
|
||
|
|
"calib/mean_conf": 0.22615686274509805,
|
||
|
|
"calib/mu_c": 0.2549295774647887,
|
||
|
|
"calib/mu_w": 0.19000000000000003,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0002745098039215687,
|
||
|
|
"calib/std_conf": 0.08723270041406593,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1378.0,
|
||
|
|
"completions/max_terminated_length": 1378.0,
|
||
|
|
"completions/mean_length": 456.234375,
|
||
|
|
"completions/mean_terminated_length": 458.0235595703125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 129.0,
|
||
|
|
"epoch": 0.15146666666666667,
|
||
|
|
"grad_norm": 0.026742398738861084,
|
||
|
|
"learning_rate": 1.6111111111111113e-06,
|
||
|
|
"loss": 0.0046,
|
||
|
|
"num_tokens": 29176158.0,
|
||
|
|
"reward": 1.4205234050750732,
|
||
|
|
"reward_std": 0.22015342116355896,
|
||
|
|
"rewards/accuracy_reward_step": 0.5546875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6656917929649353,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9366475343704224,
|
||
|
|
"step": 142
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.7243349753694581,
|
||
|
|
"calib/avg_num_step_conf": 1.078125,
|
||
|
|
"calib/ece": 0.33160000000000006,
|
||
|
|
"calib/final_conf_rate": 0.9765625,
|
||
|
|
"calib/format_rate": 0.97265625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.07523809523809522,
|
||
|
|
"calib/mean_conf": 0.24839999999999998,
|
||
|
|
"calib/mu_c": 0.27999999999999997,
|
||
|
|
"calib/mu_w": 0.20476190476190476,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98046875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.08400857099129826,
|
||
|
|
"calib/step_conf_rate": 0.98046875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2764.0,
|
||
|
|
"completions/max_terminated_length": 2764.0,
|
||
|
|
"completions/mean_length": 454.6171875,
|
||
|
|
"completions/mean_terminated_length": 454.6171875,
|
||
|
|
"completions/min_length": 156.0,
|
||
|
|
"completions/min_terminated_length": 156.0,
|
||
|
|
"epoch": 0.15253333333333333,
|
||
|
|
"grad_norm": 0.012399648316204548,
|
||
|
|
"learning_rate": 1.5833333333333333e-06,
|
||
|
|
"loss": 0.0181,
|
||
|
|
"num_tokens": 29375556.0,
|
||
|
|
"reward": 1.413042426109314,
|
||
|
|
"reward_std": 0.2350015789270401,
|
||
|
|
"rewards/accuracy_reward_step": 0.5703125,
|
||
|
|
"rewards/final_brier_reward_step": 0.656445324420929,
|
||
|
|
"rewards/format_reward_step": 0.97265625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8908417224884033,
|
||
|
|
"step": 143
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98046875,
|
||
|
|
"calib/auroc": 0.6306699958385352,
|
||
|
|
"calib/avg_num_step_conf": 1.0546875,
|
||
|
|
"calib/ece": 0.4079681274900399,
|
||
|
|
"calib/final_conf_rate": 0.98046875,
|
||
|
|
"calib/format_rate": 0.97265625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04504785684560961,
|
||
|
|
"calib/mean_conf": 0.2414342629482072,
|
||
|
|
"calib/mu_c": 0.2574074074074074,
|
||
|
|
"calib/mu_w": 0.2123595505617978,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98046875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.97265625,
|
||
|
|
"calib/pce": 0.00199203187250996,
|
||
|
|
"calib/std_conf": 0.08721187429063451,
|
||
|
|
"calib/step_conf_rate": 0.97265625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 2071.0,
|
||
|
|
"completions/max_terminated_length": 2071.0,
|
||
|
|
"completions/mean_length": 452.03125,
|
||
|
|
"completions/mean_terminated_length": 455.5905456542969,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 115.0,
|
||
|
|
"epoch": 0.1536,
|
||
|
|
"grad_norm": 0.015421283431351185,
|
||
|
|
"learning_rate": 1.5555555555555558e-06,
|
||
|
|
"loss": 0.0031,
|
||
|
|
"num_tokens": 29571084.0,
|
||
|
|
"reward": 1.446988582611084,
|
||
|
|
"reward_std": 0.2860656678676605,
|
||
|
|
"rewards/accuracy_reward_step": 0.6328125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6020312309265137,
|
||
|
|
"rewards/format_reward_step": 0.97265625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8854540586471558,
|
||
|
|
"step": 144
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.5864285714285714,
|
||
|
|
"calib/avg_num_step_conf": 1.20703125,
|
||
|
|
"calib/ece": 0.3535433070866142,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.02918181818181817,
|
||
|
|
"calib/mean_conf": 0.2566929133858268,
|
||
|
|
"calib/mu_c": 0.2681818181818182,
|
||
|
|
"calib/mu_w": 0.23900000000000002,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.001968503937007874,
|
||
|
|
"calib/std_conf": 0.08237665671673708,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2405.0,
|
||
|
|
"completions/max_terminated_length": 2405.0,
|
||
|
|
"completions/mean_length": 448.1484375,
|
||
|
|
"completions/mean_terminated_length": 448.1484375,
|
||
|
|
"completions/min_length": 141.0,
|
||
|
|
"completions/min_terminated_length": 141.0,
|
||
|
|
"epoch": 0.15466666666666667,
|
||
|
|
"grad_norm": 0.020210454240441322,
|
||
|
|
"learning_rate": 1.527777777777778e-06,
|
||
|
|
"loss": 0.0296,
|
||
|
|
"num_tokens": 29764194.0,
|
||
|
|
"reward": 1.4496335983276367,
|
||
|
|
"reward_std": 0.22999891638755798,
|
||
|
|
"rewards/accuracy_reward_step": 0.6015625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6411718130111694,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9193153381347656,
|
||
|
|
"step": 145
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6356532742802912,
|
||
|
|
"calib/avg_num_step_conf": 1.08203125,
|
||
|
|
"calib/ece": 0.18149606299212603,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04190446061372982,
|
||
|
|
"calib/mean_conf": 0.24763779527559054,
|
||
|
|
"calib/mu_c": 0.2715596330275229,
|
||
|
|
"calib/mu_w": 0.22965517241379307,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07870963975533606,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2552.0,
|
||
|
|
"completions/max_terminated_length": 2552.0,
|
||
|
|
"completions/mean_length": 479.94921875,
|
||
|
|
"completions/mean_terminated_length": 479.94921875,
|
||
|
|
"completions/min_length": 148.0,
|
||
|
|
"completions/min_terminated_length": 148.0,
|
||
|
|
"epoch": 0.15573333333333333,
|
||
|
|
"grad_norm": 0.019039005041122437,
|
||
|
|
"learning_rate": 1.5e-06,
|
||
|
|
"loss": 0.0385,
|
||
|
|
"num_tokens": 29969957.0,
|
||
|
|
"reward": 1.3201770782470703,
|
||
|
|
"reward_std": 0.22056898474693298,
|
||
|
|
"rewards/accuracy_reward_step": 0.42578125,
|
||
|
|
"rewards/final_brier_reward_step": 0.730664074420929,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.925629734992981,
|
||
|
|
"step": 146
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6030879307098475,
|
||
|
|
"calib/avg_num_step_conf": 1.17578125,
|
||
|
|
"calib/ece": 0.21594488188976374,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03393899453963467,
|
||
|
|
"calib/mean_conf": 0.2395669291338583,
|
||
|
|
"calib/mu_c": 0.25840707964601767,
|
||
|
|
"calib/mu_w": 0.224468085106383,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.00531496062992126,
|
||
|
|
"calib/std_conf": 0.09146343794719812,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 1561.0,
|
||
|
|
"completions/max_terminated_length": 1561.0,
|
||
|
|
"completions/mean_length": 499.375,
|
||
|
|
"completions/mean_terminated_length": 499.375,
|
||
|
|
"completions/min_length": 159.0,
|
||
|
|
"completions/min_terminated_length": 159.0,
|
||
|
|
"epoch": 0.1568,
|
||
|
|
"grad_norm": 0.015556562691926956,
|
||
|
|
"learning_rate": 1.4722222222222225e-06,
|
||
|
|
"loss": -0.0014,
|
||
|
|
"num_tokens": 30177157.0,
|
||
|
|
"reward": 1.3237640857696533,
|
||
|
|
"reward_std": 0.19050589203834534,
|
||
|
|
"rewards/accuracy_reward_step": 0.44140625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7136621475219727,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9114823937416077,
|
||
|
|
"step": 147
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.605428279585583,
|
||
|
|
"calib/avg_num_step_conf": 1.1171875,
|
||
|
|
"calib/ece": 0.4505882352941177,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.038246023639282006,
|
||
|
|
"calib/mean_conf": 0.2513725490196078,
|
||
|
|
"calib/mu_c": 0.26292134831460673,
|
||
|
|
"calib/mu_w": 0.22467532467532472,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.00196078431372549,
|
||
|
|
"calib/std_conf": 0.08715592535723,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1449.0,
|
||
|
|
"completions/max_terminated_length": 1449.0,
|
||
|
|
"completions/mean_length": 460.58984375,
|
||
|
|
"completions/mean_terminated_length": 462.3961181640625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 120.0,
|
||
|
|
"epoch": 0.15786666666666666,
|
||
|
|
"grad_norm": 0.025586625561118126,
|
||
|
|
"learning_rate": 1.4444444444444445e-06,
|
||
|
|
"loss": 0.0233,
|
||
|
|
"num_tokens": 30375860.0,
|
||
|
|
"reward": 1.5209028720855713,
|
||
|
|
"reward_std": 0.19411450624465942,
|
||
|
|
"rewards/accuracy_reward_step": 0.6953125,
|
||
|
|
"rewards/final_brier_reward_step": 0.5958983898162842,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9152524471282959,
|
||
|
|
"step": 148
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6476557246775475,
|
||
|
|
"calib/avg_num_step_conf": 1.30078125,
|
||
|
|
"calib/ece": 0.29023437500000004,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04970352711045911,
|
||
|
|
"calib/mean_conf": 0.23476562499999998,
|
||
|
|
"calib/mu_c": 0.2586466165413534,
|
||
|
|
"calib/mu_w": 0.2089430894308943,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.002734375000000001,
|
||
|
|
"calib/std_conf": 0.09015147707253263,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1677.0,
|
||
|
|
"completions/max_terminated_length": 1677.0,
|
||
|
|
"completions/mean_length": 506.140625,
|
||
|
|
"completions/mean_terminated_length": 508.1255187988281,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 163.0,
|
||
|
|
"epoch": 0.15893333333333334,
|
||
|
|
"grad_norm": 0.014532111585140228,
|
||
|
|
"learning_rate": 1.4166666666666667e-06,
|
||
|
|
"loss": -0.0186,
|
||
|
|
"num_tokens": 30585568.0,
|
||
|
|
"reward": 1.389301061630249,
|
||
|
|
"reward_std": 0.25378724932670593,
|
||
|
|
"rewards/accuracy_reward_step": 0.51953125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6859765648841858,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9071261286735535,
|
||
|
|
"step": 149
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5439709595959595,
|
||
|
|
"calib/avg_num_step_conf": 1.19921875,
|
||
|
|
"calib/ece": 0.26567460317460323,
|
||
|
|
"calib/final_conf_rate": 0.984375,
|
||
|
|
"calib/format_rate": 0.97265625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.013939393939393974,
|
||
|
|
"calib/mean_conf": 0.2748015873015873,
|
||
|
|
"calib/mu_c": 0.28143939393939393,
|
||
|
|
"calib/mu_w": 0.26749999999999996,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98046875,
|
||
|
|
"calib/pce": 0.008333333333333338,
|
||
|
|
"calib/std_conf": 0.07695829831811614,
|
||
|
|
"calib/step_conf_rate": 0.98046875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1974.0,
|
||
|
|
"completions/max_terminated_length": 1974.0,
|
||
|
|
"completions/mean_length": 438.8046875,
|
||
|
|
"completions/mean_terminated_length": 440.5255126953125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 147.0,
|
||
|
|
"epoch": 0.16,
|
||
|
|
"grad_norm": 0.018813475966453552,
|
||
|
|
"learning_rate": 1.3888888888888892e-06,
|
||
|
|
"loss": 0.0019,
|
||
|
|
"num_tokens": 30778542.0,
|
||
|
|
"reward": 1.364441156387329,
|
||
|
|
"reward_std": 0.25788232684135437,
|
||
|
|
"rewards/accuracy_reward_step": 0.515625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6683886647224426,
|
||
|
|
"rewards/format_reward_step": 0.97265625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8913000822067261,
|
||
|
|
"step": 150
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.653863010267969,
|
||
|
|
"calib/avg_num_step_conf": 1.08984375,
|
||
|
|
"calib/ece": 0.22766798418972337,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.97265625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04400826446280984,
|
||
|
|
"calib/mean_conf": 0.25059288537549407,
|
||
|
|
"calib/mu_c": 0.2735537190082644,
|
||
|
|
"calib/mu_w": 0.22954545454545458,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9765625,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.08130387705202266,
|
||
|
|
"calib/step_conf_rate": 0.9765625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2698.0,
|
||
|
|
"completions/max_terminated_length": 2698.0,
|
||
|
|
"completions/mean_length": 543.51953125,
|
||
|
|
"completions/mean_terminated_length": 545.6510009765625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 159.0,
|
||
|
|
"epoch": 0.16106666666666666,
|
||
|
|
"grad_norm": 0.023138921707868576,
|
||
|
|
"learning_rate": 1.3611111111111112e-06,
|
||
|
|
"loss": 0.0396,
|
||
|
|
"num_tokens": 31000387.0,
|
||
|
|
"reward": 1.333712100982666,
|
||
|
|
"reward_std": 0.24264037609100342,
|
||
|
|
"rewards/accuracy_reward_step": 0.47265625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6918749809265137,
|
||
|
|
"rewards/format_reward_step": 0.97265625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8932860493659973,
|
||
|
|
"step": 151
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6129373019325173,
|
||
|
|
"calib/avg_num_step_conf": 1.05859375,
|
||
|
|
"calib/ece": 0.22480314960629916,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98046875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03603430062760202,
|
||
|
|
"calib/mean_conf": 0.2555118110236221,
|
||
|
|
"calib/mu_c": 0.2743801652892562,
|
||
|
|
"calib/mu_w": 0.23834586466165417,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.0019685039370078744,
|
||
|
|
"calib/std_conf": 0.08889707984732463,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2537.0,
|
||
|
|
"completions/max_terminated_length": 2537.0,
|
||
|
|
"completions/mean_length": 524.52734375,
|
||
|
|
"completions/mean_terminated_length": 526.5843505859375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 175.0,
|
||
|
|
"epoch": 0.16213333333333332,
|
||
|
|
"grad_norm": 0.02300630323588848,
|
||
|
|
"learning_rate": 1.3333333333333334e-06,
|
||
|
|
"loss": -0.0155,
|
||
|
|
"num_tokens": 31215738.0,
|
||
|
|
"reward": 1.3442671298980713,
|
||
|
|
"reward_std": 0.2782951593399048,
|
||
|
|
"rewards/accuracy_reward_step": 0.47265625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6958593726158142,
|
||
|
|
"rewards/format_reward_step": 0.98046875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9181627035140991,
|
||
|
|
"step": 152
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.96875,
|
||
|
|
"calib/auroc": 0.5876557708618777,
|
||
|
|
"calib/avg_num_step_conf": 1.05078125,
|
||
|
|
"calib/ece": 0.2848790322580645,
|
||
|
|
"calib/final_conf_rate": 0.96875,
|
||
|
|
"calib/format_rate": 0.9609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03167286487897181,
|
||
|
|
"calib/mean_conf": 0.24536290322580642,
|
||
|
|
"calib/mu_c": 0.26030534351145046,
|
||
|
|
"calib/mu_w": 0.22863247863247865,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.96875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.97265625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.96484375,
|
||
|
|
"calib/pce": 0.0010080645161290335,
|
||
|
|
"calib/std_conf": 0.08548173118137481,
|
||
|
|
"calib/step_conf_rate": 0.96484375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.01171875,
|
||
|
|
"completions/max_length": 2081.0,
|
||
|
|
"completions/max_terminated_length": 2081.0,
|
||
|
|
"completions/mean_length": 523.2265625,
|
||
|
|
"completions/mean_terminated_length": 529.4308471679688,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 170.0,
|
||
|
|
"epoch": 0.1632,
|
||
|
|
"grad_norm": 0.0198051817715168,
|
||
|
|
"learning_rate": 1.3055555555555556e-06,
|
||
|
|
"loss": 0.006,
|
||
|
|
"num_tokens": 31432684.0,
|
||
|
|
"reward": 1.3532977104187012,
|
||
|
|
"reward_std": 0.2662625014781952,
|
||
|
|
"rewards/accuracy_reward_step": 0.515625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6515527963638306,
|
||
|
|
"rewards/format_reward_step": 0.9609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8944602608680725,
|
||
|
|
"step": 153
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6288372093023256,
|
||
|
|
"calib/avg_num_step_conf": 1.0390625,
|
||
|
|
"calib/ece": 0.2688976377952756,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.043665116279069716,
|
||
|
|
"calib/mean_conf": 0.23897637795275592,
|
||
|
|
"calib/mu_c": 0.26046511627906976,
|
||
|
|
"calib/mu_w": 0.21680000000000005,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.08748754373132563,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 1541.0,
|
||
|
|
"completions/max_terminated_length": 1541.0,
|
||
|
|
"completions/mean_length": 467.68359375,
|
||
|
|
"completions/mean_terminated_length": 471.36614990234375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 184.0,
|
||
|
|
"epoch": 0.16426666666666667,
|
||
|
|
"grad_norm": 0.015027433633804321,
|
||
|
|
"learning_rate": 1.2777777777777779e-06,
|
||
|
|
"loss": -0.0261,
|
||
|
|
"num_tokens": 31632531.0,
|
||
|
|
"reward": 1.3765439987182617,
|
||
|
|
"reward_std": 0.20490823686122894,
|
||
|
|
"rewards/accuracy_reward_step": 0.50390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6865234375,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9268794059753418,
|
||
|
|
"step": 154
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5754440578648599,
|
||
|
|
"calib/avg_num_step_conf": 1.16015625,
|
||
|
|
"calib/ece": 0.26503906250000003,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.025566135628395298,
|
||
|
|
"calib/mean_conf": 0.23886718749999997,
|
||
|
|
"calib/mu_c": 0.25155038759689924,
|
||
|
|
"calib/mu_w": 0.22598425196850394,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.0871818207015651,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1521.0,
|
||
|
|
"completions/max_terminated_length": 1521.0,
|
||
|
|
"completions/mean_length": 457.54296875,
|
||
|
|
"completions/mean_terminated_length": 459.3372802734375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 162.0,
|
||
|
|
"epoch": 0.16533333333333333,
|
||
|
|
"grad_norm": 0.016778098419308662,
|
||
|
|
"learning_rate": 1.25e-06,
|
||
|
|
"loss": -0.0047,
|
||
|
|
"num_tokens": 31832558.0,
|
||
|
|
"reward": 1.374552607536316,
|
||
|
|
"reward_std": 0.1923726499080658,
|
||
|
|
"rewards/accuracy_reward_step": 0.50390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6812012195587158,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9248706102371216,
|
||
|
|
"step": 155
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6032649253731344,
|
||
|
|
"calib/avg_num_step_conf": 1.17578125,
|
||
|
|
"calib/ece": 0.2856299212598426,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.033669154228855774,
|
||
|
|
"calib/mean_conf": 0.24192913385826773,
|
||
|
|
"calib/mu_c": 0.2578358208955224,
|
||
|
|
"calib/mu_w": 0.22416666666666665,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.08651054108111587,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2905.0,
|
||
|
|
"completions/max_terminated_length": 2905.0,
|
||
|
|
"completions/mean_length": 474.0703125,
|
||
|
|
"completions/mean_terminated_length": 475.929443359375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 148.0,
|
||
|
|
"epoch": 0.1664,
|
||
|
|
"grad_norm": 0.020775113254785538,
|
||
|
|
"learning_rate": 1.2222222222222223e-06,
|
||
|
|
"loss": -0.0068,
|
||
|
|
"num_tokens": 32034360.0,
|
||
|
|
"reward": 1.3807235956192017,
|
||
|
|
"reward_std": 0.24285456538200378,
|
||
|
|
"rewards/accuracy_reward_step": 0.5234375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6656738519668579,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9165468215942383,
|
||
|
|
"step": 156
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.5783989092325672,
|
||
|
|
"calib/avg_num_step_conf": 1.09765625,
|
||
|
|
"calib/ece": 0.37075098814229246,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.021821841319309226,
|
||
|
|
"calib/mean_conf": 0.23557312252964427,
|
||
|
|
"calib/mu_c": 0.24437086092715235,
|
||
|
|
"calib/mu_w": 0.22254901960784312,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.0047430830039525695,
|
||
|
|
"calib/std_conf": 0.08434452912789969,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 2779.0,
|
||
|
|
"completions/max_terminated_length": 2779.0,
|
||
|
|
"completions/mean_length": 471.4765625,
|
||
|
|
"completions/mean_terminated_length": 475.18896484375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 138.0,
|
||
|
|
"epoch": 0.16746666666666668,
|
||
|
|
"grad_norm": 0.019724013283848763,
|
||
|
|
"learning_rate": 1.1944444444444446e-06,
|
||
|
|
"loss": 0.0233,
|
||
|
|
"num_tokens": 32234466.0,
|
||
|
|
"reward": 1.4214212894439697,
|
||
|
|
"reward_std": 0.22582949697971344,
|
||
|
|
"rewards/accuracy_reward_step": 0.58984375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6209765672683716,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9031072854995728,
|
||
|
|
"step": 157
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.984375,
|
||
|
|
"calib/auroc": 0.5811827956989247,
|
||
|
|
"calib/avg_num_step_conf": 1.09375,
|
||
|
|
"calib/ece": 0.3725099601593626,
|
||
|
|
"calib/final_conf_rate": 0.98046875,
|
||
|
|
"calib/format_rate": 0.98046875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.028750000000000026,
|
||
|
|
"calib/mean_conf": 0.24900398406374502,
|
||
|
|
"calib/mu_c": 0.26,
|
||
|
|
"calib/mu_w": 0.23124999999999998,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98046875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.984375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.0019920318725099606,
|
||
|
|
"calib/std_conf": 0.08942603603191979,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.01171875,
|
||
|
|
"completions/max_length": 2498.0,
|
||
|
|
"completions/max_terminated_length": 2498.0,
|
||
|
|
"completions/mean_length": 486.0390625,
|
||
|
|
"completions/mean_terminated_length": 491.8023986816406,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 88.0,
|
||
|
|
"epoch": 0.16853333333333334,
|
||
|
|
"grad_norm": 0.01671520248055458,
|
||
|
|
"learning_rate": 1.1666666666666668e-06,
|
||
|
|
"loss": -0.0134,
|
||
|
|
"num_tokens": 32439812.0,
|
||
|
|
"reward": 1.4309093952178955,
|
||
|
|
"reward_std": 0.27989810705184937,
|
||
|
|
"rewards/accuracy_reward_step": 0.60546875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6212109327316284,
|
||
|
|
"rewards/format_reward_step": 0.98046875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8827787637710571,
|
||
|
|
"step": 158
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.637247615508485,
|
||
|
|
"calib/avg_num_step_conf": 1.0625,
|
||
|
|
"calib/ece": 0.2913725490196079,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04466740988480117,
|
||
|
|
"calib/mean_conf": 0.25921568627450986,
|
||
|
|
"calib/mu_c": 0.2797101449275362,
|
||
|
|
"calib/mu_w": 0.23504273504273504,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0047058823529411795,
|
||
|
|
"calib/std_conf": 0.08058227161751418,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2902.0,
|
||
|
|
"completions/max_terminated_length": 2902.0,
|
||
|
|
"completions/mean_length": 462.94921875,
|
||
|
|
"completions/mean_terminated_length": 464.7647399902344,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 134.0,
|
||
|
|
"epoch": 0.1696,
|
||
|
|
"grad_norm": 0.02069372683763504,
|
||
|
|
"learning_rate": 1.138888888888889e-06,
|
||
|
|
"loss": -0.0197,
|
||
|
|
"num_tokens": 32638791.0,
|
||
|
|
"reward": 1.40556001663208,
|
||
|
|
"reward_std": 0.2604949176311493,
|
||
|
|
"rewards/accuracy_reward_step": 0.5390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.67808598279953,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9238807559013367,
|
||
|
|
"step": 159
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6567470026914607,
|
||
|
|
"calib/avg_num_step_conf": 1.0859375,
|
||
|
|
"calib/ece": 0.2943359375,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05235502813799853,
|
||
|
|
"calib/mean_conf": 0.2302734375,
|
||
|
|
"calib/mu_c": 0.25522388059701495,
|
||
|
|
"calib/mu_w": 0.20286885245901642,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0005859375000000001,
|
||
|
|
"calib/std_conf": 0.08734094605014073,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1568.0,
|
||
|
|
"completions/max_terminated_length": 1568.0,
|
||
|
|
"completions/mean_length": 464.828125,
|
||
|
|
"completions/mean_terminated_length": 466.6510009765625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 195.0,
|
||
|
|
"epoch": 0.17066666666666666,
|
||
|
|
"grad_norm": 0.01723591983318329,
|
||
|
|
"learning_rate": 1.111111111111111e-06,
|
||
|
|
"loss": 0.021,
|
||
|
|
"num_tokens": 32838307.0,
|
||
|
|
"reward": 1.3904207944869995,
|
||
|
|
"reward_std": 0.2143954634666443,
|
||
|
|
"rewards/accuracy_reward_step": 0.5234375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6823534965515137,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9079135656356812,
|
||
|
|
"step": 160
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5613266583229035,
|
||
|
|
"calib/avg_num_step_conf": 1.1484375,
|
||
|
|
"calib/ece": 0.47847656250000004,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.01935857321652054,
|
||
|
|
"calib/mean_conf": 0.25980468749999996,
|
||
|
|
"calib/mu_c": 0.2649468085106383,
|
||
|
|
"calib/mu_w": 0.24558823529411775,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.001953125,
|
||
|
|
"calib/std_conf": 0.0822175618589322,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1818.0,
|
||
|
|
"completions/max_terminated_length": 1818.0,
|
||
|
|
"completions/mean_length": 454.125,
|
||
|
|
"completions/mean_terminated_length": 455.9059143066406,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 139.0,
|
||
|
|
"epoch": 0.17173333333333332,
|
||
|
|
"grad_norm": 0.013422448188066483,
|
||
|
|
"learning_rate": 1.0833333333333335e-06,
|
||
|
|
"loss": 0.0223,
|
||
|
|
"num_tokens": 33034163.0,
|
||
|
|
"reward": 1.54856538772583,
|
||
|
|
"reward_std": 0.1663738191127777,
|
||
|
|
"rewards/accuracy_reward_step": 0.734375,
|
||
|
|
"rewards/final_brier_reward_step": 0.5785152912139893,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9044186472892761,
|
||
|
|
"step": 161
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6401791135397692,
|
||
|
|
"calib/avg_num_step_conf": 1.015625,
|
||
|
|
"calib/ece": 0.46156862745098043,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.051047358834244105,
|
||
|
|
"calib/mean_conf": 0.25607843137254904,
|
||
|
|
"calib/mu_c": 0.27049180327868855,
|
||
|
|
"calib/mu_w": 0.21944444444444444,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.09305445235678816,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1722.0,
|
||
|
|
"completions/max_terminated_length": 1722.0,
|
||
|
|
"completions/mean_length": 428.890625,
|
||
|
|
"completions/mean_terminated_length": 430.57257080078125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 148.0,
|
||
|
|
"epoch": 0.1728,
|
||
|
|
"grad_norm": 0.019668465480208397,
|
||
|
|
"learning_rate": 1.0555555555555557e-06,
|
||
|
|
"loss": 0.0157,
|
||
|
|
"num_tokens": 33223783.0,
|
||
|
|
"reward": 1.5355653762817383,
|
||
|
|
"reward_std": 0.1955740749835968,
|
||
|
|
"rewards/accuracy_reward_step": 0.71484375,
|
||
|
|
"rewards/final_brier_reward_step": 0.5876562595367432,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9216366410255432,
|
||
|
|
"step": 162
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.7115226337448559,
|
||
|
|
"calib/avg_num_step_conf": 1.1015625,
|
||
|
|
"calib/ece": 0.28591269841269845,
|
||
|
|
"calib/final_conf_rate": 0.984375,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.0785470085470085,
|
||
|
|
"calib/mean_conf": 0.2501984126984127,
|
||
|
|
"calib/mu_c": 0.2866666666666667,
|
||
|
|
"calib/mu_w": 0.20811965811965819,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0001984126984126984,
|
||
|
|
"calib/std_conf": 0.10555014755211915,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.015625,
|
||
|
|
"completions/max_length": 2325.0,
|
||
|
|
"completions/max_terminated_length": 2325.0,
|
||
|
|
"completions/mean_length": 521.4453125,
|
||
|
|
"completions/mean_terminated_length": 529.7222290039062,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 128.0,
|
||
|
|
"epoch": 0.17386666666666667,
|
||
|
|
"grad_norm": 0.014891761355102062,
|
||
|
|
"learning_rate": 1.0277777777777777e-06,
|
||
|
|
"loss": -0.0231,
|
||
|
|
"num_tokens": 33437785.0,
|
||
|
|
"reward": 1.3962335586547852,
|
||
|
|
"reward_std": 0.22696751356124878,
|
||
|
|
"rewards/accuracy_reward_step": 0.52734375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6867871284484863,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9207350611686707,
|
||
|
|
"step": 163
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.6489661654135338,
|
||
|
|
"calib/avg_num_step_conf": 1.14453125,
|
||
|
|
"calib/ece": 0.28715415019762847,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05587092731829579,
|
||
|
|
"calib/mean_conf": 0.23853754940711466,
|
||
|
|
"calib/mu_c": 0.26503759398496246,
|
||
|
|
"calib/mu_w": 0.20916666666666667,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.1040059076552682,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.01171875,
|
||
|
|
"completions/max_length": 2729.0,
|
||
|
|
"completions/max_terminated_length": 2729.0,
|
||
|
|
"completions/mean_length": 555.09765625,
|
||
|
|
"completions/mean_terminated_length": 561.6798706054688,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 179.0,
|
||
|
|
"epoch": 0.17493333333333333,
|
||
|
|
"grad_norm": 0.016471445560455322,
|
||
|
|
"learning_rate": 1.0000000000000002e-06,
|
||
|
|
"loss": -0.0415,
|
||
|
|
"num_tokens": 33661706.0,
|
||
|
|
"reward": 1.3797190189361572,
|
||
|
|
"reward_std": 0.20705097913742065,
|
||
|
|
"rewards/accuracy_reward_step": 0.51953125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6772167682647705,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.90038001537323,
|
||
|
|
"step": 164
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6223961543202268,
|
||
|
|
"calib/avg_num_step_conf": 1.15234375,
|
||
|
|
"calib/ece": 0.2605882352941178,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04616356464932825,
|
||
|
|
"calib/mean_conf": 0.2613725490196079,
|
||
|
|
"calib/mu_c": 0.28345864661654135,
|
||
|
|
"calib/mu_w": 0.2372950819672131,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.00019607843137254904,
|
||
|
|
"calib/std_conf": 0.10407424966110544,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1956.0,
|
||
|
|
"completions/max_terminated_length": 1956.0,
|
||
|
|
"completions/mean_length": 550.890625,
|
||
|
|
"completions/mean_terminated_length": 553.051025390625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 149.0,
|
||
|
|
"epoch": 0.176,
|
||
|
|
"grad_norm": 0.011542506515979767,
|
||
|
|
"learning_rate": 9.722222222222224e-07,
|
||
|
|
"loss": -0.0159,
|
||
|
|
"num_tokens": 33883990.0,
|
||
|
|
"reward": 1.3936247825622559,
|
||
|
|
"reward_std": 0.17919129133224487,
|
||
|
|
"rewards/accuracy_reward_step": 0.51953125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6849511861801147,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9405341744422913,
|
||
|
|
"step": 165
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6533776301218162,
|
||
|
|
"calib/avg_num_step_conf": 1.09765625,
|
||
|
|
"calib/ece": 0.3903543307086615,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.057336655592469504,
|
||
|
|
"calib/mean_conf": 0.2710629921259842,
|
||
|
|
"calib/mu_c": 0.29047619047619044,
|
||
|
|
"calib/mu_w": 0.23313953488372094,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.09880811341224238,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 2063.0,
|
||
|
|
"completions/max_terminated_length": 2063.0,
|
||
|
|
"completions/mean_length": 550.6171875,
|
||
|
|
"completions/mean_terminated_length": 554.9527587890625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 175.0,
|
||
|
|
"epoch": 0.17706666666666668,
|
||
|
|
"grad_norm": 0.013915646821260452,
|
||
|
|
"learning_rate": 9.444444444444445e-07,
|
||
|
|
"loss": -0.0148,
|
||
|
|
"num_tokens": 34106812.0,
|
||
|
|
"reward": 1.4977242946624756,
|
||
|
|
"reward_std": 0.21183454990386963,
|
||
|
|
"rewards/accuracy_reward_step": 0.65625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6308495998382568,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9182603359222412,
|
||
|
|
"step": 166
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5557319784058432,
|
||
|
|
"calib/avg_num_step_conf": 1.08203125,
|
||
|
|
"calib/ece": 0.45333333333333337,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.02475389012384882,
|
||
|
|
"calib/mean_conf": 0.283921568627451,
|
||
|
|
"calib/mu_c": 0.2904255319148936,
|
||
|
|
"calib/mu_w": 0.2656716417910448,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.10744954915319879,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1595.0,
|
||
|
|
"completions/max_terminated_length": 1595.0,
|
||
|
|
"completions/mean_length": 527.37109375,
|
||
|
|
"completions/mean_terminated_length": 529.4392700195312,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 164.0,
|
||
|
|
"epoch": 0.17813333333333334,
|
||
|
|
"grad_norm": 0.021476641297340393,
|
||
|
|
"learning_rate": 9.166666666666666e-07,
|
||
|
|
"loss": 0.0531,
|
||
|
|
"num_tokens": 34323107.0,
|
||
|
|
"reward": 1.5472520589828491,
|
||
|
|
"reward_std": 0.21065786480903625,
|
||
|
|
"rewards/accuracy_reward_step": 0.734375,
|
||
|
|
"rewards/final_brier_reward_step": 0.5907421708106995,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8840864896774292,
|
||
|
|
"step": 167
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6603872135942614,
|
||
|
|
"calib/avg_num_step_conf": 1.0546875,
|
||
|
|
"calib/ece": 0.39570312500000004,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.0632286370917195,
|
||
|
|
"calib/mean_conf": 0.280078125,
|
||
|
|
"calib/mu_c": 0.30057803468208094,
|
||
|
|
"calib/mu_w": 0.23734939759036144,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.11160054388973369,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2102.0,
|
||
|
|
"completions/max_terminated_length": 2102.0,
|
||
|
|
"completions/mean_length": 616.66015625,
|
||
|
|
"completions/mean_terminated_length": 619.0784912109375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 190.0,
|
||
|
|
"epoch": 0.1792,
|
||
|
|
"grad_norm": 0.013876114040613174,
|
||
|
|
"learning_rate": 8.88888888888889e-07,
|
||
|
|
"loss": 0.0211,
|
||
|
|
"num_tokens": 34561324.0,
|
||
|
|
"reward": 1.5183846950531006,
|
||
|
|
"reward_std": 0.21503406763076782,
|
||
|
|
"rewards/accuracy_reward_step": 0.67578125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6360155940055847,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.903070330619812,
|
||
|
|
"step": 168
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.602853048462741,
|
||
|
|
"calib/avg_num_step_conf": 1.046875,
|
||
|
|
"calib/ece": 0.29150197628458496,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03785500260552371,
|
||
|
|
"calib/mean_conf": 0.31482213438735174,
|
||
|
|
"calib/mu_c": 0.3299342105263158,
|
||
|
|
"calib/mu_w": 0.29207920792079206,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0027667984189723317,
|
||
|
|
"calib/std_conf": 0.11258579505958703,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1404.0,
|
||
|
|
"completions/max_terminated_length": 1404.0,
|
||
|
|
"completions/mean_length": 559.39453125,
|
||
|
|
"completions/mean_terminated_length": 561.5882568359375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 174.0,
|
||
|
|
"epoch": 0.18026666666666666,
|
||
|
|
"grad_norm": 0.01111834030598402,
|
||
|
|
"learning_rate": 8.611111111111112e-07,
|
||
|
|
"loss": -0.0282,
|
||
|
|
"num_tokens": 34784393.0,
|
||
|
|
"reward": 1.4540972709655762,
|
||
|
|
"reward_std": 0.20363430678844452,
|
||
|
|
"rewards/accuracy_reward_step": 0.59765625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6738574504852295,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8967991471290588,
|
||
|
|
"step": 169
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6974222046413502,
|
||
|
|
"calib/avg_num_step_conf": 1.0625,
|
||
|
|
"calib/ece": 0.30314960629921267,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.08229166666666665,
|
||
|
|
"calib/mean_conf": 0.3188976377952756,
|
||
|
|
"calib/mu_c": 0.35,
|
||
|
|
"calib/mu_w": 0.2677083333333333,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.10884669813644965,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1629.0,
|
||
|
|
"completions/max_terminated_length": 1629.0,
|
||
|
|
"completions/mean_length": 568.24609375,
|
||
|
|
"completions/mean_terminated_length": 570.4745483398438,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 198.0,
|
||
|
|
"epoch": 0.18133333333333335,
|
||
|
|
"grad_norm": 0.010557408444583416,
|
||
|
|
"learning_rate": 8.333333333333333e-07,
|
||
|
|
"loss": -0.0077,
|
||
|
|
"num_tokens": 35009696.0,
|
||
|
|
"reward": 1.4908766746520996,
|
||
|
|
"reward_std": 0.18308508396148682,
|
||
|
|
"rewards/accuracy_reward_step": 0.62109375,
|
||
|
|
"rewards/final_brier_reward_step": 0.6905077695846558,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9121789932250977,
|
||
|
|
"step": 170
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6503187078408317,
|
||
|
|
"calib/avg_num_step_conf": 1.09375,
|
||
|
|
"calib/ece": 0.23945312500000004,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.06436660684448287,
|
||
|
|
"calib/mean_conf": 0.31914062500000007,
|
||
|
|
"calib/mu_c": 0.34755244755244746,
|
||
|
|
"calib/mu_w": 0.2831858407079646,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.10997533348260136,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1401.0,
|
||
|
|
"completions/max_terminated_length": 1401.0,
|
||
|
|
"completions/mean_length": 558.64453125,
|
||
|
|
"completions/mean_terminated_length": 560.8353271484375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 230.0,
|
||
|
|
"epoch": 0.1824,
|
||
|
|
"grad_norm": 0.012995216064155102,
|
||
|
|
"learning_rate": 8.055555555555557e-07,
|
||
|
|
"loss": -0.0082,
|
||
|
|
"num_tokens": 35235285.0,
|
||
|
|
"reward": 1.4445995092391968,
|
||
|
|
"reward_std": 0.24250145256519318,
|
||
|
|
"rewards/accuracy_reward_step": 0.55859375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7086328268051147,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9361324906349182,
|
||
|
|
"step": 171
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6476407346573092,
|
||
|
|
"calib/avg_num_step_conf": 1.01953125,
|
||
|
|
"calib/ece": 0.36156862745098045,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05843661340898909,
|
||
|
|
"calib/mean_conf": 0.34823529411764703,
|
||
|
|
"calib/mu_c": 0.3651933701657458,
|
||
|
|
"calib/mu_w": 0.30675675675675673,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.09735131211242885,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2385.0,
|
||
|
|
"completions/max_terminated_length": 2385.0,
|
||
|
|
"completions/mean_length": 539.7421875,
|
||
|
|
"completions/mean_terminated_length": 541.8588256835938,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 213.0,
|
||
|
|
"epoch": 0.18346666666666667,
|
||
|
|
"grad_norm": 0.019643718376755714,
|
||
|
|
"learning_rate": 7.777777777777779e-07,
|
||
|
|
"loss": 0.0121,
|
||
|
|
"num_tokens": 35452491.0,
|
||
|
|
"reward": 1.5684814453125,
|
||
|
|
"reward_std": 0.19160641729831696,
|
||
|
|
"rewards/accuracy_reward_step": 0.70703125,
|
||
|
|
"rewards/final_brier_reward_step": 0.6713671684265137,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9124413132667542,
|
||
|
|
"step": 172
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.4669896193771626,
|
||
|
|
"calib/avg_num_step_conf": 1.0546875,
|
||
|
|
"calib/ece": 0.3266666666666667,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": -0.011470588235294121,
|
||
|
|
"calib/mean_conf": 0.34,
|
||
|
|
"calib/mu_c": 0.33617647058823524,
|
||
|
|
"calib/mu_w": 0.34764705882352936,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.10245993956511139,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1751.0,
|
||
|
|
"completions/max_terminated_length": 1751.0,
|
||
|
|
"completions/mean_length": 617.77734375,
|
||
|
|
"completions/mean_terminated_length": 620.2000122070312,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 131.0,
|
||
|
|
"epoch": 0.18453333333333333,
|
||
|
|
"grad_norm": 0.011911396868526936,
|
||
|
|
"learning_rate": 7.5e-07,
|
||
|
|
"loss": 0.0077,
|
||
|
|
"num_tokens": 35689482.0,
|
||
|
|
"reward": 1.5004022121429443,
|
||
|
|
"reward_std": 0.2213282287120819,
|
||
|
|
"rewards/accuracy_reward_step": 0.6640625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6490429639816284,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8566482067108154,
|
||
|
|
"step": 173
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9765625,
|
||
|
|
"calib/auroc": 0.4596725288053366,
|
||
|
|
"calib/avg_num_step_conf": 1.203125,
|
||
|
|
"calib/ece": 0.31200000000000006,
|
||
|
|
"calib/final_conf_rate": 0.9765625,
|
||
|
|
"calib/format_rate": 0.9609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": -0.01675089279698133,
|
||
|
|
"calib/mean_conf": 0.32480000000000003,
|
||
|
|
"calib/mu_c": 0.3183006535947712,
|
||
|
|
"calib/mu_w": 0.33505154639175255,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98046875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.96484375,
|
||
|
|
"calib/pce": 0.0124,
|
||
|
|
"calib/std_conf": 0.11038550629498423,
|
||
|
|
"calib/step_conf_rate": 0.96484375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 2894.0,
|
||
|
|
"completions/max_terminated_length": 2894.0,
|
||
|
|
"completions/mean_length": 686.48046875,
|
||
|
|
"completions/mean_terminated_length": 691.8858032226562,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 211.0,
|
||
|
|
"epoch": 0.1856,
|
||
|
|
"grad_norm": 0.010533534921705723,
|
||
|
|
"learning_rate": 7.222222222222222e-07,
|
||
|
|
"loss": -0.005,
|
||
|
|
"num_tokens": 35945133.0,
|
||
|
|
"reward": 1.416163444519043,
|
||
|
|
"reward_std": 0.29552051424980164,
|
||
|
|
"rewards/accuracy_reward_step": 0.59765625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6329687833786011,
|
||
|
|
"rewards/format_reward_step": 0.9609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8549661636352539,
|
||
|
|
"step": 174
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.984375,
|
||
|
|
"calib/auroc": 0.6765468091998704,
|
||
|
|
"calib/avg_num_step_conf": 1.04296875,
|
||
|
|
"calib/ece": 0.09603174603174608,
|
||
|
|
"calib/final_conf_rate": 0.984375,
|
||
|
|
"calib/format_rate": 0.98046875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.06802721088435382,
|
||
|
|
"calib/mean_conf": 0.33174603174603173,
|
||
|
|
"calib/mu_c": 0.37142857142857144,
|
||
|
|
"calib/mu_w": 0.3034013605442176,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.984375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98046875,
|
||
|
|
"calib/pce": 0.005555555555555572,
|
||
|
|
"calib/std_conf": 0.10775545038385559,
|
||
|
|
"calib/step_conf_rate": 0.98046875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 2986.0,
|
||
|
|
"completions/max_terminated_length": 2986.0,
|
||
|
|
"completions/mean_length": 638.01953125,
|
||
|
|
"completions/mean_terminated_length": 643.0433349609375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 103.0,
|
||
|
|
"epoch": 0.18666666666666668,
|
||
|
|
"grad_norm": 0.009028359316289425,
|
||
|
|
"learning_rate": 6.944444444444446e-07,
|
||
|
|
"loss": -0.0385,
|
||
|
|
"num_tokens": 36189970.0,
|
||
|
|
"reward": 1.3047478199005127,
|
||
|
|
"reward_std": 0.20292183756828308,
|
||
|
|
"rewards/accuracy_reward_step": 0.41015625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7562109231948853,
|
||
|
|
"rewards/format_reward_step": 0.98046875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8893823027610779,
|
||
|
|
"step": 175
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.5794956968141326,
|
||
|
|
"calib/avg_num_step_conf": 1.03515625,
|
||
|
|
"calib/ece": 0.3505928853754941,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03269666314359049,
|
||
|
|
"calib/mean_conf": 0.3569169960474308,
|
||
|
|
"calib/mu_c": 0.36648044692737425,
|
||
|
|
"calib/mu_w": 0.33378378378378376,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.0965899866518836,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 1773.0,
|
||
|
|
"completions/max_terminated_length": 1773.0,
|
||
|
|
"completions/mean_length": 575.5625,
|
||
|
|
"completions/mean_terminated_length": 580.094482421875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 187.0,
|
||
|
|
"epoch": 0.18773333333333334,
|
||
|
|
"grad_norm": 0.012131310068070889,
|
||
|
|
"learning_rate": 6.666666666666667e-07,
|
||
|
|
"loss": -0.0189,
|
||
|
|
"num_tokens": 36417058.0,
|
||
|
|
"reward": 1.5568971633911133,
|
||
|
|
"reward_std": 0.19418621063232422,
|
||
|
|
"rewards/accuracy_reward_step": 0.69921875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6644531488418579,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9205576181411743,
|
||
|
|
"step": 176
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6526360544217688,
|
||
|
|
"calib/avg_num_step_conf": 1.0546875,
|
||
|
|
"calib/ece": 0.2535433070866142,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05884353741496601,
|
||
|
|
"calib/mean_conf": 0.36062992125984256,
|
||
|
|
"calib/mu_c": 0.3833333333333333,
|
||
|
|
"calib/mu_w": 0.3244897959183673,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.09319987860740346,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1413.0,
|
||
|
|
"completions/max_terminated_length": 1413.0,
|
||
|
|
"completions/mean_length": 570.96875,
|
||
|
|
"completions/mean_terminated_length": 573.2078857421875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 225.0,
|
||
|
|
"epoch": 0.1888,
|
||
|
|
"grad_norm": 0.016085185110569,
|
||
|
|
"learning_rate": 6.388888888888889e-07,
|
||
|
|
"loss": 0.014,
|
||
|
|
"num_tokens": 36642738.0,
|
||
|
|
"reward": 1.486448049545288,
|
||
|
|
"reward_std": 0.20098945498466492,
|
||
|
|
"rewards/accuracy_reward_step": 0.609375,
|
||
|
|
"rewards/final_brier_reward_step": 0.712343692779541,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8929798007011414,
|
||
|
|
"step": 177
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6869550173010381,
|
||
|
|
"calib/avg_num_step_conf": 1.109375,
|
||
|
|
"calib/ece": 0.3121568627450981,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.07117647058823534,
|
||
|
|
"calib/mean_conf": 0.35450980392156867,
|
||
|
|
"calib/mu_c": 0.37823529411764706,
|
||
|
|
"calib/mu_w": 0.3070588235294117,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.0956364383237669,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1244.0,
|
||
|
|
"completions/max_terminated_length": 1244.0,
|
||
|
|
"completions/mean_length": 575.24609375,
|
||
|
|
"completions/mean_terminated_length": 577.5020141601562,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 80.0,
|
||
|
|
"epoch": 0.18986666666666666,
|
||
|
|
"grad_norm": 0.020766116678714752,
|
||
|
|
"learning_rate": 6.111111111111112e-07,
|
||
|
|
"loss": 0.016,
|
||
|
|
"num_tokens": 36871753.0,
|
||
|
|
"reward": 1.5327744483947754,
|
||
|
|
"reward_std": 0.22211039066314697,
|
||
|
|
"rewards/accuracy_reward_step": 0.6640625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6946484446525574,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8996132612228394,
|
||
|
|
"step": 178
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9765625,
|
||
|
|
"calib/auroc": 0.6402533521999865,
|
||
|
|
"calib/avg_num_step_conf": 1.046875,
|
||
|
|
"calib/ece": 0.25960000000000005,
|
||
|
|
"calib/final_conf_rate": 0.9765625,
|
||
|
|
"calib/format_rate": 0.97265625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.0567818880129371,
|
||
|
|
"calib/mean_conf": 0.3564,
|
||
|
|
"calib/mu_c": 0.37843137254901954,
|
||
|
|
"calib/mu_w": 0.32164948453608244,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9765625,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9765625,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.97265625,
|
||
|
|
"calib/pce": 0.002,
|
||
|
|
"calib/std_conf": 0.09065892123779104,
|
||
|
|
"calib/step_conf_rate": 0.97265625,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.01171875,
|
||
|
|
"completions/max_length": 3037.0,
|
||
|
|
"completions/max_terminated_length": 3037.0,
|
||
|
|
"completions/mean_length": 621.015625,
|
||
|
|
"completions/mean_terminated_length": 628.3794555664062,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 207.0,
|
||
|
|
"epoch": 0.19093333333333334,
|
||
|
|
"grad_norm": 0.010622275993227959,
|
||
|
|
"learning_rate": 5.833333333333334e-07,
|
||
|
|
"loss": -0.009,
|
||
|
|
"num_tokens": 37112677.0,
|
||
|
|
"reward": 1.4636054039001465,
|
||
|
|
"reward_std": 0.2516264021396637,
|
||
|
|
"rewards/accuracy_reward_step": 0.59765625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6956249475479126,
|
||
|
|
"rewards/format_reward_step": 0.97265625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9053595066070557,
|
||
|
|
"step": 179
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5907735262389746,
|
||
|
|
"calib/avg_num_step_conf": 1.01171875,
|
||
|
|
"calib/ece": 0.23593750000000008,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03666476299257554,
|
||
|
|
"calib/mean_conf": 0.3617187500000001,
|
||
|
|
"calib/mu_c": 0.37647058823529406,
|
||
|
|
"calib/mu_w": 0.3398058252427185,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.09692224150543309,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1541.0,
|
||
|
|
"completions/max_terminated_length": 1541.0,
|
||
|
|
"completions/mean_length": 623.40234375,
|
||
|
|
"completions/mean_terminated_length": 625.8471069335938,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 196.0,
|
||
|
|
"epoch": 0.192,
|
||
|
|
"grad_norm": 0.012773178517818451,
|
||
|
|
"learning_rate": 5.555555555555555e-07,
|
||
|
|
"loss": -0.0161,
|
||
|
|
"num_tokens": 37351804.0,
|
||
|
|
"reward": 1.4770622253417969,
|
||
|
|
"reward_std": 0.1948680281639099,
|
||
|
|
"rewards/accuracy_reward_step": 0.59765625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7088280916213989,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.904654860496521,
|
||
|
|
"step": 180
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.6617485277939559,
|
||
|
|
"calib/avg_num_step_conf": 1.0234375,
|
||
|
|
"calib/ece": 0.25078740157480317,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9765625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.06619426648547211,
|
||
|
|
"calib/mean_conf": 0.3547244094488189,
|
||
|
|
"calib/mu_c": 0.3810457516339869,
|
||
|
|
"calib/mu_w": 0.3148514851485148,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98046875,
|
||
|
|
"calib/pce": 0.0015748031496062992,
|
||
|
|
"calib/std_conf": 0.10289760043767955,
|
||
|
|
"calib/step_conf_rate": 0.98046875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 3014.0,
|
||
|
|
"completions/max_terminated_length": 3014.0,
|
||
|
|
"completions/mean_length": 570.703125,
|
||
|
|
"completions/mean_terminated_length": 572.9412231445312,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 175.0,
|
||
|
|
"epoch": 0.19306666666666666,
|
||
|
|
"grad_norm": 0.011856647208333015,
|
||
|
|
"learning_rate": 5.277777777777779e-07,
|
||
|
|
"loss": 0.0644,
|
||
|
|
"num_tokens": 37579848.0,
|
||
|
|
"reward": 1.4679954051971436,
|
||
|
|
"reward_std": 0.2306535542011261,
|
||
|
|
"rewards/accuracy_reward_step": 0.59765625,
|
||
|
|
"rewards/final_brier_reward_step": 0.7007812261581421,
|
||
|
|
"rewards/format_reward_step": 0.9765625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9079186916351318,
|
||
|
|
"step": 181
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6935108856478546,
|
||
|
|
"calib/avg_num_step_conf": 1.0,
|
||
|
|
"calib/ece": 0.3135433070866142,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.07587965898682447,
|
||
|
|
"calib/mean_conf": 0.36,
|
||
|
|
"calib/mu_c": 0.38479532163742686,
|
||
|
|
"calib/mu_w": 0.3089156626506024,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.00015748031496062991,
|
||
|
|
"calib/std_conf": 0.09142786838874807,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2598.0,
|
||
|
|
"completions/max_terminated_length": 2598.0,
|
||
|
|
"completions/mean_length": 613.7734375,
|
||
|
|
"completions/mean_terminated_length": 613.7734375,
|
||
|
|
"completions/min_length": 230.0,
|
||
|
|
"completions/min_terminated_length": 230.0,
|
||
|
|
"epoch": 0.19413333333333332,
|
||
|
|
"grad_norm": 0.014810623601078987,
|
||
|
|
"learning_rate": 5.000000000000001e-07,
|
||
|
|
"loss": 0.0274,
|
||
|
|
"num_tokens": 37818814.0,
|
||
|
|
"reward": 1.5463969707489014,
|
||
|
|
"reward_std": 0.17773394286632538,
|
||
|
|
"rewards/accuracy_reward_step": 0.66796875,
|
||
|
|
"rewards/final_brier_reward_step": 0.7013999819755554,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9202882051467896,
|
||
|
|
"step": 182
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.628010878010878,
|
||
|
|
"calib/avg_num_step_conf": 1.0625,
|
||
|
|
"calib/ece": 0.24274509803921573,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.046037296037296005,
|
||
|
|
"calib/mean_conf": 0.37058823529411766,
|
||
|
|
"calib/mu_c": 0.3884615384615384,
|
||
|
|
"calib/mu_w": 0.3424242424242424,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.0007843137254901962,
|
||
|
|
"calib/std_conf": 0.09475275907216514,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1340.0,
|
||
|
|
"completions/max_terminated_length": 1340.0,
|
||
|
|
"completions/mean_length": 601.56640625,
|
||
|
|
"completions/mean_terminated_length": 603.925537109375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 202.0,
|
||
|
|
"epoch": 0.1952,
|
||
|
|
"grad_norm": 0.016529878601431847,
|
||
|
|
"learning_rate": 4.7222222222222226e-07,
|
||
|
|
"loss": 0.0203,
|
||
|
|
"num_tokens": 38055175.0,
|
||
|
|
"reward": 1.4892892837524414,
|
||
|
|
"reward_std": 0.28105586767196655,
|
||
|
|
"rewards/accuracy_reward_step": 0.61328125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7111327648162842,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.891141414642334,
|
||
|
|
"step": 183
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6555634301913537,
|
||
|
|
"calib/avg_num_step_conf": 1.03125,
|
||
|
|
"calib/ece": 0.3063241106719368,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.9765625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05875974486180008,
|
||
|
|
"calib/mean_conf": 0.3671936758893281,
|
||
|
|
"calib/mu_c": 0.38647058823529407,
|
||
|
|
"calib/mu_w": 0.327710843373494,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98046875,
|
||
|
|
"calib/pce": 0.0007905138339920949,
|
||
|
|
"calib/std_conf": 0.08706245792854107,
|
||
|
|
"calib/step_conf_rate": 0.98046875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 1799.0,
|
||
|
|
"completions/max_terminated_length": 1799.0,
|
||
|
|
"completions/mean_length": 602.84375,
|
||
|
|
"completions/mean_terminated_length": 607.590576171875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 262.0,
|
||
|
|
"epoch": 0.19626666666666667,
|
||
|
|
"grad_norm": 0.010777152143418789,
|
||
|
|
"learning_rate": 4.444444444444445e-07,
|
||
|
|
"loss": -0.0174,
|
||
|
|
"num_tokens": 38290463.0,
|
||
|
|
"reward": 1.5306932926177979,
|
||
|
|
"reward_std": 0.21238690614700317,
|
||
|
|
"rewards/accuracy_reward_step": 0.66796875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6872656345367432,
|
||
|
|
"rewards/format_reward_step": 0.9765625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.904492199420929,
|
||
|
|
"step": 184
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.95703125,
|
||
|
|
"calib/auroc": 0.7062724014336916,
|
||
|
|
"calib/avg_num_step_conf": 0.984375,
|
||
|
|
"calib/ece": 0.276326530612245,
|
||
|
|
"calib/final_conf_rate": 0.95703125,
|
||
|
|
"calib/format_rate": 0.94921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.08168458781362004,
|
||
|
|
"calib/mean_conf": 0.3661224489795919,
|
||
|
|
"calib/mu_c": 0.39612903225806445,
|
||
|
|
"calib/mu_w": 0.3144444444444444,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.95703125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.95703125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.94921875,
|
||
|
|
"calib/pce": 0.004897959183673465,
|
||
|
|
"calib/std_conf": 0.09321505367019424,
|
||
|
|
"calib/step_conf_rate": 0.94921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 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": 3025.0,
|
||
|
|
"completions/max_terminated_length": 3025.0,
|
||
|
|
"completions/mean_length": 650.30078125,
|
||
|
|
"completions/mean_terminated_length": 655.4212646484375,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 225.0,
|
||
|
|
"epoch": 0.19733333333333333,
|
||
|
|
"grad_norm": 0.01256662979722023,
|
||
|
|
"learning_rate": 4.1666666666666667e-07,
|
||
|
|
"loss": 0.0048,
|
||
|
|
"num_tokens": 38539540.0,
|
||
|
|
"reward": 1.4508123397827148,
|
||
|
|
"reward_std": 0.222275048494339,
|
||
|
|
"rewards/accuracy_reward_step": 0.60546875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6873437166213989,
|
||
|
|
"rewards/format_reward_step": 0.94921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8676245808601379,
|
||
|
|
"step": 185
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6813147295742232,
|
||
|
|
"calib/avg_num_step_conf": 0.9921875,
|
||
|
|
"calib/ece": 0.3235294117647059,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.07091484464902187,
|
||
|
|
"calib/mean_conf": 0.36666666666666664,
|
||
|
|
"calib/mu_c": 0.38863636363636367,
|
||
|
|
"calib/mu_w": 0.3177215189873418,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.09344531097808413,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2861.0,
|
||
|
|
"completions/max_terminated_length": 2861.0,
|
||
|
|
"completions/mean_length": 607.05859375,
|
||
|
|
"completions/mean_terminated_length": 607.05859375,
|
||
|
|
"completions/min_length": 190.0,
|
||
|
|
"completions/min_terminated_length": 190.0,
|
||
|
|
"epoch": 0.1984,
|
||
|
|
"grad_norm": 0.009213064797222614,
|
||
|
|
"learning_rate": 3.8888888888888895e-07,
|
||
|
|
"loss": 0.0548,
|
||
|
|
"num_tokens": 38775667.0,
|
||
|
|
"reward": 1.559600591659546,
|
||
|
|
"reward_std": 0.18579980731010437,
|
||
|
|
"rewards/accuracy_reward_step": 0.6875,
|
||
|
|
"rewards/final_brier_reward_step": 0.6948046684265137,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9128557443618774,
|
||
|
|
"step": 186
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.6369402985074628,
|
||
|
|
"calib/avg_num_step_conf": 1.05078125,
|
||
|
|
"calib/ece": 0.16181102362204727,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05512437810945264,
|
||
|
|
"calib/mean_conf": 0.365748031496063,
|
||
|
|
"calib/mu_c": 0.39179104477611937,
|
||
|
|
"calib/mu_w": 0.3366666666666667,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.09457754960785239,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 2285.0,
|
||
|
|
"completions/max_terminated_length": 2285.0,
|
||
|
|
"completions/mean_length": 634.3359375,
|
||
|
|
"completions/mean_terminated_length": 639.3306884765625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 165.0,
|
||
|
|
"epoch": 0.19946666666666665,
|
||
|
|
"grad_norm": 0.009206795133650303,
|
||
|
|
"learning_rate": 3.611111111111111e-07,
|
||
|
|
"loss": -0.0023,
|
||
|
|
"num_tokens": 39015281.0,
|
||
|
|
"reward": 1.4204821586608887,
|
||
|
|
"reward_std": 0.2516452372074127,
|
||
|
|
"rewards/accuracy_reward_step": 0.5234375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7373046875,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9229444265365601,
|
||
|
|
"step": 187
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6273310023310024,
|
||
|
|
"calib/avg_num_step_conf": 1.015625,
|
||
|
|
"calib/ece": 0.3295275590551181,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05447261072261067,
|
||
|
|
"calib/mean_conf": 0.36338582677165354,
|
||
|
|
"calib/mu_c": 0.3801136363636363,
|
||
|
|
"calib/mu_w": 0.32564102564102565,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.09410766553530815,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1914.0,
|
||
|
|
"completions/max_terminated_length": 1914.0,
|
||
|
|
"completions/mean_length": 625.09765625,
|
||
|
|
"completions/mean_terminated_length": 627.549072265625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 195.0,
|
||
|
|
"epoch": 0.20053333333333334,
|
||
|
|
"grad_norm": 0.009032557718455791,
|
||
|
|
"learning_rate": 3.3333333333333335e-07,
|
||
|
|
"loss": 0.024,
|
||
|
|
"num_tokens": 39255058.0,
|
||
|
|
"reward": 1.5570321083068848,
|
||
|
|
"reward_std": 0.19003382325172424,
|
||
|
|
"rewards/accuracy_reward_step": 0.69140625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6839843392372131,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9085971117019653,
|
||
|
|
"step": 188
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.5884168362991661,
|
||
|
|
"calib/avg_num_step_conf": 1.01171875,
|
||
|
|
"calib/ece": 0.24724409448818907,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.9921875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.029588285507912482,
|
||
|
|
"calib/mean_conf": 0.3708661417322835,
|
||
|
|
"calib/mu_c": 0.38216560509554137,
|
||
|
|
"calib/mu_w": 0.3525773195876289,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.08927550582456467,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2851.0,
|
||
|
|
"completions/max_terminated_length": 2851.0,
|
||
|
|
"completions/mean_length": 600.6015625,
|
||
|
|
"completions/mean_terminated_length": 602.9569091796875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 256.0,
|
||
|
|
"epoch": 0.2016,
|
||
|
|
"grad_norm": 0.00952502153813839,
|
||
|
|
"learning_rate": 3.055555555555556e-07,
|
||
|
|
"loss": -0.0013,
|
||
|
|
"num_tokens": 39492260.0,
|
||
|
|
"reward": 1.4871289730072021,
|
||
|
|
"reward_std": 0.22185944020748138,
|
||
|
|
"rewards/accuracy_reward_step": 0.61328125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7032812833786011,
|
||
|
|
"rewards/format_reward_step": 0.9921875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8982037305831909,
|
||
|
|
"step": 189
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5954418725280426,
|
||
|
|
"calib/avg_num_step_conf": 1.01171875,
|
||
|
|
"calib/ece": 0.2550781250000001,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 0.99609375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03285353044154826,
|
||
|
|
"calib/mean_conf": 0.366796875,
|
||
|
|
"calib/mu_c": 0.37924528301886784,
|
||
|
|
"calib/mu_w": 0.3463917525773196,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.99609375,
|
||
|
|
"calib/pce": 0.000390625,
|
||
|
|
"calib/std_conf": 0.08356727822679387,
|
||
|
|
"calib/step_conf_rate": 0.99609375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1795.0,
|
||
|
|
"completions/max_terminated_length": 1795.0,
|
||
|
|
"completions/mean_length": 653.4296875,
|
||
|
|
"completions/mean_terminated_length": 655.9921875,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 190.0,
|
||
|
|
"epoch": 0.20266666666666666,
|
||
|
|
"grad_norm": 0.015201202593743801,
|
||
|
|
"learning_rate": 2.7777777777777776e-07,
|
||
|
|
"loss": 0.0272,
|
||
|
|
"num_tokens": 39740826.0,
|
||
|
|
"reward": 1.5011475086212158,
|
||
|
|
"reward_std": 0.15234965085983276,
|
||
|
|
"rewards/accuracy_reward_step": 0.62109375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7049218416213989,
|
||
|
|
"rewards/format_reward_step": 0.99609375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9150592088699341,
|
||
|
|
"step": 190
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.6745073108709473,
|
||
|
|
"calib/avg_num_step_conf": 1.0703125,
|
||
|
|
"calib/ece": 0.1964426877470356,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.06468531468531474,
|
||
|
|
"calib/mean_conf": 0.3774703557312253,
|
||
|
|
"calib/mu_c": 0.40559440559440557,
|
||
|
|
"calib/mu_w": 0.34090909090909083,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.004347826086956521,
|
||
|
|
"calib/std_conf": 0.09823704950740778,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2133.0,
|
||
|
|
"completions/max_terminated_length": 2133.0,
|
||
|
|
"completions/mean_length": 587.34375,
|
||
|
|
"completions/mean_terminated_length": 589.6470947265625,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 202.0,
|
||
|
|
"epoch": 0.20373333333333332,
|
||
|
|
"grad_norm": 0.012601793743669987,
|
||
|
|
"learning_rate": 2.5000000000000004e-07,
|
||
|
|
"loss": 0.0447,
|
||
|
|
"num_tokens": 39971034.0,
|
||
|
|
"reward": 1.4446088075637817,
|
||
|
|
"reward_std": 0.1962060034275055,
|
||
|
|
"rewards/accuracy_reward_step": 0.55859375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7324609756469727,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.893200695514679,
|
||
|
|
"step": 191
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.6367849153084052,
|
||
|
|
"calib/avg_num_step_conf": 1.0546875,
|
||
|
|
"calib/ece": 0.23011811023622056,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98046875,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.048191115372323345,
|
||
|
|
"calib/mean_conf": 0.37303149606299213,
|
||
|
|
"calib/mu_c": 0.3929530201342281,
|
||
|
|
"calib/mu_w": 0.34476190476190477,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.00826771653543307,
|
||
|
|
"calib/std_conf": 0.09736217288446995,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 2149.0,
|
||
|
|
"completions/max_terminated_length": 2149.0,
|
||
|
|
"completions/mean_length": 617.0859375,
|
||
|
|
"completions/mean_terminated_length": 621.9448852539062,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 164.0,
|
||
|
|
"epoch": 0.2048,
|
||
|
|
"grad_norm": 0.01687338761985302,
|
||
|
|
"learning_rate": 2.2222222222222224e-07,
|
||
|
|
"loss": 0.0049,
|
||
|
|
"num_tokens": 40209664.0,
|
||
|
|
"reward": 1.4559879302978516,
|
||
|
|
"reward_std": 0.24735161662101746,
|
||
|
|
"rewards/accuracy_reward_step": 0.58203125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7122167944908142,
|
||
|
|
"rewards/format_reward_step": 0.98046875,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8948308229446411,
|
||
|
|
"step": 192
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.564126942495574,
|
||
|
|
"calib/avg_num_step_conf": 0.99609375,
|
||
|
|
"calib/ece": 0.2158730158730159,
|
||
|
|
"calib/final_conf_rate": 0.984375,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.029060389482656823,
|
||
|
|
"calib/mean_conf": 0.3936507936507936,
|
||
|
|
"calib/mu_c": 0.40529801324503306,
|
||
|
|
"calib/mu_w": 0.37623762376237624,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.005158730158730159,
|
||
|
|
"calib/std_conf": 0.08613945111882626,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0078125,
|
||
|
|
"completions/max_length": 1519.0,
|
||
|
|
"completions/max_terminated_length": 1519.0,
|
||
|
|
"completions/mean_length": 568.9140625,
|
||
|
|
"completions/mean_terminated_length": 573.3936767578125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 238.0,
|
||
|
|
"epoch": 0.20586666666666667,
|
||
|
|
"grad_norm": 0.013111384585499763,
|
||
|
|
"learning_rate": 1.9444444444444447e-07,
|
||
|
|
"loss": 0.0052,
|
||
|
|
"num_tokens": 40436698.0,
|
||
|
|
"reward": 1.4665391445159912,
|
||
|
|
"reward_std": 0.24395251274108887,
|
||
|
|
"rewards/accuracy_reward_step": 0.59375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7128125429153442,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8842817544937134,
|
||
|
|
"step": 193
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.622751124437781,
|
||
|
|
"calib/avg_num_step_conf": 1.03125,
|
||
|
|
"calib/ece": 0.17821522309711285,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.04470264867566215,
|
||
|
|
"calib/mean_conf": 0.3650918635170603,
|
||
|
|
"calib/mu_c": 0.3855072463768116,
|
||
|
|
"calib/mu_w": 0.34080459770114946,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.09950518591905787,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1691.0,
|
||
|
|
"completions/max_terminated_length": 1691.0,
|
||
|
|
"completions/mean_length": 594.44921875,
|
||
|
|
"completions/mean_terminated_length": 596.7804565429688,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 217.0,
|
||
|
|
"epoch": 0.20693333333333333,
|
||
|
|
"grad_norm": 0.010096187703311443,
|
||
|
|
"learning_rate": 1.6666666666666668e-07,
|
||
|
|
"loss": 0.0033,
|
||
|
|
"num_tokens": 40670501.0,
|
||
|
|
"reward": 1.4261138439178467,
|
||
|
|
"reward_std": 0.19851510226726532,
|
||
|
|
"rewards/accuracy_reward_step": 0.5390625,
|
||
|
|
"rewards/final_brier_reward_step": 0.723745584487915,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9147768020629883,
|
||
|
|
"step": 194
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.99609375,
|
||
|
|
"calib/auroc": 0.5836825396825397,
|
||
|
|
"calib/avg_num_step_conf": 1.02734375,
|
||
|
|
"calib/ece": 0.21058823529411766,
|
||
|
|
"calib/final_conf_rate": 0.99609375,
|
||
|
|
"calib/format_rate": 0.98828125,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.032095238095237955,
|
||
|
|
"calib/mean_conf": 0.38078431372549015,
|
||
|
|
"calib/mu_c": 0.3939999999999999,
|
||
|
|
"calib/mu_w": 0.36190476190476195,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.99609375,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.9921875,
|
||
|
|
"calib/pce": 0.0015686274509803923,
|
||
|
|
"calib/std_conf": 0.09279627987081357,
|
||
|
|
"calib/step_conf_rate": 0.9921875,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2938.0,
|
||
|
|
"completions/max_terminated_length": 2938.0,
|
||
|
|
"completions/mean_length": 604.9453125,
|
||
|
|
"completions/mean_terminated_length": 607.3176879882812,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 172.0,
|
||
|
|
"epoch": 0.208,
|
||
|
|
"grad_norm": 0.012461402453482151,
|
||
|
|
"learning_rate": 1.3888888888888888e-07,
|
||
|
|
"loss": 0.005,
|
||
|
|
"num_tokens": 40907031.0,
|
||
|
|
"reward": 1.4583547115325928,
|
||
|
|
"reward_std": 0.2264489382505417,
|
||
|
|
"rewards/accuracy_reward_step": 0.5859375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7117577791213989,
|
||
|
|
"rewards/format_reward_step": 0.98828125,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8802154064178467,
|
||
|
|
"step": 195
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.5707578145521408,
|
||
|
|
"calib/avg_num_step_conf": 1.0,
|
||
|
|
"calib/ece": 0.21445312500000008,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.029012345679012397,
|
||
|
|
"calib/mean_conf": 0.4183593750000001,
|
||
|
|
"calib/mu_c": 0.42901234567901236,
|
||
|
|
"calib/mu_w": 0.39999999999999997,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.0,
|
||
|
|
"calib/std_conf": 0.07815526757429325,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1503.0,
|
||
|
|
"completions/max_terminated_length": 1503.0,
|
||
|
|
"completions/mean_length": 479.5625,
|
||
|
|
"completions/mean_terminated_length": 481.44317626953125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 208.0,
|
||
|
|
"epoch": 0.20906666666666668,
|
||
|
|
"grad_norm": 0.01966879330575466,
|
||
|
|
"learning_rate": 1.1111111111111112e-07,
|
||
|
|
"loss": -0.0071,
|
||
|
|
"num_tokens": 41108023.0,
|
||
|
|
"reward": 1.5250952243804932,
|
||
|
|
"reward_std": 0.17813417315483093,
|
||
|
|
"rewards/accuracy_reward_step": 0.6328125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7290234565734863,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9110841751098633,
|
||
|
|
"step": 196
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 1.0,
|
||
|
|
"calib/auroc": 0.6216659446747942,
|
||
|
|
"calib/avg_num_step_conf": 1.00390625,
|
||
|
|
"calib/ece": 0.18750000000000006,
|
||
|
|
"calib/final_conf_rate": 1.0,
|
||
|
|
"calib/format_rate": 1.0,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.05338201621387456,
|
||
|
|
"calib/mean_conf": 0.38203125000000004,
|
||
|
|
"calib/mu_c": 0.4055944055944055,
|
||
|
|
"calib/mu_w": 0.35221238938053095,
|
||
|
|
"calib/nonempty_final_conf_rate": 1.0,
|
||
|
|
"calib/nonempty_reasoning_rate": 1.0,
|
||
|
|
"calib/nonempty_step_conf_rate": 1.0,
|
||
|
|
"calib/pce": 0.00546875,
|
||
|
|
"calib/std_conf": 0.09390287015548299,
|
||
|
|
"calib/step_conf_rate": 1.0,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 1963.0,
|
||
|
|
"completions/max_terminated_length": 1963.0,
|
||
|
|
"completions/mean_length": 636.0234375,
|
||
|
|
"completions/mean_terminated_length": 638.5177001953125,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 124.0,
|
||
|
|
"epoch": 0.21013333333333334,
|
||
|
|
"grad_norm": 0.011499550193548203,
|
||
|
|
"learning_rate": 8.333333333333334e-08,
|
||
|
|
"loss": -0.0092,
|
||
|
|
"num_tokens": 41351581.0,
|
||
|
|
"reward": 1.4608979225158691,
|
||
|
|
"reward_std": 0.24005961418151855,
|
||
|
|
"rewards/accuracy_reward_step": 0.55859375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7397655844688416,
|
||
|
|
"rewards/format_reward_step": 1.0,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9296857118606567,
|
||
|
|
"step": 197
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.9921875,
|
||
|
|
"calib/auroc": 0.5870192307692308,
|
||
|
|
"calib/avg_num_step_conf": 1.0078125,
|
||
|
|
"calib/ece": 0.2122047244094489,
|
||
|
|
"calib/final_conf_rate": 0.9921875,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.03142307692307689,
|
||
|
|
"calib/mean_conf": 0.3791338582677165,
|
||
|
|
"calib/mu_c": 0.39199999999999996,
|
||
|
|
"calib/mu_w": 0.3605769230769231,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.9921875,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.99609375,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.0003937007874015748,
|
||
|
|
"calib/std_conf": 0.07835436987695427,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2624.0,
|
||
|
|
"completions/max_terminated_length": 2624.0,
|
||
|
|
"completions/mean_length": 575.80859375,
|
||
|
|
"completions/mean_terminated_length": 578.0667114257812,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 160.0,
|
||
|
|
"epoch": 0.2112,
|
||
|
|
"grad_norm": 0.00982653722167015,
|
||
|
|
"learning_rate": 5.555555555555556e-08,
|
||
|
|
"loss": 0.0295,
|
||
|
|
"num_tokens": 41580052.0,
|
||
|
|
"reward": 1.4629411697387695,
|
||
|
|
"reward_std": 0.24547307193279266,
|
||
|
|
"rewards/accuracy_reward_step": 0.5859375,
|
||
|
|
"rewards/final_brier_reward_step": 0.7108592987060547,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.9050459265708923,
|
||
|
|
"step": 198
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.5179347826086957,
|
||
|
|
"calib/avg_num_step_conf": 0.99609375,
|
||
|
|
"calib/ece": 0.25515873015873014,
|
||
|
|
"calib/final_conf_rate": 0.984375,
|
||
|
|
"calib/format_rate": 0.97265625,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.008016304347826186,
|
||
|
|
"calib/mean_conf": 0.3876984126984127,
|
||
|
|
"calib/mu_c": 0.390625,
|
||
|
|
"calib/mu_w": 0.3826086956521738,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.984375,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.98828125,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.984375,
|
||
|
|
"calib/pce": 0.003968253968253968,
|
||
|
|
"calib/std_conf": 0.08845824679503829,
|
||
|
|
"calib/step_conf_rate": 0.984375,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.00390625,
|
||
|
|
"completions/max_length": 2545.0,
|
||
|
|
"completions/max_terminated_length": 2545.0,
|
||
|
|
"completions/mean_length": 639.46484375,
|
||
|
|
"completions/mean_terminated_length": 641.9725952148438,
|
||
|
|
"completions/min_length": 0.0,
|
||
|
|
"completions/min_terminated_length": 197.0,
|
||
|
|
"epoch": 0.21226666666666666,
|
||
|
|
"grad_norm": 0.012449882924556732,
|
||
|
|
"learning_rate": 2.777777777777778e-08,
|
||
|
|
"loss": 0.0039,
|
||
|
|
"num_tokens": 41823635.0,
|
||
|
|
"reward": 1.4842878580093384,
|
||
|
|
"reward_std": 0.2170136272907257,
|
||
|
|
"rewards/accuracy_reward_step": 0.62890625,
|
||
|
|
"rewards/final_brier_reward_step": 0.6845703125,
|
||
|
|
"rewards/format_reward_step": 0.97265625,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8851981163024902,
|
||
|
|
"step": 199
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"calib/answer_extract_rate": 0.98828125,
|
||
|
|
"calib/auroc": 0.7157153729071537,
|
||
|
|
"calib/avg_num_step_conf": 1.01953125,
|
||
|
|
"calib/ece": 0.32964426877470365,
|
||
|
|
"calib/final_conf_rate": 0.98828125,
|
||
|
|
"calib/format_rate": 0.984375,
|
||
|
|
"calib/frac_conf_gt_0.9": 0.0,
|
||
|
|
"calib/gap": 0.07338660578386608,
|
||
|
|
"calib/mean_conf": 0.39604743083003946,
|
||
|
|
"calib/mu_c": 0.4172222222222222,
|
||
|
|
"calib/mu_w": 0.34383561643835614,
|
||
|
|
"calib/nonempty_final_conf_rate": 0.98828125,
|
||
|
|
"calib/nonempty_reasoning_rate": 0.9921875,
|
||
|
|
"calib/nonempty_step_conf_rate": 0.98828125,
|
||
|
|
"calib/pce": 0.007114624505928857,
|
||
|
|
"calib/std_conf": 0.07892677926720582,
|
||
|
|
"calib/step_conf_rate": 0.98828125,
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 2503.0,
|
||
|
|
"completions/max_terminated_length": 2503.0,
|
||
|
|
"completions/mean_length": 621.4921875,
|
||
|
|
"completions/mean_terminated_length": 621.4921875,
|
||
|
|
"completions/min_length": 177.0,
|
||
|
|
"completions/min_terminated_length": 177.0,
|
||
|
|
"epoch": 0.21333333333333335,
|
||
|
|
"grad_norm": 0.013002941384911537,
|
||
|
|
"learning_rate": 0.0,
|
||
|
|
"loss": 0.009,
|
||
|
|
"num_tokens": 42066465.0,
|
||
|
|
"reward": 1.572641134262085,
|
||
|
|
"reward_std": 0.18552997708320618,
|
||
|
|
"rewards/accuracy_reward_step": 0.703125,
|
||
|
|
"rewards/final_brier_reward_step": 0.7069531083106995,
|
||
|
|
"rewards/format_reward_step": 0.984375,
|
||
|
|
"rewards/stepwise_brier_reward": 0.8829084038734436,
|
||
|
|
"step": 200
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 0.21333333333333335,
|
||
|
|
"step": 200,
|
||
|
|
"total_flos": 0.0,
|
||
|
|
"train_loss": 0.011818162081181072,
|
||
|
|
"train_runtime": 8954.9185,
|
||
|
|
"train_samples_per_second": 5.718,
|
||
|
|
"train_steps_per_second": 0.022
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"logging_steps": 1,
|
||
|
|
"max_steps": 200,
|
||
|
|
"num_input_tokens_seen": 42066465,
|
||
|
|
"num_train_epochs": 1,
|
||
|
|
"save_steps": 20,
|
||
|
|
"stateful_callbacks": {
|
||
|
|
"TrainerControl": {
|
||
|
|
"args": {
|
||
|
|
"should_epoch_stop": false,
|
||
|
|
"should_evaluate": false,
|
||
|
|
"should_log": false,
|
||
|
|
"should_save": true,
|
||
|
|
"should_training_stop": true
|
||
|
|
},
|
||
|
|
"attributes": {}
|
||
|
|
}
|
||
|
|
},
|
||
|
|
"total_flos": 0.0,
|
||
|
|
"train_batch_size": 8,
|
||
|
|
"trial_name": null,
|
||
|
|
"trial_params": null
|
||
|
|
}
|