Files
RLCR-v4-ks-uniqueness-cov0-…/trainer_state.json
ModelHub XC b521eb5a4c 初始化项目,由ModelHub XC社区提供模型
Model: hector-gr/RLCR-v4-ks-uniqueness-cov0-entropy100-noece-noaurc-scaletrue-highcov-accgated-hotpot
Source: Original Platform
2026-05-09 19:39:27 +08:00

9805 lines
614 KiB
JSON

{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.9984,
"eval_steps": 50,
"global_step": 312,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"calibration/aurc": 0.6369179374624646,
"calibration/batch_distribution_entropy": 0.6544487092366394,
"calibration/batch_entropy_100bins": 0.4802970152360862,
"calibration/batch_entropy_10bins": 0.6544487092366394,
"calibration/batch_entropy_50bins": 0.5623008598890478,
"calibration/batch_uniqueness": 0.7240216143942608,
"calibration/confidence_entropy": 0.34818559548888905,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.0,
"calibration/coverage@25%": 0.0,
"calibration/coverage@30%": 0.0,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.6544487092366394,
"calibration/distribution_entropy_100": 0.4802970152360862,
"calibration/ece": 0.49848305354173855,
"calibration/mean_confidence": 0.7906346560520442,
"calibration/unique_confidence_per_question": 0.03203125,
"calibration/unique_confidences": 16.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0365234375,
"completions/max_length": 1507.2,
"completions/max_terminated_length": 1507.2,
"completions/mean_length": 216.43837890625,
"completions/mean_terminated_length": 224.64193115234374,
"completions/min_length": 0.0,
"completions/min_terminated_length": 2.0,
"epoch": 0.016,
"grad_norm": 0.3191711902618408,
"learning_rate": 3.1249999999999997e-07,
"loss": 0.0256,
"num_tokens": 17060361.0,
"reward": 0.6802164673805237,
"reward_std": 0.5657365441322326,
"rewards/accgated_coverage_0": 0.3000900387763977,
"rewards/accgated_coverage_1": 0.3000900387763977,
"rewards/accgated_coverage_10": 0.3000900387763977,
"rewards/accgated_coverage_15": 0.3000900387763977,
"rewards/accgated_coverage_20": 0.3000900387763977,
"rewards/accgated_coverage_25": 0.3000900387763977,
"rewards/accgated_coverage_5": 0.3000900387763977,
"rewards/accuracy_reward": 0.21787109375,
"rewards/brier_reward": 0.37398988008499146,
"rewards/confidence_uniqueness_reward": 0.48599093556404116,
"rewards/format_reward": 0.6810546875,
"rewards/frontier_entropy_batch_reward": -0.6530755043029786,
"signal/accgated_coverage_0/centered_abs_mean": 0.2924555778503418,
"signal/accgated_coverage_0/group_std_mean": 0.34396286606788634,
"signal/accgated_coverage_0/group_zero_std_frac": 0.003125,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.05832238420844078,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.029245557636022566,
"signal/accgated_coverage_1/centered_abs_mean": 0.2924555778503418,
"signal/accgated_coverage_1/group_std_mean": 0.34396286606788634,
"signal/accgated_coverage_1/group_zero_std_frac": 0.003125,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.05832238420844078,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.029245557636022566,
"signal/accgated_coverage_10/centered_abs_mean": 0.2924555778503418,
"signal/accgated_coverage_10/group_std_mean": 0.34396286606788634,
"signal/accgated_coverage_10/group_zero_std_frac": 0.003125,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.05832238420844078,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.029245557636022566,
"signal/accgated_coverage_15/centered_abs_mean": 0.2924555778503418,
"signal/accgated_coverage_15/group_std_mean": 0.34396286606788634,
"signal/accgated_coverage_15/group_zero_std_frac": 0.003125,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05832238420844078,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.029245557636022566,
"signal/accgated_coverage_20/centered_abs_mean": 0.2924555778503418,
"signal/accgated_coverage_20/group_std_mean": 0.34396286606788634,
"signal/accgated_coverage_20/group_zero_std_frac": 0.003125,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.05832238420844078,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.029245557636022566,
"signal/accgated_coverage_25/centered_abs_mean": 0.2924555778503418,
"signal/accgated_coverage_25/group_std_mean": 0.34396286606788634,
"signal/accgated_coverage_25/group_zero_std_frac": 0.003125,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.05832238420844078,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.029245557636022566,
"signal/accgated_coverage_5/centered_abs_mean": 0.2924555778503418,
"signal/accgated_coverage_5/group_std_mean": 0.34396286606788634,
"signal/accgated_coverage_5/group_zero_std_frac": 0.003125,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.05832238420844078,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.029245557636022566,
"signal/accuracy_reward/centered_abs_mean": 0.240631103515625,
"signal/accuracy_reward/group_std_mean": 0.2825372278690338,
"signal/accuracy_reward/group_zero_std_frac": 0.31875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.23846413791179658,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.1203155517578125,
"signal/advantage_abs_mean": 0.8286440968513489,
"signal/advantage_pre_scale_abs_mean": 0.47917195558547976,
"signal/advantage_pre_scale_std": 0.5831026792526245,
"signal/advantage_std": 0.984294056892395,
"signal/brier_reward/centered_abs_mean": 0.31949453353881835,
"signal/brier_reward/group_std_mean": 0.36506720781326296,
"signal/brier_reward/group_zero_std_frac": 0.003125,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.06380620226264,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.03194945380091667,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.298015820980072,
"signal/confidence_uniqueness_reward/group_std_mean": 0.3490760505199432,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.05968900471925735,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.029801582172513007,
"signal/format_reward/centered_abs_mean": 0.4048583984375,
"signal/format_reward/group_std_mean": 0.4546546757221222,
"signal/format_reward/group_zero_std_frac": 0.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.4052778363227844,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.20242919921875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.4261767566204071,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.47186354994773866,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.08531290292739868,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.04261767566204071,
"step": 5
},
{
"calibration/aurc": 0.6669354113138501,
"calibration/batch_distribution_entropy": 0.6540258655475888,
"calibration/batch_entropy_100bins": 0.48497332806195087,
"calibration/batch_entropy_10bins": 0.6540258655475888,
"calibration/batch_entropy_50bins": 0.5666634646401097,
"calibration/batch_uniqueness": 0.7307180399404241,
"calibration/confidence_entropy": 0.3438339658444833,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.0,
"calibration/coverage@25%": 0.0,
"calibration/coverage@30%": 0.0,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.6540258655475888,
"calibration/distribution_entropy_100": 0.48497332806195087,
"calibration/ece": 0.5078477034527543,
"calibration/mean_confidence": 0.788065953476565,
"calibration/unique_confidence_per_question": 0.03515625,
"calibration/unique_confidences": 18.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0341796875,
"completions/max_length": 1519.4,
"completions/max_terminated_length": 1519.4,
"completions/mean_length": 204.69599609375,
"completions/mean_terminated_length": 211.93881225585938,
"completions/min_length": 0.0,
"completions/min_terminated_length": 2.0,
"epoch": 0.032,
"grad_norm": 0.3231641352176666,
"learning_rate": 6.249999999999999e-07,
"loss": 0.0072,
"num_tokens": 34256800.0,
"reward": 0.7205589890480042,
"reward_std": 0.5330297231674195,
"rewards/accgated_coverage_0": 0.31446969509124756,
"rewards/accgated_coverage_1": 0.31446969509124756,
"rewards/accgated_coverage_10": 0.31446969509124756,
"rewards/accgated_coverage_15": 0.31446969509124756,
"rewards/accgated_coverage_20": 0.31446969509124756,
"rewards/accgated_coverage_25": 0.31446969509124756,
"rewards/accgated_coverage_5": 0.31446969509124756,
"rewards/accuracy_reward": 0.22109375,
"rewards/brier_reward": 0.39473122358322144,
"rewards/confidence_uniqueness_reward": 0.5247102797031402,
"rewards/format_reward": 0.7365234375,
"rewards/frontier_entropy_batch_reward": -0.7032256126403809,
"signal/accgated_coverage_0/centered_abs_mean": 0.2804561495780945,
"signal/accgated_coverage_0/group_std_mean": 0.3354907214641571,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.05935393273830414,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.028045616298913955,
"signal/accgated_coverage_1/centered_abs_mean": 0.2804561495780945,
"signal/accgated_coverage_1/group_std_mean": 0.3354907214641571,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.05935393273830414,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.028045616298913955,
"signal/accgated_coverage_10/centered_abs_mean": 0.2804561495780945,
"signal/accgated_coverage_10/group_std_mean": 0.3354907214641571,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.05935393273830414,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.028045616298913955,
"signal/accgated_coverage_15/centered_abs_mean": 0.2804561495780945,
"signal/accgated_coverage_15/group_std_mean": 0.3354907214641571,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05935393273830414,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.028045616298913955,
"signal/accgated_coverage_20/centered_abs_mean": 0.2804561495780945,
"signal/accgated_coverage_20/group_std_mean": 0.3354907214641571,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.05935393273830414,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.028045616298913955,
"signal/accgated_coverage_25/centered_abs_mean": 0.2804561495780945,
"signal/accgated_coverage_25/group_std_mean": 0.3354907214641571,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.05935393273830414,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.028045616298913955,
"signal/accgated_coverage_5/centered_abs_mean": 0.2804561495780945,
"signal/accgated_coverage_5/group_std_mean": 0.3354907214641571,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.05935393273830414,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.028045616298913955,
"signal/accuracy_reward/centered_abs_mean": 0.2223388671875,
"signal/accuracy_reward/group_std_mean": 0.26838217973709105,
"signal/accuracy_reward/group_zero_std_frac": 0.325,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.23507187962532045,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.11116943359375,
"signal/advantage_abs_mean": 0.8027427673339844,
"signal/advantage_pre_scale_abs_mean": 0.43977790474891665,
"signal/advantage_pre_scale_std": 0.5510475993156433,
"signal/advantage_std": 0.9842826008796692,
"signal/brier_reward/centered_abs_mean": 0.30625501871109007,
"signal/brier_reward/group_std_mean": 0.35521684885025023,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.06480256989598274,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.03062550202012062,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.27534674406051635,
"signal/confidence_uniqueness_reward/group_std_mean": 0.333311527967453,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.05818134024739265,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0275346752256155,
"signal/format_reward/centered_abs_mean": 0.35655517578125,
"signal/format_reward/group_std_mean": 0.42446200251579286,
"signal/format_reward/group_zero_std_frac": 0.0,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.3764032065868378,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.178277587890625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.387843918800354,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.45033844113349913,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.08198202103376388,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.03878439143300057,
"step": 10
},
{
"calibration/aurc": 0.5845214219403985,
"calibration/batch_distribution_entropy": 0.6655310308077376,
"calibration/batch_entropy_100bins": 0.48673023491865663,
"calibration/batch_entropy_10bins": 0.6655310308077376,
"calibration/batch_entropy_50bins": 0.5686679566607937,
"calibration/batch_uniqueness": 0.7217716714165819,
"calibration/buffer_distribution_entropy": 0.6711236351590234,
"calibration/buffer_entropy_100bins": 0.49463996292765533,
"calibration/buffer_entropy_10bins": 0.6711236351590234,
"calibration/buffer_entropy_50bins": 0.5778489809993154,
"calibration/confidence_entropy": 0.36437029673633065,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.0,
"calibration/coverage@25%": 0.0,
"calibration/coverage@30%": 0.0,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.6655310308077376,
"calibration/distribution_entropy_100": 0.48673023491865663,
"calibration/ece": 0.4528321319023771,
"calibration/mean_confidence": 0.786073370374098,
"calibration/unique_confidence_per_question": 0.040625,
"calibration/unique_confidences": 20.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.023046875,
"completions/max_length": 1498.0,
"completions/max_terminated_length": 1498.0,
"completions/mean_length": 170.6353515625,
"completions/mean_terminated_length": 174.8092071533203,
"completions/min_length": 0.0,
"completions/min_terminated_length": 2.6,
"epoch": 0.048,
"grad_norm": 0.05429961904883385,
"learning_rate": 9.374999999999999e-07,
"loss": 0.0016,
"num_tokens": 51052842.0,
"reward": 0.8294067621231079,
"reward_std": 0.41685991287231444,
"rewards/accgated_coverage_0": 0.31299112644046545,
"rewards/accgated_coverage_1": 0.31299112644046545,
"rewards/accgated_coverage_10": 0.31299112644046545,
"rewards/accgated_coverage_15": 0.31299112644046545,
"rewards/accgated_coverage_20": 0.31299112644046545,
"rewards/accgated_coverage_25": 0.31299112644046545,
"rewards/accgated_coverage_5": 0.31299112644046545,
"rewards/accuracy_reward": 0.273828125,
"rewards/brier_reward": 0.5012387871742249,
"rewards/confidence_uniqueness_reward": 0.6541430830955506,
"rewards/format_reward": 0.88193359375,
"rewards/frontier_entropy_batch_reward": -0.8310608983039856,
"signal/accgated_coverage_0/centered_abs_mean": 0.22104704901576042,
"signal/accgated_coverage_0/group_std_mean": 0.26555165722966195,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.05563354790210724,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.02210470554418862,
"signal/accgated_coverage_1/centered_abs_mean": 0.22104704901576042,
"signal/accgated_coverage_1/group_std_mean": 0.26555165722966195,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.05563354790210724,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.02210470554418862,
"signal/accgated_coverage_10/centered_abs_mean": 0.22104704901576042,
"signal/accgated_coverage_10/group_std_mean": 0.26555165722966195,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.05563354790210724,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.02210470554418862,
"signal/accgated_coverage_15/centered_abs_mean": 0.22104704901576042,
"signal/accgated_coverage_15/group_std_mean": 0.26555165722966195,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05563354790210724,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.02210470554418862,
"signal/accgated_coverage_20/centered_abs_mean": 0.22104704901576042,
"signal/accgated_coverage_20/group_std_mean": 0.26555165722966195,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.05563354790210724,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.02210470554418862,
"signal/accgated_coverage_25/centered_abs_mean": 0.22104704901576042,
"signal/accgated_coverage_25/group_std_mean": 0.26555165722966195,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.05563354790210724,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.02210470554418862,
"signal/accgated_coverage_5/centered_abs_mean": 0.22104704901576042,
"signal/accgated_coverage_5/group_std_mean": 0.26555165722966195,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.05563354790210724,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.02210470554418862,
"signal/accuracy_reward/centered_abs_mean": 0.20126953125,
"signal/accuracy_reward/group_std_mean": 0.25166312158107756,
"signal/accuracy_reward/group_zero_std_frac": 0.340625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.32761208415031434,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.100634765625,
"signal/advantage_abs_mean": 0.7497124195098877,
"signal/advantage_pre_scale_abs_mean": 0.32911559343338015,
"signal/advantage_pre_scale_std": 0.43396281599998476,
"signal/advantage_std": 0.9841608762741089,
"signal/brier_reward/centered_abs_mean": 0.27257916927337644,
"signal/brier_reward/group_std_mean": 0.3263565719127655,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.08823587894439697,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.02725791819393635,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.18984188437461852,
"signal/confidence_uniqueness_reward/group_std_mean": 0.25445507764816283,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.057601611316204074,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.01898418851196766,
"signal/format_reward/centered_abs_mean": 0.190679931640625,
"signal/format_reward/group_std_mean": 0.2829218327999115,
"signal/format_reward/group_zero_std_frac": 0.125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.26141563057899475,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0953399658203125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2620670020580292,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3645237565040588,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.021875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.07975108250975609,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02620669938623905,
"step": 15
},
{
"calibration/aurc": 0.5226028450449948,
"calibration/batch_distribution_entropy": 0.8157085947501281,
"calibration/batch_entropy_100bins": 0.5598161244489184,
"calibration/batch_entropy_10bins": 0.8157085947501281,
"calibration/batch_entropy_50bins": 0.6495203344870534,
"calibration/batch_uniqueness": 0.7932460733858593,
"calibration/buffer_distribution_entropy": 0.6895597806789104,
"calibration/buffer_entropy_100bins": 0.5034696601544146,
"calibration/buffer_entropy_10bins": 0.6895597806789104,
"calibration/buffer_entropy_50bins": 0.5874436816277322,
"calibration/confidence_entropy": 0.4358751798675248,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.0,
"calibration/coverage@25%": 0.0,
"calibration/coverage@30%": 0.0,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.8157085947501281,
"calibration/distribution_entropy_100": 0.5598161244489184,
"calibration/ece": 0.2793788661180069,
"calibration/mean_confidence": 0.6612075107813843,
"calibration/unique_confidence_per_question": 0.06015625,
"calibration/unique_confidences": 30.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00419921875,
"completions/max_length": 1218.8,
"completions/max_terminated_length": 1218.8,
"completions/mean_length": 131.40615234375,
"completions/mean_terminated_length": 131.96847229003907,
"completions/min_length": 0.0,
"completions/min_terminated_length": 10.0,
"epoch": 0.064,
"grad_norm": 0.1383822113275528,
"learning_rate": 1e-06,
"loss": -0.0202,
"num_tokens": 67316841.0,
"reward": 0.7213197231292725,
"reward_std": 0.17918029129505159,
"rewards/accgated_coverage_0": 0.012151591293513776,
"rewards/accgated_coverage_1": 0.012151591293513776,
"rewards/accgated_coverage_10": 0.012151591293513776,
"rewards/accgated_coverage_15": 0.012151591293513776,
"rewards/accgated_coverage_20": 0.012151591293513776,
"rewards/accgated_coverage_25": 0.012151591293513776,
"rewards/accgated_coverage_5": 0.012151591293513776,
"rewards/accuracy_reward": 0.3421875,
"rewards/brier_reward": 0.646466600894928,
"rewards/confidence_uniqueness_reward": 0.7833222866058349,
"rewards/format_reward": 0.97734375,
"rewards/frontier_entropy_batch_reward": -0.8993089914321899,
"signal/accgated_coverage_0/centered_abs_mean": 0.02158209141343832,
"signal/accgated_coverage_0/group_std_mean": 0.031163923814892767,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.016024397686123848,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0021582091692835094,
"signal/accgated_coverage_1/centered_abs_mean": 0.02158209141343832,
"signal/accgated_coverage_1/group_std_mean": 0.031163923814892767,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.016024397686123848,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0021582091692835094,
"signal/accgated_coverage_10/centered_abs_mean": 0.02158209141343832,
"signal/accgated_coverage_10/group_std_mean": 0.031163923814892767,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.016024397686123848,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0021582091692835094,
"signal/accgated_coverage_15/centered_abs_mean": 0.02158209141343832,
"signal/accgated_coverage_15/group_std_mean": 0.031163923814892767,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.016024397686123848,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0021582091692835094,
"signal/accgated_coverage_20/centered_abs_mean": 0.02158209141343832,
"signal/accgated_coverage_20/group_std_mean": 0.031163923814892767,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.016024397686123848,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0021582091692835094,
"signal/accgated_coverage_25/centered_abs_mean": 0.02158209141343832,
"signal/accgated_coverage_25/group_std_mean": 0.031163923814892767,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.016024397686123848,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0021582091692835094,
"signal/accgated_coverage_5/centered_abs_mean": 0.02158209141343832,
"signal/accgated_coverage_5/group_std_mean": 0.031163923814892767,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.016024397686123848,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0021582091692835094,
"signal/accuracy_reward/centered_abs_mean": 0.19715576171875,
"signal/accuracy_reward/group_std_mean": 0.24958559572696687,
"signal/accuracy_reward/group_zero_std_frac": 0.334375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7264934659004212,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.098577880859375,
"signal/advantage_abs_mean": 0.7055983662605285,
"signal/advantage_pre_scale_abs_mean": 0.13239798396825792,
"signal/advantage_pre_scale_std": 0.19309431314468384,
"signal/advantage_std": 0.9837452530860901,
"signal/brier_reward/centered_abs_mean": 0.23120042979717254,
"signal/brier_reward/group_std_mean": 0.28611327409744264,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17408487796783448,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.023120043426752092,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.11088034957647323,
"signal/confidence_uniqueness_reward/group_std_mean": 0.15136729925870895,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.08330316171050071,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.011088035069406033,
"signal/format_reward/centered_abs_mean": 0.0405517578125,
"signal/format_reward/group_std_mean": 0.09191280752420425,
"signal/format_reward/group_zero_std_frac": 0.559375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.1461832731962204,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.02027587890625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.17341790199279786,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.2991870820522308,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.06875,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.13263704776763915,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0173417903482914,
"step": 20
},
{
"calibration/aurc": 0.6530375665324215,
"calibration/batch_distribution_entropy": 0.9169398257072003,
"calibration/batch_entropy_100bins": 0.7480897335648299,
"calibration/batch_entropy_10bins": 0.9169398257072003,
"calibration/batch_entropy_50bins": 0.8159506209188093,
"calibration/batch_uniqueness": 0.8852481109497786,
"calibration/buffer_distribution_entropy": 0.7872363824408404,
"calibration/buffer_entropy_100bins": 0.5662417953154824,
"calibration/buffer_entropy_10bins": 0.7872363824408404,
"calibration/buffer_entropy_50bins": 0.6546115137639402,
"calibration/confidence_entropy": 0.47576488492860325,
"calibration/coverage@0%": 0.0,
"calibration/coverage@1%": 0.0,
"calibration/coverage@10%": 0.0,
"calibration/coverage@15%": 0.0,
"calibration/coverage@20%": 0.0,
"calibration/coverage@25%": 0.0,
"calibration/coverage@30%": 0.0,
"calibration/coverage@5%": 0.0,
"calibration/distribution_entropy_10": 0.9169398257072003,
"calibration/distribution_entropy_100": 0.7480897335648299,
"calibration/ece": 0.20308442109150784,
"calibration/mean_confidence": 0.39197013951131837,
"calibration/unique_confidence_per_question": 0.20234375,
"calibration/unique_confidences": 103.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00419921875,
"completions/max_length": 1086.8,
"completions/max_terminated_length": 1086.8,
"completions/mean_length": 116.26494140625,
"completions/mean_terminated_length": 116.75207977294922,
"completions/min_length": 0.0,
"completions/min_terminated_length": 9.4,
"epoch": 0.08,
"grad_norm": 0.09559965133666992,
"learning_rate": 1e-06,
"loss": -0.0244,
"num_tokens": 83440546.0,
"reward": 0.7711252927780151,
"reward_std": 0.1475716471672058,
"rewards/accgated_coverage_0": 0.01681157825514674,
"rewards/accgated_coverage_1": 0.01681157825514674,
"rewards/accgated_coverage_10": 0.01681157825514674,
"rewards/accgated_coverage_15": 0.01681157825514674,
"rewards/accgated_coverage_20": 0.01681157825514674,
"rewards/accgated_coverage_25": 0.01681157825514674,
"rewards/accgated_coverage_5": 0.01681157825514674,
"rewards/accuracy_reward": 0.33994140625,
"rewards/brier_reward": 0.7291219353675842,
"rewards/confidence_uniqueness_reward": 0.8762930989265442,
"rewards/format_reward": 0.987890625,
"rewards/frontier_entropy_batch_reward": -0.6510035574436188,
"signal/accgated_coverage_0/centered_abs_mean": 0.0394746832549572,
"signal/accgated_coverage_0/group_std_mean": 0.050183466821908954,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.03596835248172283,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.003947468474507332,
"signal/accgated_coverage_1/centered_abs_mean": 0.0394746832549572,
"signal/accgated_coverage_1/group_std_mean": 0.050183466821908954,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.03596835248172283,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.003947468474507332,
"signal/accgated_coverage_10/centered_abs_mean": 0.0394746832549572,
"signal/accgated_coverage_10/group_std_mean": 0.050183466821908954,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.03596835248172283,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003947468474507332,
"signal/accgated_coverage_15/centered_abs_mean": 0.0394746832549572,
"signal/accgated_coverage_15/group_std_mean": 0.050183466821908954,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.03596835248172283,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003947468474507332,
"signal/accgated_coverage_20/centered_abs_mean": 0.0394746832549572,
"signal/accgated_coverage_20/group_std_mean": 0.050183466821908954,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03596835248172283,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.003947468474507332,
"signal/accgated_coverage_25/centered_abs_mean": 0.0394746832549572,
"signal/accgated_coverage_25/group_std_mean": 0.050183466821908954,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03596835248172283,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.003947468474507332,
"signal/accgated_coverage_5/centered_abs_mean": 0.0394746832549572,
"signal/accgated_coverage_5/group_std_mean": 0.050183466821908954,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.03596835248172283,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.003947468474507332,
"signal/accuracy_reward/centered_abs_mean": 0.193365478515625,
"signal/accuracy_reward/group_std_mean": 0.24426858127117157,
"signal/accuracy_reward/group_zero_std_frac": 0.353125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.8779962182044982,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0966827392578125,
"signal/advantage_abs_mean": 0.7505025267601013,
"signal/advantage_pre_scale_abs_mean": 0.11283858716487885,
"signal/advantage_pre_scale_std": 0.16124872267246246,
"signal/advantage_std": 0.9835973143577575,
"signal/brier_reward/centered_abs_mean": 0.20754425525665282,
"signal/brier_reward/group_std_mean": 0.2633790075778961,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18865779936313629,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.02075442597270012,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.08122085779905319,
"signal/confidence_uniqueness_reward/group_std_mean": 0.10770698338747024,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.07371854782104492,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.008122085314244031,
"signal/format_reward/centered_abs_mean": 0.01826171875,
"signal/format_reward/group_std_mean": 0.036469937115907666,
"signal/format_reward/group_zero_std_frac": 0.834375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.08346155881881714,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.009130859375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.4041131794452667,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.4963274121284485,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.003125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.3669182360172272,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.04041131883859635,
"step": 25
},
{
"calibration/aurc": 0.6624974602632341,
"calibration/batch_distribution_entropy": 0.8124735735427375,
"calibration/batch_entropy_100bins": 0.8750873779109286,
"calibration/batch_entropy_10bins": 0.8124735735427375,
"calibration/batch_entropy_50bins": 0.8715266836984741,
"calibration/batch_uniqueness": 0.9185722562656657,
"calibration/buffer_distribution_entropy": 0.8849393060327049,
"calibration/buffer_entropy_100bins": 0.6852681120906572,
"calibration/buffer_entropy_10bins": 0.8849393060327049,
"calibration/buffer_entropy_50bins": 0.7634561651669772,
"calibration/confidence_entropy": 0.43257229173609046,
"calibration/coverage@0%": 0.0003929273084479371,
"calibration/coverage@1%": 0.0003929273084479371,
"calibration/coverage@10%": 0.0003929273084479371,
"calibration/coverage@15%": 0.0003929273084479371,
"calibration/coverage@20%": 0.0003929273084479371,
"calibration/coverage@25%": 0.0003929273084479371,
"calibration/coverage@30%": 0.007059593975114603,
"calibration/coverage@5%": 0.0003929273084479371,
"calibration/distribution_entropy_10": 0.8124735735427375,
"calibration/distribution_entropy_100": 0.8750873779109286,
"calibration/ece": 0.1876806253437428,
"calibration/mean_confidence": 0.24786234308538582,
"calibration/unique_confidence_per_question": 0.77109375,
"calibration/unique_confidences": 394.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.005078125,
"completions/max_length": 1022.8,
"completions/max_terminated_length": 1022.8,
"completions/mean_length": 109.1837890625,
"completions/mean_terminated_length": 109.74313659667969,
"completions/min_length": 0.0,
"completions/min_terminated_length": 17.8,
"epoch": 0.096,
"grad_norm": 0.6592833995819092,
"learning_rate": 1e-06,
"loss": -0.0251,
"num_tokens": 99603196.0,
"reward": 0.7912370562553406,
"reward_std": 0.1335388630628586,
"rewards/accgated_coverage_0": 0.021468784660100937,
"rewards/accgated_coverage_1": 0.021468784660100937,
"rewards/accgated_coverage_10": 0.021468784660100937,
"rewards/accgated_coverage_15": 0.021468784660100937,
"rewards/accgated_coverage_20": 0.021468784660100937,
"rewards/accgated_coverage_25": 0.021468784660100937,
"rewards/accgated_coverage_5": 0.021468784660100937,
"rewards/accuracy_reward": 0.3462890625,
"rewards/brier_reward": 0.7245596408843994,
"rewards/confidence_uniqueness_reward": 0.910044002532959,
"rewards/format_reward": 0.9875,
"rewards/frontier_entropy_batch_reward": -0.5414599895477294,
"signal/accgated_coverage_0/centered_abs_mean": 0.04524085596203804,
"signal/accgated_coverage_0/group_std_mean": 0.05522804409265518,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.05592176169157028,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004524085624143481,
"signal/accgated_coverage_1/centered_abs_mean": 0.04524085596203804,
"signal/accgated_coverage_1/group_std_mean": 0.05522804409265518,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.05592176169157028,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004524085624143481,
"signal/accgated_coverage_10/centered_abs_mean": 0.04524085596203804,
"signal/accgated_coverage_10/group_std_mean": 0.05522804409265518,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.05592176169157028,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004524085624143481,
"signal/accgated_coverage_15/centered_abs_mean": 0.04524085596203804,
"signal/accgated_coverage_15/group_std_mean": 0.05522804409265518,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05592176169157028,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004524085624143481,
"signal/accgated_coverage_20/centered_abs_mean": 0.04524085596203804,
"signal/accgated_coverage_20/group_std_mean": 0.05522804409265518,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.05592176169157028,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.004524085624143481,
"signal/accgated_coverage_25/centered_abs_mean": 0.04524085596203804,
"signal/accgated_coverage_25/group_std_mean": 0.05522804409265518,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.05592176169157028,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.004524085624143481,
"signal/accgated_coverage_5/centered_abs_mean": 0.04524085596203804,
"signal/accgated_coverage_5/group_std_mean": 0.05522804409265518,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.05592176169157028,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004524085624143481,
"signal/accuracy_reward/centered_abs_mean": 0.189306640625,
"signal/accuracy_reward/group_std_mean": 0.23722708523273467,
"signal/accuracy_reward/group_zero_std_frac": 0.375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.1745991706848145,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0946533203125,
"signal/advantage_abs_mean": 0.732611107826233,
"signal/advantage_pre_scale_abs_mean": 0.0980033040046692,
"signal/advantage_pre_scale_std": 0.15151599943637847,
"signal/advantage_std": 0.9832700371742249,
"signal/brier_reward/centered_abs_mean": 0.20171859860420227,
"signal/brier_reward/group_std_mean": 0.2550782173871994,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.2500972092151642,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.020171860232949257,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.049207459390163424,
"signal/confidence_uniqueness_reward/group_std_mean": 0.078475920855999,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0607869565486908,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004920745920389891,
"signal/format_reward/centered_abs_mean": 0.02218017578125,
"signal/format_reward/group_std_mean": 0.04925174191594124,
"signal/format_reward/group_zero_std_frac": 0.7625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.13655894845724106,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.011090087890625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.45344639420509336,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.5132026433944702,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5628592371940613,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.04534464031457901,
"step": 30
},
{
"calibration/aurc": 0.5471933123926485,
"calibration/batch_distribution_entropy": 0.9532739169200232,
"calibration/batch_entropy_100bins": 0.9535472720446684,
"calibration/batch_entropy_10bins": 0.9532739169200232,
"calibration/batch_entropy_50bins": 0.9605380064253927,
"calibration/batch_uniqueness": 0.9484135706576053,
"calibration/buffer_distribution_entropy": 0.9322783167621298,
"calibration/buffer_entropy_100bins": 0.7762173508473715,
"calibration/buffer_entropy_10bins": 0.9322783167621298,
"calibration/buffer_entropy_50bins": 0.837588344198819,
"calibration/confidence_entropy": 0.5519745383798715,
"calibration/coverage@0%": 0.004315314878843422,
"calibration/coverage@1%": 0.004315314878843422,
"calibration/coverage@10%": 0.004315314878843422,
"calibration/coverage@15%": 0.004706704311328745,
"calibration/coverage@20%": 0.006272262041270037,
"calibration/coverage@25%": 0.008229209203696651,
"calibration/coverage@30%": 0.023995950365590605,
"calibration/coverage@5%": 0.004315314878843422,
"calibration/distribution_entropy_10": 0.9532739169200232,
"calibration/distribution_entropy_100": 0.9535472720446684,
"calibration/ece": 0.1871538553902457,
"calibration/mean_confidence": 0.4430844668477844,
"calibration/unique_confidence_per_question": 0.9890625,
"calibration/unique_confidences": 506.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00263671875,
"completions/max_length": 984.8,
"completions/max_terminated_length": 984.8,
"completions/mean_length": 105.637109375,
"completions/mean_terminated_length": 105.9163101196289,
"completions/min_length": 0.0,
"completions/min_terminated_length": 13.0,
"epoch": 0.112,
"grad_norm": 0.2867770195007324,
"learning_rate": 1e-06,
"loss": -0.0202,
"num_tokens": 115794392.0,
"reward": 0.8346109986305237,
"reward_std": 0.13620110154151915,
"rewards/accgated_coverage_0": 0.00985442753881216,
"rewards/accgated_coverage_1": 0.00985442753881216,
"rewards/accgated_coverage_10": 0.00985442753881216,
"rewards/accgated_coverage_15": 0.00985442753881216,
"rewards/accgated_coverage_20": 0.00985442753881216,
"rewards/accgated_coverage_25": 0.00985442753881216,
"rewards/accgated_coverage_5": 0.00985442753881216,
"rewards/accuracy_reward": 0.38779296875,
"rewards/brier_reward": 0.7143089652061463,
"rewards/confidence_uniqueness_reward": 0.9423018455505371,
"rewards/format_reward": 0.9939453125,
"rewards/frontier_entropy_batch_reward": -0.2881730914115906,
"signal/accgated_coverage_0/centered_abs_mean": 0.04202499836683273,
"signal/accgated_coverage_0/group_std_mean": 0.05275702103972435,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.04252778291702271,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004202499752864242,
"signal/accgated_coverage_1/centered_abs_mean": 0.04202499836683273,
"signal/accgated_coverage_1/group_std_mean": 0.05275702103972435,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.04252778291702271,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004202499752864242,
"signal/accgated_coverage_10/centered_abs_mean": 0.04202499836683273,
"signal/accgated_coverage_10/group_std_mean": 0.05275702103972435,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.04252778291702271,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004202499752864242,
"signal/accgated_coverage_15/centered_abs_mean": 0.04202499836683273,
"signal/accgated_coverage_15/group_std_mean": 0.05275702103972435,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.04252778291702271,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004202499752864242,
"signal/accgated_coverage_20/centered_abs_mean": 0.04202499836683273,
"signal/accgated_coverage_20/group_std_mean": 0.05275702103972435,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.04252778291702271,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.004202499752864242,
"signal/accgated_coverage_25/centered_abs_mean": 0.04202499836683273,
"signal/accgated_coverage_25/group_std_mean": 0.05275702103972435,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.04252778291702271,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.004202499752864242,
"signal/accgated_coverage_5/centered_abs_mean": 0.04202499836683273,
"signal/accgated_coverage_5/group_std_mean": 0.05275702103972435,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.04252778291702271,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004202499752864242,
"signal/accuracy_reward/centered_abs_mean": 0.185931396484375,
"signal/accuracy_reward/group_std_mean": 0.23456257879734038,
"signal/accuracy_reward/group_zero_std_frac": 0.365625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.965952455997467,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0929656982421875,
"signal/advantage_abs_mean": 0.7708975553512574,
"signal/advantage_pre_scale_abs_mean": 0.10621223896741867,
"signal/advantage_pre_scale_std": 0.15235058665275575,
"signal/advantage_std": 0.9834662795066833,
"signal/brier_reward/centered_abs_mean": 0.19554105401039124,
"signal/brier_reward/group_std_mean": 0.24241564869880677,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.2030681014060974,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.019554105401039124,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.02477550767362118,
"signal/confidence_uniqueness_reward/group_std_mean": 0.041399688273668286,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.026273638755083085,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0024775508558377624,
"signal/format_reward/centered_abs_mean": 0.01055908203125,
"signal/format_reward/group_std_mean": 0.024568172357976436,
"signal/format_reward/group_zero_std_frac": 0.878125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.05671740174293518,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.005279541015625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.3586492359638214,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.42883986234664917,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.37724814414978025,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.03586492240428925,
"step": 35
},
{
"calibration/aurc": 0.5562746804831212,
"calibration/batch_distribution_entropy": 0.9784437549906835,
"calibration/batch_entropy_100bins": 0.9690982151357795,
"calibration/batch_entropy_10bins": 0.9784437549906835,
"calibration/batch_entropy_50bins": 0.9776671487784686,
"calibration/batch_uniqueness": 0.9535524873769825,
"calibration/buffer_distribution_entropy": 0.9560775766869393,
"calibration/buffer_entropy_100bins": 0.834081194801799,
"calibration/buffer_entropy_10bins": 0.9560775766869393,
"calibration/buffer_entropy_50bins": 0.8830149707864867,
"calibration/confidence_entropy": 0.5230787621794859,
"calibration/coverage@0%": 0.000392156862745098,
"calibration/coverage@1%": 0.000392156862745098,
"calibration/coverage@10%": 0.000392156862745098,
"calibration/coverage@15%": 0.000392156862745098,
"calibration/coverage@20%": 0.000392156862745098,
"calibration/coverage@25%": 0.005480219485054295,
"calibration/coverage@30%": 0.014090786999731398,
"calibration/coverage@5%": 0.000392156862745098,
"calibration/distribution_entropy_10": 0.9784437549906835,
"calibration/distribution_entropy_100": 0.9690982151357795,
"calibration/ece": 0.21167051982873195,
"calibration/mean_confidence": 0.43923672562763344,
"calibration/unique_confidence_per_question": 0.997265625,
"calibration/unique_confidences": 510.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0029296875,
"completions/max_length": 978.6,
"completions/max_terminated_length": 978.6,
"completions/mean_length": 120.215234375,
"completions/mean_terminated_length": 120.57055206298828,
"completions/min_length": 0.0,
"completions/min_terminated_length": 53.2,
"epoch": 0.128,
"grad_norm": 0.02596159279346466,
"learning_rate": 1e-06,
"loss": -0.0137,
"num_tokens": 131942068.0,
"reward": 0.8621386885643005,
"reward_std": 0.12643432021141052,
"rewards/accgated_coverage_0": 0.011271145474165679,
"rewards/accgated_coverage_1": 0.011271145474165679,
"rewards/accgated_coverage_10": 0.011271145474165679,
"rewards/accgated_coverage_15": 0.011271145474165679,
"rewards/accgated_coverage_20": 0.011271145474165679,
"rewards/accgated_coverage_25": 0.011271145474165679,
"rewards/accgated_coverage_5": 0.011271145474165679,
"rewards/accuracy_reward": 0.4208984375,
"rewards/brier_reward": 0.7081641912460327,
"rewards/confidence_uniqueness_reward": 0.9498059868812561,
"rewards/format_reward": 0.996484375,
"rewards/frontier_entropy_batch_reward": -0.20239529609680176,
"signal/accgated_coverage_0/centered_abs_mean": 0.05066698864102363,
"signal/accgated_coverage_0/group_std_mean": 0.06311498582363129,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.05274929255247116,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.005066698882728815,
"signal/accgated_coverage_1/centered_abs_mean": 0.05066698864102363,
"signal/accgated_coverage_1/group_std_mean": 0.06311498582363129,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.05274929255247116,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.005066698882728815,
"signal/accgated_coverage_10/centered_abs_mean": 0.05066698864102363,
"signal/accgated_coverage_10/group_std_mean": 0.06311498582363129,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.05274929255247116,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.005066698882728815,
"signal/accgated_coverage_15/centered_abs_mean": 0.05066698864102363,
"signal/accgated_coverage_15/group_std_mean": 0.06311498582363129,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05274929255247116,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.005066698882728815,
"signal/accgated_coverage_20/centered_abs_mean": 0.05066698864102363,
"signal/accgated_coverage_20/group_std_mean": 0.06311498582363129,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.05274929255247116,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.005066698882728815,
"signal/accgated_coverage_25/centered_abs_mean": 0.05066698864102363,
"signal/accgated_coverage_25/group_std_mean": 0.06311498582363129,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.05274929255247116,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.005066698882728815,
"signal/accgated_coverage_5/centered_abs_mean": 0.05066698864102363,
"signal/accgated_coverage_5/group_std_mean": 0.06311498582363129,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.05274929255247116,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.005066698882728815,
"signal/accuracy_reward/centered_abs_mean": 0.1521240234375,
"signal/accuracy_reward/group_std_mean": 0.19842498898506164,
"signal/accuracy_reward/group_zero_std_frac": 0.440625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7887407898902893,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.07606201171875,
"signal/advantage_abs_mean": 0.7775111436843872,
"signal/advantage_pre_scale_abs_mean": 0.09791817367076874,
"signal/advantage_pre_scale_std": 0.13843732476234435,
"signal/advantage_std": 0.9834481239318847,
"signal/brier_reward/centered_abs_mean": 0.2133061707019806,
"signal/brier_reward/group_std_mean": 0.2618970155715942,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.22628326416015626,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.021330617368221283,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01804858408868313,
"signal/confidence_uniqueness_reward/group_std_mean": 0.033051389083266255,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.019152047485113143,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001804858399555087,
"signal/format_reward/centered_abs_mean": 0.00679931640625,
"signal/format_reward/group_std_mean": 0.019551069661974906,
"signal/format_reward/group_zero_std_frac": 0.890625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.03566441833972931,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.003399658203125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2966335892677307,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.37520670890808105,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.3161166191101074,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.029663359001278877,
"step": 40
},
{
"calibration/aurc": 0.358392498307488,
"calibration/batch_distribution_entropy": 0.9727884666420131,
"calibration/batch_entropy_100bins": 0.9658152998471852,
"calibration/batch_entropy_10bins": 0.9727884666420131,
"calibration/batch_entropy_50bins": 0.9735809230320702,
"calibration/batch_uniqueness": 0.9521793794728179,
"calibration/buffer_distribution_entropy": 0.9683937333285725,
"calibration/buffer_entropy_100bins": 0.8686393961796532,
"calibration/buffer_entropy_10bins": 0.9683937333285725,
"calibration/buffer_entropy_50bins": 0.9091305845412911,
"calibration/confidence_entropy": 0.4743727500369258,
"calibration/coverage@0%": 0.004701295828703686,
"calibration/coverage@1%": 0.004701295828703686,
"calibration/coverage@10%": 0.004701295828703686,
"calibration/coverage@15%": 0.05275738006249543,
"calibration/coverage@20%": 0.18635265892746608,
"calibration/coverage@25%": 0.206667461235914,
"calibration/coverage@30%": 0.2090173357067886,
"calibration/coverage@5%": 0.004701295828703686,
"calibration/distribution_entropy_10": 0.9727884666420131,
"calibration/distribution_entropy_100": 0.9658152998471852,
"calibration/ece": 0.288734558167821,
"calibration/mean_confidence": 0.4082907115715816,
"calibration/unique_confidence_per_question": 0.996875,
"calibration/unique_confidences": 510.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0033203125,
"completions/max_length": 1159.6,
"completions/max_terminated_length": 1159.6,
"completions/mean_length": 131.62626953125,
"completions/mean_terminated_length": 132.06433715820313,
"completions/min_length": 0.0,
"completions/min_terminated_length": 54.4,
"epoch": 0.144,
"grad_norm": 0.014015128836035728,
"learning_rate": 1e-06,
"loss": -0.0091,
"num_tokens": 148240353.0,
"reward": 0.8941228747367859,
"reward_std": 0.13708886206150056,
"rewards/accgated_coverage_0": -0.00866492409259081,
"rewards/accgated_coverage_1": -0.00866492409259081,
"rewards/accgated_coverage_10": -0.00866492409259081,
"rewards/accgated_coverage_15": -0.00866492409259081,
"rewards/accgated_coverage_20": -0.00866492409259081,
"rewards/accgated_coverage_25": -0.00866492409259081,
"rewards/accgated_coverage_5": -0.00866492409259081,
"rewards/accuracy_reward": 0.526953125,
"rewards/brier_reward": 0.6824822306632996,
"rewards/confidence_uniqueness_reward": 0.9471188902854919,
"rewards/format_reward": 0.9966796875,
"rewards/frontier_entropy_batch_reward": -0.24588188529014587,
"signal/accgated_coverage_0/centered_abs_mean": 0.09038410186767579,
"signal/accgated_coverage_0/group_std_mean": 0.10992505997419358,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08354411572217942,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.009038410056382418,
"signal/accgated_coverage_1/centered_abs_mean": 0.09038410186767579,
"signal/accgated_coverage_1/group_std_mean": 0.10992505997419358,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08354411572217942,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.009038410056382418,
"signal/accgated_coverage_10/centered_abs_mean": 0.09038410186767579,
"signal/accgated_coverage_10/group_std_mean": 0.10992505997419358,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08354411572217942,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.009038410056382418,
"signal/accgated_coverage_15/centered_abs_mean": 0.09038410186767579,
"signal/accgated_coverage_15/group_std_mean": 0.10992505997419358,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.08354411572217942,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.009038410056382418,
"signal/accgated_coverage_20/centered_abs_mean": 0.09038410186767579,
"signal/accgated_coverage_20/group_std_mean": 0.10992505997419358,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.08354411572217942,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.009038410056382418,
"signal/accgated_coverage_25/centered_abs_mean": 0.09038410186767579,
"signal/accgated_coverage_25/group_std_mean": 0.10992505997419358,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.08354411572217942,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.009038410056382418,
"signal/accgated_coverage_5/centered_abs_mean": 0.09038410186767579,
"signal/accgated_coverage_5/group_std_mean": 0.10992505997419358,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08354411572217942,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.009038410056382418,
"signal/accuracy_reward/centered_abs_mean": 0.1458251953125,
"signal/accuracy_reward/group_std_mean": 0.19198355376720427,
"signal/accuracy_reward/group_zero_std_frac": 0.453125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6883184671401977,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.07291259765625,
"signal/advantage_abs_mean": 0.787965190410614,
"signal/advantage_pre_scale_abs_mean": 0.10850597620010376,
"signal/advantage_pre_scale_std": 0.145897975564003,
"signal/advantage_std": 0.9835744142532349,
"signal/brier_reward/centered_abs_mean": 0.23881636261940004,
"signal/brier_reward/group_std_mean": 0.287902706861496,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.222304630279541,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.023881636559963226,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.021583887562155724,
"signal/confidence_uniqueness_reward/group_std_mean": 0.03647680841386318,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.020256298407912255,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.002158388658426702,
"signal/format_reward/centered_abs_mean": 0.006396484375,
"signal/format_reward/group_std_mean": 0.018050693720579148,
"signal/format_reward/group_zero_std_frac": 0.9,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.02993999347090721,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0031982421875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.34290409088134766,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.418064683675766,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.32126058340072633,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.034290410578250885,
"step": 45
},
{
"calibration/aurc": 0.4303138823686969,
"calibration/batch_distribution_entropy": 0.9872966941622152,
"calibration/batch_entropy_100bins": 0.9719368590983816,
"calibration/batch_entropy_10bins": 0.9872966941622152,
"calibration/batch_entropy_50bins": 0.9820715952358517,
"calibration/batch_uniqueness": 0.954329758674562,
"calibration/buffer_distribution_entropy": 0.9750856011339903,
"calibration/buffer_entropy_100bins": 0.8930578685467975,
"calibration/buffer_entropy_10bins": 0.9750856011339903,
"calibration/buffer_entropy_50bins": 0.9272801203148754,
"calibration/confidence_entropy": 0.4805351308262821,
"calibration/coverage@0%": 0.006290792099582226,
"calibration/coverage@1%": 0.006290792099582226,
"calibration/coverage@10%": 0.007865595249188525,
"calibration/coverage@15%": 0.01180493271366227,
"calibration/coverage@20%": 0.025935819130211625,
"calibration/coverage@25%": 0.04049894153571988,
"calibration/coverage@30%": 0.07473628950438871,
"calibration/coverage@5%": 0.006290792099582226,
"calibration/distribution_entropy_10": 0.9872966941622152,
"calibration/distribution_entropy_100": 0.9719368590983816,
"calibration/ece": 0.17536238067846233,
"calibration/mean_confidence": 0.5142108289417581,
"calibration/unique_confidence_per_question": 0.9921875,
"calibration/unique_confidences": 508.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00361328125,
"completions/max_length": 1150.6,
"completions/max_terminated_length": 1150.6,
"completions/mean_length": 134.39892578125,
"completions/mean_terminated_length": 134.89354248046874,
"completions/min_length": 0.0,
"completions/min_terminated_length": 57.6,
"epoch": 0.16,
"grad_norm": 0.012396840378642082,
"learning_rate": 1e-06,
"loss": -0.0187,
"num_tokens": 164637526.0,
"reward": 0.8867763638496399,
"reward_std": 0.13014983087778093,
"rewards/accgated_coverage_0": 0.015360437519848346,
"rewards/accgated_coverage_1": 0.015360437519848346,
"rewards/accgated_coverage_10": 0.015360437519848346,
"rewards/accgated_coverage_15": 0.015360437519848346,
"rewards/accgated_coverage_20": 0.015360437519848346,
"rewards/accgated_coverage_25": 0.015360437519848346,
"rewards/accgated_coverage_5": 0.015360437519848346,
"rewards/accuracy_reward": 0.4568359375,
"rewards/brier_reward": 0.7095378875732422,
"rewards/confidence_uniqueness_reward": 0.9516965389251709,
"rewards/format_reward": 0.99609375,
"rewards/frontier_entropy_batch_reward": -0.16564251780509948,
"signal/accgated_coverage_0/centered_abs_mean": 0.05440087839961052,
"signal/accgated_coverage_0/group_std_mean": 0.06851492002606392,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.05290875434875488,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.005440088175237179,
"signal/accgated_coverage_1/centered_abs_mean": 0.05440087839961052,
"signal/accgated_coverage_1/group_std_mean": 0.06851492002606392,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.05290875434875488,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.005440088175237179,
"signal/accgated_coverage_10/centered_abs_mean": 0.05440087839961052,
"signal/accgated_coverage_10/group_std_mean": 0.06851492002606392,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.05290875434875488,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.005440088175237179,
"signal/accgated_coverage_15/centered_abs_mean": 0.05440087839961052,
"signal/accgated_coverage_15/group_std_mean": 0.06851492002606392,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05290875434875488,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.005440088175237179,
"signal/accgated_coverage_20/centered_abs_mean": 0.05440087839961052,
"signal/accgated_coverage_20/group_std_mean": 0.06851492002606392,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.05290875434875488,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.005440088175237179,
"signal/accgated_coverage_25/centered_abs_mean": 0.05440087839961052,
"signal/accgated_coverage_25/group_std_mean": 0.06851492002606392,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.05290875434875488,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.005440088175237179,
"signal/accgated_coverage_5/centered_abs_mean": 0.05440087839961052,
"signal/accgated_coverage_5/group_std_mean": 0.06851492002606392,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.05290875434875488,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.005440088175237179,
"signal/accuracy_reward/centered_abs_mean": 0.13243408203125,
"signal/accuracy_reward/group_std_mean": 0.17721919417381288,
"signal/accuracy_reward/group_zero_std_frac": 0.48125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6435363292694092,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.066217041015625,
"signal/advantage_abs_mean": 0.7713564515113831,
"signal/advantage_pre_scale_abs_mean": 0.1001865029335022,
"signal/advantage_pre_scale_std": 0.14206840097904205,
"signal/advantage_std": 0.9835274338722229,
"signal/brier_reward/centered_abs_mean": 0.22864238023757935,
"signal/brier_reward/group_std_mean": 0.2773905754089355,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.22308139503002167,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.022864237800240515,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.016879253648221494,
"signal/confidence_uniqueness_reward/group_std_mean": 0.031111791357398035,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.016394594684243202,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0016879254020750523,
"signal/format_reward/centered_abs_mean": 0.00748291015625,
"signal/format_reward/group_std_mean": 0.019742927886545657,
"signal/format_reward/group_zero_std_frac": 0.896875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.03598161712288857,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.003741455078125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2529976963996887,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.33269529342651366,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.24671269357204437,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025299769639968873,
"step": 50
},
{
"epoch": 0.16,
"eval_completions/clipped_ratio": 0.001953125,
"eval_completions/max_length": 830.25,
"eval_completions/max_terminated_length": 830.25,
"eval_completions/mean_length": 137.92631912231445,
"eval_completions/mean_terminated_length": 138.21495056152344,
"eval_completions/min_length": 49.25,
"eval_completions/min_terminated_length": 67.0,
"eval_loss": 0.0,
"eval_num_tokens": 164637526.0,
"eval_reward": 0.768775075674057,
"eval_reward_std": 0.22393060475587845,
"eval_rewards/accgated_coverage_0": 0.017385049490258098,
"eval_rewards/accgated_coverage_1": 0.017385049490258098,
"eval_rewards/accgated_coverage_10": 0.017385049490258098,
"eval_rewards/accgated_coverage_15": 0.017385049490258098,
"eval_rewards/accgated_coverage_20": 0.017385049490258098,
"eval_rewards/accgated_coverage_25": 0.017385049490258098,
"eval_rewards/accgated_coverage_5": 0.017385049490258098,
"eval_rewards/accuracy_reward": 0.390625,
"eval_rewards/brier_reward": 0.7255421131849289,
"eval_rewards/confidence_uniqueness_reward": 0.8952006697654724,
"eval_rewards/format_reward": 0.998046875,
"eval_rewards/frontier_entropy_batch_reward": -0.998046875,
"eval_runtime": 36.4841,
"eval_samples_per_second": 13.705,
"eval_signal/accgated_coverage_0/centered_abs_mean": 0.05017893575131893,
"eval_signal/accgated_coverage_0/group_std_mean": 0.06709024310112,
"eval_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.022471013944596052,
"eval_signal/accgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.005017893738113344,
"eval_signal/accgated_coverage_1/centered_abs_mean": 0.05017893575131893,
"eval_signal/accgated_coverage_1/group_std_mean": 0.06709024310112,
"eval_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.022471013944596052,
"eval_signal/accgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.005017893738113344,
"eval_signal/accgated_coverage_10/centered_abs_mean": 0.05017893575131893,
"eval_signal/accgated_coverage_10/group_std_mean": 0.06709024310112,
"eval_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.022471013944596052,
"eval_signal/accgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.005017893738113344,
"eval_signal/accgated_coverage_15/centered_abs_mean": 0.05017893575131893,
"eval_signal/accgated_coverage_15/group_std_mean": 0.06709024310112,
"eval_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.022471013944596052,
"eval_signal/accgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.005017893738113344,
"eval_signal/accgated_coverage_20/centered_abs_mean": 0.05017893575131893,
"eval_signal/accgated_coverage_20/group_std_mean": 0.06709024310112,
"eval_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.022471013944596052,
"eval_signal/accgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.005017893738113344,
"eval_signal/accgated_coverage_25/centered_abs_mean": 0.05017893575131893,
"eval_signal/accgated_coverage_25/group_std_mean": 0.06709024310112,
"eval_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.022471013944596052,
"eval_signal/accgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.005017893738113344,
"eval_signal/accgated_coverage_5/centered_abs_mean": 0.05017893575131893,
"eval_signal/accgated_coverage_5/group_std_mean": 0.06709024310112,
"eval_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.022471013944596052,
"eval_signal/accgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.005017893738113344,
"eval_signal/accuracy_reward/centered_abs_mean": 0.456787109375,
"eval_signal/accuracy_reward/group_std_mean": 0.4841504618525505,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.0233747214078903,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.2283935546875,
"eval_signal/advantage_abs_mean": 0.8952026814222336,
"eval_signal/advantage_pre_scale_abs_mean": 0.20100067555904388,
"eval_signal/advantage_pre_scale_std": 0.22167641296982765,
"eval_signal/advantage_std": 0.9876750856637955,
"eval_signal/brier_reward/centered_abs_mean": 0.2335420735180378,
"eval_signal/brier_reward/group_std_mean": 0.28004275262355804,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.10474211536347866,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.023354208562523127,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.04180317558348179,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.0545404851436615,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.01872535515576601,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.00418031751178205,
"eval_signal/format_reward/centered_abs_mean": 0.0037841796875,
"eval_signal/format_reward/group_std_mean": 0.011048543266952038,
"eval_signal/format_reward/group_zero_std_frac": 0.9375,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.008115066215395927,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.00189208984375,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0037841796875,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.011048543266952038,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.9375,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.001623013406060636,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0003784179862122983,
"eval_steps_per_second": 0.11,
"step": 50
},
{
"epoch": 0.16,
"step": 50,
"train_probe_completions/clipped_ratio": 0.00390625,
"train_probe_completions/max_length": 846.75,
"train_probe_completions/max_terminated_length": 846.75,
"train_probe_completions/mean_length": 135.55926895141602,
"train_probe_completions/mean_terminated_length": 136.09591674804688,
"train_probe_completions/min_length": 34.5,
"train_probe_completions/min_terminated_length": 69.5,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 164637526.0,
"train_probe_reward": 0.8164596557617188,
"train_probe_reward_std": 0.22054458037018776,
"train_probe_rewards/accgated_coverage_0": 0.018788411282002926,
"train_probe_rewards/accgated_coverage_1": 0.018788411282002926,
"train_probe_rewards/accgated_coverage_10": 0.018788411282002926,
"train_probe_rewards/accgated_coverage_15": 0.018788411282002926,
"train_probe_rewards/accgated_coverage_20": 0.018788411282002926,
"train_probe_rewards/accgated_coverage_25": 0.018788411282002926,
"train_probe_rewards/accgated_coverage_5": 0.018788411282002926,
"train_probe_rewards/accuracy_reward": 0.48046875,
"train_probe_rewards/brier_reward": 0.7549779415130615,
"train_probe_rewards/confidence_uniqueness_reward": 0.8913810551166534,
"train_probe_rewards/format_reward": 0.99609375,
"train_probe_rewards/frontier_entropy_batch_reward": -0.99609375,
"train_probe_runtime": 43.9926,
"train_probe_samples_per_second": 11.366,
"train_probe_signal/accgated_coverage_0/centered_abs_mean": 0.0745154544711113,
"train_probe_signal/accgated_coverage_0/group_std_mean": 0.09976488724350929,
"train_probe_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.034343925304710865,
"train_probe_signal/accgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.007451545679941773,
"train_probe_signal/accgated_coverage_1/centered_abs_mean": 0.0745154544711113,
"train_probe_signal/accgated_coverage_1/group_std_mean": 0.09976488724350929,
"train_probe_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.034343925304710865,
"train_probe_signal/accgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.007451545679941773,
"train_probe_signal/accgated_coverage_10/centered_abs_mean": 0.0745154544711113,
"train_probe_signal/accgated_coverage_10/group_std_mean": 0.09976488724350929,
"train_probe_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.034343925304710865,
"train_probe_signal/accgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.007451545679941773,
"train_probe_signal/accgated_coverage_15/centered_abs_mean": 0.0745154544711113,
"train_probe_signal/accgated_coverage_15/group_std_mean": 0.09976488724350929,
"train_probe_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.034343925304710865,
"train_probe_signal/accgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.007451545679941773,
"train_probe_signal/accgated_coverage_20/centered_abs_mean": 0.0745154544711113,
"train_probe_signal/accgated_coverage_20/group_std_mean": 0.09976488724350929,
"train_probe_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.034343925304710865,
"train_probe_signal/accgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.007451545679941773,
"train_probe_signal/accgated_coverage_25/centered_abs_mean": 0.0745154544711113,
"train_probe_signal/accgated_coverage_25/group_std_mean": 0.09976488724350929,
"train_probe_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.034343925304710865,
"train_probe_signal/accgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.007451545679941773,
"train_probe_signal/accgated_coverage_5/centered_abs_mean": 0.0745154544711113,
"train_probe_signal/accgated_coverage_5/group_std_mean": 0.09976488724350929,
"train_probe_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.034343925304710865,
"train_probe_signal/accgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.007451545679941773,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.48583984375,
"train_probe_signal/accuracy_reward/group_std_mean": 0.5006761774420738,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.1079604029655457,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.242919921875,
"train_probe_signal/advantage_abs_mean": 0.904426246881485,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.19941693544387817,
"train_probe_signal/advantage_pre_scale_std": 0.21841463819146156,
"train_probe_signal/advantage_std": 0.9876678287982941,
"train_probe_signal/brier_reward/centered_abs_mean": 0.2183300368487835,
"train_probe_signal/brier_reward/group_std_mean": 0.2686193361878395,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.09959661029279232,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.02183300396427512,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.04378090798854828,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.060883630998432636,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.01995938364416361,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004378090729005635,
"train_probe_signal/format_reward/centered_abs_mean": 0.007568359375,
"train_probe_signal/format_reward/group_std_mean": 0.022097086533904076,
"train_probe_signal/format_reward/group_zero_std_frac": 0.875,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.017395395785570145,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.0037841796875,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.007568359375,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.022097086533904076,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.875,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0034790795762091875,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0007568359724245965,
"train_probe_steps_per_second": 0.091
},
{
"calibration/aurc": 0.4106372154416068,
"calibration/batch_distribution_entropy": 0.990430322832305,
"calibration/batch_entropy_100bins": 0.9750504968819012,
"calibration/batch_entropy_10bins": 0.990430322832305,
"calibration/batch_entropy_50bins": 0.984369266304293,
"calibration/batch_uniqueness": 0.9555020219154182,
"calibration/buffer_distribution_entropy": 0.9796954716481563,
"calibration/buffer_entropy_100bins": 0.9133396770462998,
"calibration/buffer_entropy_10bins": 0.9796954716481563,
"calibration/buffer_entropy_50bins": 0.9418611674698566,
"calibration/confidence_entropy": 0.49630968504306505,
"calibration/coverage@0%": 0.0023475721624266143,
"calibration/coverage@1%": 0.0023475721624266143,
"calibration/coverage@10%": 0.0023475721624266143,
"calibration/coverage@15%": 0.003521740459882583,
"calibration/coverage@20%": 0.009010401677794405,
"calibration/coverage@25%": 0.05394145286539274,
"calibration/coverage@30%": 0.19196675887820883,
"calibration/coverage@5%": 0.0023475721624266143,
"calibration/distribution_entropy_10": 0.990430322832305,
"calibration/distribution_entropy_100": 0.9750504968819012,
"calibration/ece": 0.17155433868669684,
"calibration/mean_confidence": 0.4980011347780395,
"calibration/unique_confidence_per_question": 0.997265625,
"calibration/unique_confidences": 510.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.001171875,
"completions/max_length": 1217.2,
"completions/max_terminated_length": 1217.2,
"completions/mean_length": 134.47626953125,
"completions/mean_terminated_length": 134.6348846435547,
"completions/min_length": 0.0,
"completions/min_terminated_length": 58.6,
"epoch": 0.176,
"grad_norm": 0.011045852676033974,
"learning_rate": 1e-06,
"loss": -0.0065,
"num_tokens": 181251683.0,
"reward": 0.8919636964797973,
"reward_std": 0.10855261534452439,
"rewards/accgated_coverage_0": 0.015519830770790577,
"rewards/accgated_coverage_1": 0.015519830770790577,
"rewards/accgated_coverage_10": 0.015519830770790577,
"rewards/accgated_coverage_15": 0.015519830770790577,
"rewards/accgated_coverage_20": 0.015519830770790577,
"rewards/accgated_coverage_25": 0.015519830770790577,
"rewards/accgated_coverage_5": 0.015519830770790577,
"rewards/accuracy_reward": 0.45791015625,
"rewards/brier_reward": 0.7383434176445007,
"rewards/confidence_uniqueness_reward": 0.9536823749542236,
"rewards/format_reward": 0.9986328125,
"rewards/frontier_entropy_batch_reward": -0.16374244391918183,
"signal/accgated_coverage_0/centered_abs_mean": 0.05432665124535561,
"signal/accgated_coverage_0/group_std_mean": 0.06869832873344421,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0641322560608387,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.005432665068656206,
"signal/accgated_coverage_1/centered_abs_mean": 0.05432665124535561,
"signal/accgated_coverage_1/group_std_mean": 0.06869832873344421,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0641322560608387,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.005432665068656206,
"signal/accgated_coverage_10/centered_abs_mean": 0.05432665124535561,
"signal/accgated_coverage_10/group_std_mean": 0.06869832873344421,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.0641322560608387,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.005432665068656206,
"signal/accgated_coverage_15/centered_abs_mean": 0.05432665124535561,
"signal/accgated_coverage_15/group_std_mean": 0.06869832873344421,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.0641322560608387,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.005432665068656206,
"signal/accgated_coverage_20/centered_abs_mean": 0.05432665124535561,
"signal/accgated_coverage_20/group_std_mean": 0.06869832873344421,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.0641322560608387,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.005432665068656206,
"signal/accgated_coverage_25/centered_abs_mean": 0.05432665124535561,
"signal/accgated_coverage_25/group_std_mean": 0.06869832873344421,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.0641322560608387,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.005432665068656206,
"signal/accgated_coverage_5/centered_abs_mean": 0.05432665124535561,
"signal/accgated_coverage_5/group_std_mean": 0.06869832873344421,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.0641322560608387,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.005432665068656206,
"signal/accuracy_reward/centered_abs_mean": 0.103558349609375,
"signal/accuracy_reward/group_std_mean": 0.14366039037704467,
"signal/accuracy_reward/group_zero_std_frac": 0.56875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6132125377655029,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0517791748046875,
"signal/advantage_abs_mean": 0.7724816560745239,
"signal/advantage_pre_scale_abs_mean": 0.08388731181621552,
"signal/advantage_pre_scale_std": 0.12010505795478821,
"signal/advantage_std": 0.9832999467849731,
"signal/brier_reward/centered_abs_mean": 0.19642902314662933,
"signal/brier_reward/group_std_mean": 0.2467724412679672,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.23659427165985109,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.01964290253818035,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.013406224921345711,
"signal/confidence_uniqueness_reward/group_std_mean": 0.0211888175457716,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.016216434165835382,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013406225014477968,
"signal/format_reward/centered_abs_mean": 0.00264892578125,
"signal/format_reward/group_std_mean": 0.0077339803334325555,
"signal/format_reward/group_zero_std_frac": 0.95625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.01582014877349138,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.001324462890625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.24987545609474182,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.32901384234428405,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.3033535420894623,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.024987546354532243,
"step": 55
},
{
"calibration/aurc": 0.3206109694951687,
"calibration/batch_distribution_entropy": 0.9755862880539994,
"calibration/batch_entropy_100bins": 0.9650828536987974,
"calibration/batch_entropy_10bins": 0.9755862880539994,
"calibration/batch_entropy_50bins": 0.9726825049858414,
"calibration/batch_uniqueness": 0.9516161754625777,
"calibration/buffer_distribution_entropy": 0.9832107728820045,
"calibration/buffer_entropy_100bins": 0.9268917363616765,
"calibration/buffer_entropy_10bins": 0.9832107728820045,
"calibration/buffer_entropy_50bins": 0.9515345823767067,
"calibration/confidence_entropy": 0.4689407321390662,
"calibration/coverage@0%": 0.0031341911764705883,
"calibration/coverage@1%": 0.0031341911764705883,
"calibration/coverage@10%": 0.06849724264705884,
"calibration/coverage@15%": 0.0954779411764706,
"calibration/coverage@20%": 0.18240501551648058,
"calibration/coverage@25%": 0.28765231691991866,
"calibration/coverage@30%": 0.4113470469523426,
"calibration/coverage@5%": 0.0031341911764705883,
"calibration/distribution_entropy_10": 0.9755862880539994,
"calibration/distribution_entropy_100": 0.9650828536987974,
"calibration/ece": 0.12412214373509349,
"calibration/mean_confidence": 0.45758849056826667,
"calibration/unique_confidence_per_question": 0.9921875,
"calibration/unique_confidences": 508.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.001953125,
"completions/max_length": 1098.0,
"completions/max_terminated_length": 1098.0,
"completions/mean_length": 142.01201171875,
"completions/mean_terminated_length": 142.28759765625,
"completions/min_length": 0.0,
"completions/min_terminated_length": 61.4,
"epoch": 0.192,
"grad_norm": 0.012371505610644817,
"learning_rate": 1e-06,
"loss": -0.0151,
"num_tokens": 197520702.0,
"reward": 0.9012426376342774,
"reward_std": 0.10214150249958039,
"rewards/accgated_coverage_0": 0.023549126461148262,
"rewards/accgated_coverage_1": 0.023549126461148262,
"rewards/accgated_coverage_10": 0.023549126461148262,
"rewards/accgated_coverage_15": 0.023549126461148262,
"rewards/accgated_coverage_20": 0.023549126461148262,
"rewards/accgated_coverage_25": 0.023549126461148262,
"rewards/accgated_coverage_5": 0.023549126461148262,
"rewards/accuracy_reward": 0.47041015625,
"rewards/brier_reward": 0.7599319815635681,
"rewards/confidence_uniqueness_reward": 0.9506913304328919,
"rewards/format_reward": 0.99794921875,
"rewards/frontier_entropy_batch_reward": -0.20483775436878204,
"signal/accgated_coverage_0/centered_abs_mean": 0.053258597105741504,
"signal/accgated_coverage_0/group_std_mean": 0.06790817081928253,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07255900055170059,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.005325859598815441,
"signal/accgated_coverage_1/centered_abs_mean": 0.053258597105741504,
"signal/accgated_coverage_1/group_std_mean": 0.06790817081928253,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07255900055170059,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.005325859598815441,
"signal/accgated_coverage_10/centered_abs_mean": 0.053258597105741504,
"signal/accgated_coverage_10/group_std_mean": 0.06790817081928253,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07255900055170059,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.005325859598815441,
"signal/accgated_coverage_15/centered_abs_mean": 0.053258597105741504,
"signal/accgated_coverage_15/group_std_mean": 0.06790817081928253,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.07255900055170059,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.005325859598815441,
"signal/accgated_coverage_20/centered_abs_mean": 0.053258597105741504,
"signal/accgated_coverage_20/group_std_mean": 0.06790817081928253,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.07255900055170059,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.005325859598815441,
"signal/accgated_coverage_25/centered_abs_mean": 0.053258597105741504,
"signal/accgated_coverage_25/group_std_mean": 0.06790817081928253,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.07255900055170059,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.005325859598815441,
"signal/accgated_coverage_5/centered_abs_mean": 0.053258597105741504,
"signal/accgated_coverage_5/group_std_mean": 0.06790817081928253,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07255900055170059,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.005325859598815441,
"signal/accuracy_reward/centered_abs_mean": 0.109832763671875,
"signal/accuracy_reward/group_std_mean": 0.1481279566884041,
"signal/accuracy_reward/group_zero_std_frac": 0.56875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7526550054550171,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0549163818359375,
"signal/advantage_abs_mean": 0.7651453971862793,
"signal/advantage_pre_scale_abs_mean": 0.07803591638803482,
"signal/advantage_pre_scale_std": 0.11695106476545333,
"signal/advantage_std": 0.983116090297699,
"signal/brier_reward/centered_abs_mean": 0.17573796808719636,
"signal/brier_reward/group_std_mean": 0.2217639535665512,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.2452210396528244,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.017573796957731248,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01655643284320831,
"signal/confidence_uniqueness_reward/group_std_mean": 0.026009998098015786,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02359951063990593,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001655643223784864,
"signal/format_reward/centered_abs_mean": 0.003924560546875,
"signal/format_reward/group_std_mean": 0.010532832238823175,
"signal/format_reward/group_zero_std_frac": 0.94375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.029064977914094924,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0019622802734375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2817984759807587,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3581483006477356,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.3971118688583374,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028179847821593284,
"step": 60
},
{
"calibration/aurc": 0.30736133671836796,
"calibration/batch_distribution_entropy": 0.9788404724233372,
"calibration/batch_entropy_100bins": 0.9702944467634456,
"calibration/batch_entropy_10bins": 0.9788404724233372,
"calibration/batch_entropy_50bins": 0.9768523437767802,
"calibration/batch_uniqueness": 0.9528387196153367,
"calibration/buffer_distribution_entropy": 0.9849736454723137,
"calibration/buffer_entropy_100bins": 0.937161370661754,
"calibration/buffer_entropy_10bins": 0.9849736454723137,
"calibration/buffer_entropy_50bins": 0.9585719203167763,
"calibration/confidence_entropy": 0.4869858085322839,
"calibration/coverage@0%": 0.014070923746210811,
"calibration/coverage@1%": 0.014070923746210811,
"calibration/coverage@10%": 0.01914904874621081,
"calibration/coverage@15%": 0.0738518673736618,
"calibration/coverage@20%": 0.20920164469897545,
"calibration/coverage@25%": 0.4817726499846514,
"calibration/coverage@30%": 0.5740948100130463,
"calibration/coverage@5%": 0.015242798746210812,
"calibration/distribution_entropy_10": 0.9788404724233372,
"calibration/distribution_entropy_100": 0.9702944467634456,
"calibration/ece": 0.1384404856132969,
"calibration/mean_confidence": 0.5396573794647559,
"calibration/unique_confidence_per_question": 0.98984375,
"calibration/unique_confidences": 506.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00224609375,
"completions/max_length": 1364.8,
"completions/max_terminated_length": 1364.8,
"completions/mean_length": 166.62919921875,
"completions/mean_terminated_length": 167.01428833007813,
"completions/min_length": 0.0,
"completions/min_terminated_length": 61.2,
"epoch": 0.208,
"grad_norm": 0.0094198789447546,
"learning_rate": 1e-06,
"loss": -0.0126,
"num_tokens": 214259209.0,
"reward": 0.9341094017028808,
"reward_std": 0.10703354924917222,
"rewards/accgated_coverage_0": 0.019850392639636994,
"rewards/accgated_coverage_1": 0.019850392639636994,
"rewards/accgated_coverage_10": 0.019850392639636994,
"rewards/accgated_coverage_15": 0.019850392639636994,
"rewards/accgated_coverage_20": 0.019850392639636994,
"rewards/accgated_coverage_25": 0.019850392639636994,
"rewards/accgated_coverage_5": 0.019850392639636994,
"rewards/accuracy_reward": 0.53671875,
"rewards/brier_reward": 0.7639217257499695,
"rewards/confidence_uniqueness_reward": 0.9516273498535156,
"rewards/format_reward": 0.99775390625,
"rewards/frontier_entropy_batch_reward": -0.18577099740505218,
"signal/accgated_coverage_0/centered_abs_mean": 0.05513238981366157,
"signal/accgated_coverage_0/group_std_mean": 0.07063145563006401,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07241926714777946,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.005513238999992609,
"signal/accgated_coverage_1/centered_abs_mean": 0.05513238981366157,
"signal/accgated_coverage_1/group_std_mean": 0.07063145563006401,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07241926714777946,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.005513238999992609,
"signal/accgated_coverage_10/centered_abs_mean": 0.05513238981366157,
"signal/accgated_coverage_10/group_std_mean": 0.07063145563006401,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07241926714777946,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.005513238999992609,
"signal/accgated_coverage_15/centered_abs_mean": 0.05513238981366157,
"signal/accgated_coverage_15/group_std_mean": 0.07063145563006401,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.07241926714777946,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.005513238999992609,
"signal/accgated_coverage_20/centered_abs_mean": 0.05513238981366157,
"signal/accgated_coverage_20/group_std_mean": 0.07063145563006401,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.07241926714777946,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.005513238999992609,
"signal/accgated_coverage_25/centered_abs_mean": 0.05513238981366157,
"signal/accgated_coverage_25/group_std_mean": 0.07063145563006401,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.07241926714777946,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.005513238999992609,
"signal/accgated_coverage_5/centered_abs_mean": 0.05513238981366157,
"signal/accgated_coverage_5/group_std_mean": 0.07063145563006401,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07241926714777946,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.005513238999992609,
"signal/accuracy_reward/centered_abs_mean": 0.1126708984375,
"signal/accuracy_reward/group_std_mean": 0.15192099213600158,
"signal/accuracy_reward/group_zero_std_frac": 0.553125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7383570551872254,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.05633544921875,
"signal/advantage_abs_mean": 0.765564787387848,
"signal/advantage_pre_scale_abs_mean": 0.0810088112950325,
"signal/advantage_pre_scale_std": 0.12141841053962707,
"signal/advantage_std": 0.9831967949867249,
"signal/brier_reward/centered_abs_mean": 0.15333881378173828,
"signal/brier_reward/group_std_mean": 0.1960354208946228,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.20132073760032654,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.015333881601691246,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.015461865067481994,
"signal/confidence_uniqueness_reward/group_std_mean": 0.026078763976693155,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.020331867039203644,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001546186488121748,
"signal/format_reward/centered_abs_mean": 0.004339599609375,
"signal/format_reward/group_std_mean": 0.012369516305625439,
"signal/format_reward/group_zero_std_frac": 0.93125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.028545215353369714,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0021697998046875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26688307523727417,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.34284729361534116,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.3510095775127411,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026688307896256445,
"step": 65
},
{
"calibration/aurc": 0.24897785878025808,
"calibration/batch_distribution_entropy": 0.9786310217822584,
"calibration/batch_entropy_100bins": 0.9679615594022571,
"calibration/batch_entropy_10bins": 0.9786310217822584,
"calibration/batch_entropy_50bins": 0.9753836785259289,
"calibration/batch_uniqueness": 0.9529156935085942,
"calibration/buffer_distribution_entropy": 0.9866267893592152,
"calibration/buffer_entropy_100bins": 0.9453132123652406,
"calibration/buffer_entropy_10bins": 0.9866267893592152,
"calibration/buffer_entropy_50bins": 0.9642049302026777,
"calibration/confidence_entropy": 0.45265012705154717,
"calibration/coverage@0%": 0.05099147705341979,
"calibration/coverage@1%": 0.08721194949436467,
"calibration/coverage@10%": 0.29273316880886213,
"calibration/coverage@15%": 0.3781763596186506,
"calibration/coverage@20%": 0.4715224182684885,
"calibration/coverage@25%": 0.553866566311564,
"calibration/coverage@30%": 0.6342438508375791,
"calibration/coverage@5%": 0.178317652848541,
"calibration/distribution_entropy_10": 0.9786310217822584,
"calibration/distribution_entropy_100": 0.9679615594022571,
"calibration/ece": 0.1434194087647127,
"calibration/mean_confidence": 0.487639239213222,
"calibration/unique_confidence_per_question": 0.99453125,
"calibration/unique_confidences": 509.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00341796875,
"completions/max_length": 1403.6,
"completions/max_terminated_length": 1403.6,
"completions/mean_length": 185.75966796875,
"completions/mean_terminated_length": 186.40545959472655,
"completions/min_length": 0.0,
"completions/min_terminated_length": 68.2,
"epoch": 0.224,
"grad_norm": 0.010321831330657005,
"learning_rate": 1e-06,
"loss": -0.0126,
"num_tokens": 231314572.0,
"reward": 0.9223137736320496,
"reward_std": 0.10135262459516525,
"rewards/accgated_coverage_0": 0.029962604865431786,
"rewards/accgated_coverage_1": 0.029962604865431786,
"rewards/accgated_coverage_10": 0.029962604865431786,
"rewards/accgated_coverage_15": 0.029962604865431786,
"rewards/accgated_coverage_20": 0.029962604865431786,
"rewards/accgated_coverage_25": 0.029962604865431786,
"rewards/accgated_coverage_5": 0.029962604865431786,
"rewards/accuracy_reward": 0.5005859375,
"rewards/brier_reward": 0.7950564026832581,
"rewards/confidence_uniqueness_reward": 0.9484570026397705,
"rewards/format_reward": 0.99658203125,
"rewards/frontier_entropy_batch_reward": -0.21595357358455658,
"signal/accgated_coverage_0/centered_abs_mean": 0.04504996985197067,
"signal/accgated_coverage_0/group_std_mean": 0.05802757143974304,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06829853430390358,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004504997003823519,
"signal/accgated_coverage_1/centered_abs_mean": 0.04504996985197067,
"signal/accgated_coverage_1/group_std_mean": 0.05802757143974304,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06829853430390358,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004504997003823519,
"signal/accgated_coverage_10/centered_abs_mean": 0.04504996985197067,
"signal/accgated_coverage_10/group_std_mean": 0.05802757143974304,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06829853430390358,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004504997003823519,
"signal/accgated_coverage_15/centered_abs_mean": 0.04504996985197067,
"signal/accgated_coverage_15/group_std_mean": 0.05802757143974304,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06829853430390358,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004504997003823519,
"signal/accgated_coverage_20/centered_abs_mean": 0.04504996985197067,
"signal/accgated_coverage_20/group_std_mean": 0.05802757143974304,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.06829853430390358,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.004504997003823519,
"signal/accgated_coverage_25/centered_abs_mean": 0.04504996985197067,
"signal/accgated_coverage_25/group_std_mean": 0.05802757143974304,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.06829853430390358,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.004504997003823519,
"signal/accgated_coverage_5/centered_abs_mean": 0.04504996985197067,
"signal/accgated_coverage_5/group_std_mean": 0.05802757143974304,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06829853430390358,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004504997003823519,
"signal/accuracy_reward/centered_abs_mean": 0.10576171875,
"signal/accuracy_reward/group_std_mean": 0.14589932560920715,
"signal/accuracy_reward/group_zero_std_frac": 0.5625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.782033097743988,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.052880859375,
"signal/advantage_abs_mean": 0.748574697971344,
"signal/advantage_pre_scale_abs_mean": 0.07407019734382629,
"signal/advantage_pre_scale_std": 0.11904896944761276,
"signal/advantage_std": 0.9830135345458985,
"signal/brier_reward/centered_abs_mean": 0.13088436424732208,
"signal/brier_reward/group_std_mean": 0.17098954319953918,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19508094787597657,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.013088436797261238,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01858687624335289,
"signal/confidence_uniqueness_reward/group_std_mean": 0.03158754706382751,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.028095318004488945,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0018586876103654504,
"signal/format_reward/centered_abs_mean": 0.006524658203125,
"signal/format_reward/group_std_mean": 0.01692157853394747,
"signal/format_reward/group_zero_std_frac": 0.9125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.04965321645140648,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0032623291015625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.277821546792984,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3522256791591644,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.41727418303489683,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027782154828310014,
"step": 70
},
{
"calibration/aurc": 0.31751850465221687,
"calibration/batch_distribution_entropy": 0.956348942137771,
"calibration/batch_entropy_100bins": 0.9568860779531516,
"calibration/batch_entropy_10bins": 0.956348942137771,
"calibration/batch_entropy_50bins": 0.9621369016843297,
"calibration/batch_uniqueness": 0.947364440651922,
"calibration/buffer_distribution_entropy": 0.9880632817517311,
"calibration/buffer_entropy_100bins": 0.9514874565015052,
"calibration/buffer_entropy_10bins": 0.9880632817517311,
"calibration/buffer_entropy_50bins": 0.968264382427623,
"calibration/confidence_entropy": 0.48313191127807376,
"calibration/coverage@0%": 0.020716920899656377,
"calibration/coverage@1%": 0.020716920899656377,
"calibration/coverage@10%": 0.11332349241875588,
"calibration/coverage@15%": 0.20407270968804414,
"calibration/coverage@20%": 0.27950927286700467,
"calibration/coverage@25%": 0.3773100367216512,
"calibration/coverage@30%": 0.5007840489063439,
"calibration/coverage@5%": 0.07696692089965637,
"calibration/distribution_entropy_10": 0.956348942137771,
"calibration/distribution_entropy_100": 0.9568860779531516,
"calibration/ece": 0.13730135940697005,
"calibration/mean_confidence": 0.47356822433853196,
"calibration/unique_confidence_per_question": 0.99765625,
"calibration/unique_confidences": 510.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.002734375,
"completions/max_length": 1368.0,
"completions/max_terminated_length": 1368.0,
"completions/mean_length": 185.9521484375,
"completions/mean_terminated_length": 186.46139526367188,
"completions/min_length": 0.0,
"completions/min_terminated_length": 69.6,
"epoch": 0.24,
"grad_norm": 0.01096147671341896,
"learning_rate": 1e-06,
"loss": -0.0055,
"num_tokens": 248470402.0,
"reward": 0.9386130571365356,
"reward_std": 0.09798669219017028,
"rewards/accgated_coverage_0": 0.0202577177900821,
"rewards/accgated_coverage_1": 0.0202577177900821,
"rewards/accgated_coverage_10": 0.0202577177900821,
"rewards/accgated_coverage_15": 0.0202577177900821,
"rewards/accgated_coverage_20": 0.0202577177900821,
"rewards/accgated_coverage_25": 0.0202577177900821,
"rewards/accgated_coverage_5": 0.0202577177900821,
"rewards/accuracy_reward": 0.54970703125,
"rewards/brier_reward": 0.7799974203109741,
"rewards/confidence_uniqueness_reward": 0.9481614828109741,
"rewards/format_reward": 0.99697265625,
"rewards/frontier_entropy_batch_reward": -0.21723068058490752,
"signal/accgated_coverage_0/centered_abs_mean": 0.04972954392433167,
"signal/accgated_coverage_0/group_std_mean": 0.06520479694008827,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0738181822001934,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004972954513505101,
"signal/accgated_coverage_1/centered_abs_mean": 0.04972954392433167,
"signal/accgated_coverage_1/group_std_mean": 0.06520479694008827,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0738181822001934,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004972954513505101,
"signal/accgated_coverage_10/centered_abs_mean": 0.04972954392433167,
"signal/accgated_coverage_10/group_std_mean": 0.06520479694008827,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.0738181822001934,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004972954513505101,
"signal/accgated_coverage_15/centered_abs_mean": 0.04972954392433167,
"signal/accgated_coverage_15/group_std_mean": 0.06520479694008827,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.0738181822001934,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004972954513505101,
"signal/accgated_coverage_20/centered_abs_mean": 0.04972954392433167,
"signal/accgated_coverage_20/group_std_mean": 0.06520479694008827,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.0738181822001934,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.004972954513505101,
"signal/accgated_coverage_25/centered_abs_mean": 0.04972954392433167,
"signal/accgated_coverage_25/group_std_mean": 0.06520479694008827,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.0738181822001934,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.004972954513505101,
"signal/accgated_coverage_5/centered_abs_mean": 0.04972954392433167,
"signal/accgated_coverage_5/group_std_mean": 0.06520479694008827,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.0738181822001934,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004972954513505101,
"signal/accuracy_reward/centered_abs_mean": 0.104913330078125,
"signal/accuracy_reward/group_std_mean": 0.14136361330747604,
"signal/accuracy_reward/group_zero_std_frac": 0.5875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7884438276290894,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0524566650390625,
"signal/advantage_abs_mean": 0.7568434357643128,
"signal/advantage_pre_scale_abs_mean": 0.07361374348402024,
"signal/advantage_pre_scale_std": 0.11564632058143616,
"signal/advantage_std": 0.9830055475234986,
"signal/brier_reward/centered_abs_mean": 0.12480638921260834,
"signal/brier_reward/group_std_mean": 0.1614906221628189,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18828485012054444,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.012480638921260834,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.017779985442757607,
"signal/confidence_uniqueness_reward/group_std_mean": 0.0302621740847826,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02683054730296135,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0017779986374080181,
"signal/format_reward/centered_abs_mean": 0.005804443359375,
"signal/format_reward/group_std_mean": 0.015720795281231402,
"signal/format_reward/group_zero_std_frac": 0.915625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.04360398724675178,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0029022216796875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2762296974658966,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3503751754760742,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.41613702178001405,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027622969821095467,
"step": 75
},
{
"calibration/aurc": 0.22884921663755264,
"calibration/batch_distribution_entropy": 0.940143201628295,
"calibration/batch_entropy_100bins": 0.9472027085636586,
"calibration/batch_entropy_10bins": 0.940143201628295,
"calibration/batch_entropy_50bins": 0.9527094285346083,
"calibration/batch_uniqueness": 0.9453500551884538,
"calibration/buffer_distribution_entropy": 0.9887657988155965,
"calibration/buffer_entropy_100bins": 0.9568530594053511,
"calibration/buffer_entropy_10bins": 0.9887657988155965,
"calibration/buffer_entropy_50bins": 0.971612294122699,
"calibration/confidence_entropy": 0.4441226940602814,
"calibration/coverage@0%": 0.051953125,
"calibration/coverage@1%": 0.0546875,
"calibration/coverage@10%": 0.312890625,
"calibration/coverage@15%": 0.43828125,
"calibration/coverage@20%": 0.5296875,
"calibration/coverage@25%": 0.633984375,
"calibration/coverage@30%": 0.699609375,
"calibration/coverage@5%": 0.192578125,
"calibration/distribution_entropy_10": 0.940143201628295,
"calibration/distribution_entropy_100": 0.9472027085636586,
"calibration/ece": 0.09887963897240112,
"calibration/mean_confidence": 0.5611375089943056,
"calibration/unique_confidence_per_question": 0.9984375,
"calibration/unique_confidences": 511.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00029296875,
"completions/max_length": 1041.4,
"completions/max_terminated_length": 1041.4,
"completions/mean_length": 157.0951171875,
"completions/mean_terminated_length": 157.14105224609375,
"completions/min_length": 28.2,
"completions/min_terminated_length": 68.2,
"epoch": 0.256,
"grad_norm": 0.014753947965800762,
"learning_rate": 1e-06,
"loss": -0.0095,
"num_tokens": 265133872.0,
"reward": 0.931930661201477,
"reward_std": 0.08377985209226609,
"rewards/accgated_coverage_0": 0.031152778863906862,
"rewards/accgated_coverage_1": 0.031152778863906862,
"rewards/accgated_coverage_10": 0.031152778863906862,
"rewards/accgated_coverage_15": 0.031152778863906862,
"rewards/accgated_coverage_20": 0.031152778863906862,
"rewards/accgated_coverage_25": 0.031152778863906862,
"rewards/accgated_coverage_5": 0.031152778863906862,
"rewards/accuracy_reward": 0.5291015625,
"rewards/brier_reward": 0.8004519820213318,
"rewards/confidence_uniqueness_reward": 0.946687388420105,
"rewards/format_reward": 0.99970703125,
"rewards/frontier_entropy_batch_reward": -0.2899448871612549,
"signal/accgated_coverage_0/centered_abs_mean": 0.03722344189882278,
"signal/accgated_coverage_0/group_std_mean": 0.048278050124645235,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06856199353933334,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.003722344199195504,
"signal/accgated_coverage_1/centered_abs_mean": 0.03722344189882278,
"signal/accgated_coverage_1/group_std_mean": 0.048278050124645235,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06856199353933334,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.003722344199195504,
"signal/accgated_coverage_10/centered_abs_mean": 0.03722344189882278,
"signal/accgated_coverage_10/group_std_mean": 0.048278050124645235,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06856199353933334,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003722344199195504,
"signal/accgated_coverage_15/centered_abs_mean": 0.03722344189882278,
"signal/accgated_coverage_15/group_std_mean": 0.048278050124645235,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06856199353933334,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003722344199195504,
"signal/accgated_coverage_20/centered_abs_mean": 0.03722344189882278,
"signal/accgated_coverage_20/group_std_mean": 0.048278050124645235,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.06856199353933334,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.003722344199195504,
"signal/accgated_coverage_25/centered_abs_mean": 0.03722344189882278,
"signal/accgated_coverage_25/group_std_mean": 0.048278050124645235,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.06856199353933334,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.003722344199195504,
"signal/accgated_coverage_5/centered_abs_mean": 0.03722344189882278,
"signal/accgated_coverage_5/group_std_mean": 0.048278050124645235,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06856199353933334,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.003722344199195504,
"signal/accuracy_reward/centered_abs_mean": 0.0857177734375,
"signal/accuracy_reward/group_std_mean": 0.11554728895425796,
"signal/accuracy_reward/group_zero_std_frac": 0.6625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7797396779060364,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.04285888671875,
"signal/advantage_abs_mean": 0.7714757680892944,
"signal/advantage_pre_scale_abs_mean": 0.06428103595972061,
"signal/advantage_pre_scale_std": 0.10194826424121857,
"signal/advantage_std": 0.982670783996582,
"signal/brier_reward/centered_abs_mean": 0.11013664156198502,
"signal/brier_reward/group_std_mean": 0.14308430552482604,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.2023284763097763,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.0110136641189456,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01524948924779892,
"signal/confidence_uniqueness_reward/group_std_mean": 0.019918569177389146,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.028465356677770615,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0015249490039423107,
"signal/format_reward/centered_abs_mean": 0.000567626953125,
"signal/format_reward/group_std_mean": 0.0016572814900428056,
"signal/format_reward/group_zero_std_frac": 0.990625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0053635423071682455,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0002838134765625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.29177509546279906,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.36490934491157534,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5390966415405274,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.029177509248256683,
"step": 80
},
{
"calibration/aurc": 0.36564841608929805,
"calibration/batch_distribution_entropy": 0.957468112154109,
"calibration/batch_entropy_100bins": 0.9566819499229304,
"calibration/batch_entropy_10bins": 0.957468112154109,
"calibration/batch_entropy_50bins": 0.9630234515940392,
"calibration/batch_uniqueness": 0.9490738663425103,
"calibration/buffer_distribution_entropy": 0.9889740169929319,
"calibration/buffer_entropy_100bins": 0.9613072856653708,
"calibration/buffer_entropy_10bins": 0.9889740169929319,
"calibration/buffer_entropy_50bins": 0.9743176890154303,
"calibration/confidence_entropy": 0.48782677996991497,
"calibration/coverage@0%": 0.020349192759295497,
"calibration/coverage@1%": 0.020349192759295497,
"calibration/coverage@10%": 0.11617309809197651,
"calibration/coverage@15%": 0.15489542563600783,
"calibration/coverage@20%": 0.25336350293542076,
"calibration/coverage@25%": 0.3303731959393346,
"calibration/coverage@30%": 0.4034774033757339,
"calibration/coverage@5%": 0.028959760273972603,
"calibration/distribution_entropy_10": 0.957468112154109,
"calibration/distribution_entropy_100": 0.9566819499229304,
"calibration/ece": 0.10895358361970231,
"calibration/mean_confidence": 0.5134070589388305,
"calibration/unique_confidence_per_question": 0.998046875,
"calibration/unique_confidences": 511.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00048828125,
"completions/max_length": 1094.0,
"completions/max_terminated_length": 1094.0,
"completions/mean_length": 164.3265625,
"completions/mean_terminated_length": 164.40765991210938,
"completions/min_length": 14.0,
"completions/min_terminated_length": 68.4,
"epoch": 0.272,
"grad_norm": 0.026652803644537926,
"learning_rate": 1e-06,
"loss": -0.0025,
"num_tokens": 281782272.0,
"reward": 0.9219348549842834,
"reward_std": 0.07897711098194123,
"rewards/accgated_coverage_0": 0.035489321872591975,
"rewards/accgated_coverage_1": 0.035489321872591975,
"rewards/accgated_coverage_10": 0.035489321872591975,
"rewards/accgated_coverage_15": 0.035489321872591975,
"rewards/accgated_coverage_20": 0.035489321872591975,
"rewards/accgated_coverage_25": 0.035489321872591975,
"rewards/accgated_coverage_5": 0.035489321872591975,
"rewards/accuracy_reward": 0.4953125,
"rewards/brier_reward": 0.8048377990722656,
"rewards/confidence_uniqueness_reward": 0.9480871677398681,
"rewards/format_reward": 0.99951171875,
"rewards/frontier_entropy_batch_reward": -0.2561229050159454,
"signal/accgated_coverage_0/centered_abs_mean": 0.03831992819905281,
"signal/accgated_coverage_0/group_std_mean": 0.04869794398546219,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07128713354468345,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0038319929037243126,
"signal/accgated_coverage_1/centered_abs_mean": 0.03831992819905281,
"signal/accgated_coverage_1/group_std_mean": 0.04869794398546219,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07128713354468345,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0038319929037243126,
"signal/accgated_coverage_10/centered_abs_mean": 0.03831992819905281,
"signal/accgated_coverage_10/group_std_mean": 0.04869794398546219,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07128713354468345,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0038319929037243126,
"signal/accgated_coverage_15/centered_abs_mean": 0.03831992819905281,
"signal/accgated_coverage_15/group_std_mean": 0.04869794398546219,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.07128713354468345,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0038319929037243126,
"signal/accgated_coverage_20/centered_abs_mean": 0.03831992819905281,
"signal/accgated_coverage_20/group_std_mean": 0.04869794398546219,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.07128713354468345,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0038319929037243126,
"signal/accgated_coverage_25/centered_abs_mean": 0.03831992819905281,
"signal/accgated_coverage_25/group_std_mean": 0.04869794398546219,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.07128713354468345,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0038319929037243126,
"signal/accgated_coverage_5/centered_abs_mean": 0.03831992819905281,
"signal/accgated_coverage_5/group_std_mean": 0.04869794398546219,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07128713354468345,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0038319929037243126,
"signal/accuracy_reward/centered_abs_mean": 0.09384765625,
"signal/accuracy_reward/group_std_mean": 0.1219882532954216,
"signal/accuracy_reward/group_zero_std_frac": 0.659375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.8743485927581787,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.046923828125,
"signal/advantage_abs_mean": 0.7757304549217224,
"signal/advantage_pre_scale_abs_mean": 0.06117977201938629,
"signal/advantage_pre_scale_std": 0.09556406885385513,
"signal/advantage_std": 0.9826070427894592,
"signal/brier_reward/centered_abs_mean": 0.10229732543230056,
"signal/brier_reward/group_std_mean": 0.1321443647146225,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19441723823547363,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010229732654988766,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01509471219033003,
"signal/confidence_uniqueness_reward/group_std_mean": 0.020363390818238257,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02915481775999069,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0015094712376594543,
"signal/format_reward/centered_abs_mean": 0.000946044921875,
"signal/format_reward/group_std_mean": 0.0027621358167380095,
"signal/format_reward/group_zero_std_frac": 0.984375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.008784599788486957,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0004730224609375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2876576542854309,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.35843446254730227,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5512084603309632,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028765766695141793,
"step": 85
},
{
"calibration/aurc": 0.28182059563434,
"calibration/batch_distribution_entropy": 0.9653119622141771,
"calibration/batch_entropy_100bins": 0.9603247640228526,
"calibration/batch_entropy_10bins": 0.9653119622141771,
"calibration/batch_entropy_50bins": 0.9686853582338767,
"calibration/batch_uniqueness": 0.9500460498930237,
"calibration/buffer_distribution_entropy": 0.9902214301106218,
"calibration/buffer_entropy_100bins": 0.965152281306531,
"calibration/buffer_entropy_10bins": 0.9902214301106218,
"calibration/buffer_entropy_50bins": 0.9769406780328171,
"calibration/confidence_entropy": 0.4688582044303282,
"calibration/coverage@0%": 0.059375,
"calibration/coverage@1%": 0.06953125,
"calibration/coverage@10%": 0.1625,
"calibration/coverage@15%": 0.21171875,
"calibration/coverage@20%": 0.30234375,
"calibration/coverage@25%": 0.38010258683953035,
"calibration/coverage@30%": 0.46526036570450097,
"calibration/coverage@5%": 0.111328125,
"calibration/distribution_entropy_10": 0.9653119622141771,
"calibration/distribution_entropy_100": 0.9603247640228526,
"calibration/ece": 0.13033918330650543,
"calibration/mean_confidence": 0.5502270186091901,
"calibration/unique_confidence_per_question": 0.998828125,
"calibration/unique_confidences": 511.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00048828125,
"completions/max_length": 1076.2,
"completions/max_terminated_length": 1076.2,
"completions/mean_length": 169.6982421875,
"completions/mean_terminated_length": 169.78143920898438,
"completions/min_length": 26.8,
"completions/min_terminated_length": 69.0,
"epoch": 0.288,
"grad_norm": 0.016800779849290848,
"learning_rate": 1e-06,
"loss": -0.0069,
"num_tokens": 298478158.0,
"reward": 0.936176872253418,
"reward_std": 0.07905197143554688,
"rewards/accgated_coverage_0": 0.02806321382522583,
"rewards/accgated_coverage_1": 0.02806321382522583,
"rewards/accgated_coverage_10": 0.02806321382522583,
"rewards/accgated_coverage_15": 0.02806321382522583,
"rewards/accgated_coverage_20": 0.02806321382522583,
"rewards/accgated_coverage_25": 0.02806321382522583,
"rewards/accgated_coverage_5": 0.02806321382522583,
"rewards/accuracy_reward": 0.53037109375,
"rewards/brier_reward": 0.8030895829200745,
"rewards/confidence_uniqueness_reward": 0.9500627875328064,
"rewards/format_reward": 0.9994140625,
"rewards/frontier_entropy_batch_reward": -0.23675192296504974,
"signal/accgated_coverage_0/centered_abs_mean": 0.03898896276950836,
"signal/accgated_coverage_0/group_std_mean": 0.051898601651191714,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07011319771409034,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0038988963235169647,
"signal/accgated_coverage_1/centered_abs_mean": 0.03898896276950836,
"signal/accgated_coverage_1/group_std_mean": 0.051898601651191714,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07011319771409034,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0038988963235169647,
"signal/accgated_coverage_10/centered_abs_mean": 0.03898896276950836,
"signal/accgated_coverage_10/group_std_mean": 0.051898601651191714,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07011319771409034,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0038988963235169647,
"signal/accgated_coverage_15/centered_abs_mean": 0.03898896276950836,
"signal/accgated_coverage_15/group_std_mean": 0.051898601651191714,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.07011319771409034,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0038988963235169647,
"signal/accgated_coverage_20/centered_abs_mean": 0.03898896276950836,
"signal/accgated_coverage_20/group_std_mean": 0.051898601651191714,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.07011319771409034,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0038988963235169647,
"signal/accgated_coverage_25/centered_abs_mean": 0.03898896276950836,
"signal/accgated_coverage_25/group_std_mean": 0.051898601651191714,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.07011319771409034,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0038988963235169647,
"signal/accgated_coverage_5/centered_abs_mean": 0.03898896276950836,
"signal/accgated_coverage_5/group_std_mean": 0.051898601651191714,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07011319771409034,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0038988963235169647,
"signal/accuracy_reward/centered_abs_mean": 0.085369873046875,
"signal/accuracy_reward/group_std_mean": 0.12051956355571747,
"signal/accuracy_reward/group_zero_std_frac": 0.625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7666887879371643,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0426849365234375,
"signal/advantage_abs_mean": 0.7691376447677613,
"signal/advantage_pre_scale_abs_mean": 0.06049362197518349,
"signal/advantage_pre_scale_std": 0.0944889172911644,
"signal/advantage_std": 0.982718551158905,
"signal/brier_reward/centered_abs_mean": 0.09939528107643128,
"signal/brier_reward/group_std_mean": 0.1313742071390152,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1792702257633209,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009939528070390224,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.014033356122672558,
"signal/confidence_uniqueness_reward/group_std_mean": 0.01928463317453861,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.025189964473247527,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014033355982974172,
"signal/format_reward/centered_abs_mean": 0.00113525390625,
"signal/format_reward/group_std_mean": 0.0033145629335194827,
"signal/format_reward/group_zero_std_frac": 0.98125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.010245678573846817,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000567626953125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.27966817915439607,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3521494925022125,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5021585345268249,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027966818213462828,
"step": 90
},
{
"calibration/aurc": 0.23723227049722134,
"calibration/batch_distribution_entropy": 0.9422197436527993,
"calibration/batch_entropy_100bins": 0.947783957400096,
"calibration/batch_entropy_10bins": 0.9422197436527993,
"calibration/batch_entropy_50bins": 0.954083116064902,
"calibration/batch_uniqueness": 0.9452045138781496,
"calibration/buffer_distribution_entropy": 0.9900522677228037,
"calibration/buffer_entropy_100bins": 0.9680836998392779,
"calibration/buffer_entropy_10bins": 0.9900522677228037,
"calibration/buffer_entropy_50bins": 0.9787217262984678,
"calibration/confidence_entropy": 0.4621249593732072,
"calibration/coverage@0%": 0.023495698793215912,
"calibration/coverage@1%": 0.023495698793215912,
"calibration/coverage@10%": 0.2475995321073635,
"calibration/coverage@15%": 0.3165586268754077,
"calibration/coverage@20%": 0.49132807104005993,
"calibration/coverage@25%": 0.5872471856893442,
"calibration/coverage@30%": 0.7033937654685163,
"calibration/coverage@5%": 0.10935811653428495,
"calibration/distribution_entropy_10": 0.9422197436527993,
"calibration/distribution_entropy_100": 0.947783957400096,
"calibration/ece": 0.09217412824116121,
"calibration/mean_confidence": 0.5705067030468787,
"calibration/unique_confidence_per_question": 0.99609375,
"calibration/unique_confidences": 510.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0015625,
"completions/max_length": 1207.0,
"completions/max_terminated_length": 1207.0,
"completions/mean_length": 186.823046875,
"completions/mean_terminated_length": 187.12020874023438,
"completions/min_length": 0.0,
"completions/min_terminated_length": 70.4,
"epoch": 0.304,
"grad_norm": 0.010825703851878643,
"learning_rate": 1e-06,
"loss": -0.0065,
"num_tokens": 315321178.0,
"reward": 0.9288538575172425,
"reward_std": 0.08966293185949326,
"rewards/accgated_coverage_0": 0.03332295194268227,
"rewards/accgated_coverage_1": 0.03332295194268227,
"rewards/accgated_coverage_10": 0.03332295194268227,
"rewards/accgated_coverage_15": 0.03332295194268227,
"rewards/accgated_coverage_20": 0.03332295194268227,
"rewards/accgated_coverage_25": 0.03332295194268227,
"rewards/accgated_coverage_5": 0.03332295194268227,
"rewards/accuracy_reward": 0.5171875,
"rewards/brier_reward": 0.7961657524108887,
"rewards/confidence_uniqueness_reward": 0.9471974849700928,
"rewards/format_reward": 0.9982421875,
"rewards/frontier_entropy_batch_reward": -0.26523380279541015,
"signal/accgated_coverage_0/centered_abs_mean": 0.03919667080044746,
"signal/accgated_coverage_0/group_std_mean": 0.05230322778224945,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06422111466526985,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.003919667098671198,
"signal/accgated_coverage_1/centered_abs_mean": 0.03919667080044746,
"signal/accgated_coverage_1/group_std_mean": 0.05230322778224945,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06422111466526985,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.003919667098671198,
"signal/accgated_coverage_10/centered_abs_mean": 0.03919667080044746,
"signal/accgated_coverage_10/group_std_mean": 0.05230322778224945,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06422111466526985,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003919667098671198,
"signal/accgated_coverage_15/centered_abs_mean": 0.03919667080044746,
"signal/accgated_coverage_15/group_std_mean": 0.05230322778224945,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06422111466526985,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003919667098671198,
"signal/accgated_coverage_20/centered_abs_mean": 0.03919667080044746,
"signal/accgated_coverage_20/group_std_mean": 0.05230322778224945,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.06422111466526985,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.003919667098671198,
"signal/accgated_coverage_25/centered_abs_mean": 0.03919667080044746,
"signal/accgated_coverage_25/group_std_mean": 0.05230322778224945,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.06422111466526985,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.003919667098671198,
"signal/accgated_coverage_5/centered_abs_mean": 0.03919667080044746,
"signal/accgated_coverage_5/group_std_mean": 0.05230322778224945,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06422111466526985,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.003919667098671198,
"signal/accuracy_reward/centered_abs_mean": 0.09920654296875,
"signal/accuracy_reward/group_std_mean": 0.1397227242588997,
"signal/accuracy_reward/group_zero_std_frac": 0.565625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.794974434375763,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.049603271484375,
"signal/advantage_abs_mean": 0.7518903851509094,
"signal/advantage_pre_scale_abs_mean": 0.06635611951351165,
"signal/advantage_pre_scale_std": 0.10549905151128769,
"signal/advantage_std": 0.9828608155250549,
"signal/brier_reward/centered_abs_mean": 0.11066959798336029,
"signal/brier_reward/group_std_mean": 0.14510378390550613,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17895411252975463,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.011066960357129573,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.017177840508520604,
"signal/confidence_uniqueness_reward/group_std_mean": 0.026433903351426126,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.028798480331897736,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0017177840461954474,
"signal/format_reward/centered_abs_mean": 0.0033935546875,
"signal/format_reward/group_std_mean": 0.009607380395755172,
"signal/format_reward/group_zero_std_frac": 0.946875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.027656373381614686,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.00169677734375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.3032440781593323,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.37698500156402587,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.504073965549469,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.030324406921863556,
"step": 95
},
{
"calibration/aurc": 0.2275224855872522,
"calibration/batch_distribution_entropy": 0.9753810136394889,
"calibration/batch_entropy_100bins": 0.9648556923284033,
"calibration/batch_entropy_10bins": 0.9753810136394889,
"calibration/batch_entropy_50bins": 0.9722959661127977,
"calibration/batch_uniqueness": 0.9518503373092599,
"calibration/buffer_distribution_entropy": 0.9900236597562163,
"calibration/buffer_entropy_100bins": 0.9706581733267374,
"calibration/buffer_entropy_10bins": 0.9900236597562163,
"calibration/buffer_entropy_50bins": 0.9802897616899188,
"calibration/confidence_entropy": 0.47064194756192534,
"calibration/coverage@0%": 0.05064769504314496,
"calibration/coverage@1%": 0.1294096670470742,
"calibration/coverage@10%": 0.37583722468604336,
"calibration/coverage@15%": 0.4228594189394815,
"calibration/coverage@20%": 0.4687304950835934,
"calibration/coverage@25%": 0.5803865358059552,
"calibration/coverage@30%": 0.6935250769828479,
"calibration/coverage@5%": 0.2833929525886975,
"calibration/distribution_entropy_10": 0.9753810136394889,
"calibration/distribution_entropy_100": 0.9648556923284033,
"calibration/ece": 0.16037479730933518,
"calibration/mean_confidence": 0.5303404974916495,
"calibration/unique_confidence_per_question": 0.996484375,
"calibration/unique_confidences": 510.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00263671875,
"completions/max_length": 1374.2,
"completions/max_terminated_length": 1374.2,
"completions/mean_length": 210.96181640625,
"completions/mean_terminated_length": 211.51818237304687,
"completions/min_length": 0.0,
"completions/min_terminated_length": 77.0,
"epoch": 0.32,
"grad_norm": 0.011879896745085716,
"learning_rate": 1e-06,
"loss": -0.003,
"num_tokens": 332570131.0,
"reward": 0.9364314079284668,
"reward_std": 0.0843667134642601,
"rewards/accgated_coverage_0": 0.024320645071566106,
"rewards/accgated_coverage_1": 0.024320645071566106,
"rewards/accgated_coverage_10": 0.024320645071566106,
"rewards/accgated_coverage_15": 0.024320645071566106,
"rewards/accgated_coverage_20": 0.024320645071566106,
"rewards/accgated_coverage_25": 0.024320645071566106,
"rewards/accgated_coverage_5": 0.024320645071566106,
"rewards/accuracy_reward": 0.5388671875,
"rewards/brier_reward": 0.8030460000038147,
"rewards/confidence_uniqueness_reward": 0.9485006332397461,
"rewards/format_reward": 0.9970703125,
"rewards/frontier_entropy_batch_reward": -0.23716461658477783,
"signal/accgated_coverage_0/centered_abs_mean": 0.04264157675206661,
"signal/accgated_coverage_0/group_std_mean": 0.05623520240187645,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07137260884046555,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004264157637953759,
"signal/accgated_coverage_1/centered_abs_mean": 0.04264157675206661,
"signal/accgated_coverage_1/group_std_mean": 0.05623520240187645,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07137260884046555,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004264157637953759,
"signal/accgated_coverage_10/centered_abs_mean": 0.04264157675206661,
"signal/accgated_coverage_10/group_std_mean": 0.05623520240187645,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07137260884046555,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004264157637953759,
"signal/accgated_coverage_15/centered_abs_mean": 0.04264157675206661,
"signal/accgated_coverage_15/group_std_mean": 0.05623520240187645,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.07137260884046555,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004264157637953759,
"signal/accgated_coverage_20/centered_abs_mean": 0.04264157675206661,
"signal/accgated_coverage_20/group_std_mean": 0.05623520240187645,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.07137260884046555,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.004264157637953759,
"signal/accgated_coverage_25/centered_abs_mean": 0.04264157675206661,
"signal/accgated_coverage_25/group_std_mean": 0.05623520240187645,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.07137260884046555,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.004264157637953759,
"signal/accgated_coverage_5/centered_abs_mean": 0.04264157675206661,
"signal/accgated_coverage_5/group_std_mean": 0.05623520240187645,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07137260884046555,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004264157637953759,
"signal/accuracy_reward/centered_abs_mean": 0.08082275390625,
"signal/accuracy_reward/group_std_mean": 0.11435425579547882,
"signal/accuracy_reward/group_zero_std_frac": 0.646875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.692106032371521,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.040411376953125,
"signal/advantage_abs_mean": 0.7543508172035217,
"signal/advantage_pre_scale_abs_mean": 0.06142271608114243,
"signal/advantage_pre_scale_std": 0.10094139575958253,
"signal/advantage_std": 0.9828068971633911,
"signal/brier_reward/centered_abs_mean": 0.09856459945440292,
"signal/brier_reward/group_std_mean": 0.13214722275733948,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.16857999861240386,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009856459870934486,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.017486925423145293,
"signal/confidence_uniqueness_reward/group_std_mean": 0.029220127686858176,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.029981967806816102,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0017486925702542067,
"signal/format_reward/centered_abs_mean": 0.00557861328125,
"signal/format_reward/group_std_mean": 0.014740523602813482,
"signal/format_reward/group_zero_std_frac": 0.921875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.04804762434214353,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.002789306640625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.29434717893600465,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3675897717475891,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5020530700683594,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.029434718191623688,
"step": 100
},
{
"epoch": 0.32,
"eval_completions/clipped_ratio": 0.006061422413793094,
"eval_completions/max_length": 784.25,
"eval_completions/max_terminated_length": 784.25,
"eval_completions/mean_length": 202.25774383544922,
"eval_completions/mean_terminated_length": 203.5218505859375,
"eval_completions/min_length": 42.75,
"eval_completions/min_terminated_length": 89.0,
"eval_loss": 0.0,
"eval_num_tokens": 332570131.0,
"eval_reward": 0.7964471578598022,
"eval_reward_std": 0.21640057116746902,
"eval_rewards/accgated_coverage_0": 0.0308312950655818,
"eval_rewards/accgated_coverage_1": 0.0308312950655818,
"eval_rewards/accgated_coverage_10": 0.0308312950655818,
"eval_rewards/accgated_coverage_15": 0.0308312950655818,
"eval_rewards/accgated_coverage_20": 0.0308312950655818,
"eval_rewards/accgated_coverage_25": 0.0308312950655818,
"eval_rewards/accgated_coverage_5": 0.0308312950655818,
"eval_rewards/accuracy_reward": 0.41796875,
"eval_rewards/brier_reward": 0.7879670858383179,
"eval_rewards/confidence_uniqueness_reward": 0.8942792564630508,
"eval_rewards/format_reward": 0.994140625,
"eval_rewards/frontier_entropy_batch_reward": -0.994140625,
"eval_runtime": 44.1578,
"eval_samples_per_second": 11.323,
"eval_signal/accgated_coverage_0/centered_abs_mean": 0.06945010460913181,
"eval_signal/accgated_coverage_0/group_std_mean": 0.08581146877259016,
"eval_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.03240918926894665,
"eval_signal/accgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.006945010623894632,
"eval_signal/accgated_coverage_1/centered_abs_mean": 0.06945010460913181,
"eval_signal/accgated_coverage_1/group_std_mean": 0.08581146877259016,
"eval_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.03240918926894665,
"eval_signal/accgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.006945010623894632,
"eval_signal/accgated_coverage_10/centered_abs_mean": 0.06945010460913181,
"eval_signal/accgated_coverage_10/group_std_mean": 0.08581146877259016,
"eval_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.03240918926894665,
"eval_signal/accgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.006945010623894632,
"eval_signal/accgated_coverage_15/centered_abs_mean": 0.06945010460913181,
"eval_signal/accgated_coverage_15/group_std_mean": 0.08581146877259016,
"eval_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.03240918926894665,
"eval_signal/accgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.006945010623894632,
"eval_signal/accgated_coverage_20/centered_abs_mean": 0.06945010460913181,
"eval_signal/accgated_coverage_20/group_std_mean": 0.08581146877259016,
"eval_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03240918926894665,
"eval_signal/accgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.006945010623894632,
"eval_signal/accgated_coverage_25/centered_abs_mean": 0.06945010460913181,
"eval_signal/accgated_coverage_25/group_std_mean": 0.08581146877259016,
"eval_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03240918926894665,
"eval_signal/accgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.006945010623894632,
"eval_signal/accgated_coverage_5/centered_abs_mean": 0.06945010460913181,
"eval_signal/accgated_coverage_5/group_std_mean": 0.08581146877259016,
"eval_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.03240918926894665,
"eval_signal/accgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.006945010623894632,
"eval_signal/accuracy_reward/centered_abs_mean": 0.461669921875,
"eval_signal/accuracy_reward/group_std_mean": 0.4874906465411186,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.0757964253425598,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.2308349609375,
"eval_signal/advantage_abs_mean": 0.8804308474063873,
"eval_signal/advantage_pre_scale_abs_mean": 0.19009095057845116,
"eval_signal/advantage_pre_scale_std": 0.2147204391658306,
"eval_signal/advantage_std": 0.9876579642295837,
"eval_signal/brier_reward/centered_abs_mean": 0.20276524871587753,
"eval_signal/brier_reward/group_std_mean": 0.2558363601565361,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.09448201768100262,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.020276525989174843,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.04477469716221094,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.0656670443713665,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.020837026182562113,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004477469890844077,
"eval_signal/format_reward/centered_abs_mean": 0.0113525390625,
"eval_signal/format_reward/group_std_mean": 0.03314562980085611,
"eval_signal/format_reward/group_zero_std_frac": 0.8125,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.026225777342915535,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.00567626953125,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0113525390625,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.03314562980085611,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.8125,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.005245155654847622,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0011352539586368948,
"eval_steps_per_second": 0.091,
"step": 100
},
{
"epoch": 0.32,
"step": 100,
"train_probe_completions/clipped_ratio": 0.006061422413793094,
"train_probe_completions/max_length": 1097.5,
"train_probe_completions/max_terminated_length": 1097.5,
"train_probe_completions/mean_length": 211.7452163696289,
"train_probe_completions/mean_terminated_length": 213.03786087036133,
"train_probe_completions/min_length": 25.0,
"train_probe_completions/min_terminated_length": 90.25,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 332570131.0,
"train_probe_reward": 0.875739112496376,
"train_probe_reward_std": 0.2000570446252823,
"train_probe_rewards/accgated_coverage_0": 0.03413898125290871,
"train_probe_rewards/accgated_coverage_1": 0.03413898125290871,
"train_probe_rewards/accgated_coverage_10": 0.03413898125290871,
"train_probe_rewards/accgated_coverage_15": 0.03413898125290871,
"train_probe_rewards/accgated_coverage_20": 0.03413898125290871,
"train_probe_rewards/accgated_coverage_25": 0.03413898125290871,
"train_probe_rewards/accgated_coverage_5": 0.03413898125290871,
"train_probe_rewards/accuracy_reward": 0.5703125,
"train_probe_rewards/brier_reward": 0.8066689521074295,
"train_probe_rewards/confidence_uniqueness_reward": 0.8914366960525513,
"train_probe_rewards/format_reward": 0.9921875,
"train_probe_rewards/frontier_entropy_batch_reward": -0.9921875,
"train_probe_runtime": 51.3464,
"train_probe_samples_per_second": 9.738,
"train_probe_signal/accgated_coverage_0/centered_abs_mean": 0.12030644342303276,
"train_probe_signal/accgated_coverage_0/group_std_mean": 0.15389259718358517,
"train_probe_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06323447171598673,
"train_probe_signal/accgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.012030644342303276,
"train_probe_signal/accgated_coverage_1/centered_abs_mean": 0.12030644342303276,
"train_probe_signal/accgated_coverage_1/group_std_mean": 0.15389259718358517,
"train_probe_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06323447171598673,
"train_probe_signal/accgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.012030644342303276,
"train_probe_signal/accgated_coverage_10/centered_abs_mean": 0.12030644342303276,
"train_probe_signal/accgated_coverage_10/group_std_mean": 0.15389259718358517,
"train_probe_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06323447171598673,
"train_probe_signal/accgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.012030644342303276,
"train_probe_signal/accgated_coverage_15/centered_abs_mean": 0.12030644342303276,
"train_probe_signal/accgated_coverage_15/group_std_mean": 0.15389259718358517,
"train_probe_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06323447171598673,
"train_probe_signal/accgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.012030644342303276,
"train_probe_signal/accgated_coverage_20/centered_abs_mean": 0.12030644342303276,
"train_probe_signal/accgated_coverage_20/group_std_mean": 0.15389259718358517,
"train_probe_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.06323447171598673,
"train_probe_signal/accgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.012030644342303276,
"train_probe_signal/accgated_coverage_25/centered_abs_mean": 0.12030644342303276,
"train_probe_signal/accgated_coverage_25/group_std_mean": 0.15389259718358517,
"train_probe_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.06323447171598673,
"train_probe_signal/accgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.012030644342303276,
"train_probe_signal/accgated_coverage_5/centered_abs_mean": 0.12030644342303276,
"train_probe_signal/accgated_coverage_5/group_std_mean": 0.15389259718358517,
"train_probe_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06323447171598673,
"train_probe_signal/accgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.012030644342303276,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.476806640625,
"train_probe_signal/accuracy_reward/group_std_mean": 0.4959462359547615,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.223237693309784,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.2384033203125,
"train_probe_signal/advantage_abs_mean": 0.8665360659360886,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.17254489660263062,
"train_probe_signal/advantage_pre_scale_std": 0.19940952584147453,
"train_probe_signal/advantage_std": 0.9876102358102798,
"train_probe_signal/brier_reward/centered_abs_mean": 0.1892232410609722,
"train_probe_signal/brier_reward/group_std_mean": 0.24852639064192772,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.09680720418691635,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.01892232336103916,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.04895491898059845,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.07097366638481617,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.024634152185171843,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004895492049399763,
"train_probe_signal/format_reward/centered_abs_mean": 0.014892578125,
"train_probe_signal/format_reward/group_std_mean": 0.037468004040420055,
"train_probe_signal/format_reward/group_zero_std_frac": 0.8125,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.03672218322753906,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.0074462890625,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.014892578125,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.037468004040420055,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.8125,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.007344436482526362,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0014892578474245965,
"train_probe_steps_per_second": 0.078
},
{
"calibration/aurc": 0.2938701333815237,
"calibration/batch_distribution_entropy": 0.9630968456387267,
"calibration/batch_entropy_100bins": 0.9608776222329464,
"calibration/batch_entropy_10bins": 0.9630968456387267,
"calibration/batch_entropy_50bins": 0.9673227230811096,
"calibration/batch_uniqueness": 0.9498737233245051,
"calibration/buffer_distribution_entropy": 0.991554186686024,
"calibration/buffer_entropy_100bins": 0.9760885578242668,
"calibration/buffer_entropy_10bins": 0.991554186686024,
"calibration/buffer_entropy_50bins": 0.9839386471662424,
"calibration/confidence_entropy": 0.47559780956426767,
"calibration/coverage@0%": 0.028955257775762834,
"calibration/coverage@1%": 0.028955257775762834,
"calibration/coverage@10%": 0.07320167182147234,
"calibration/coverage@15%": 0.17581575308235764,
"calibration/coverage@20%": 0.25887114260989535,
"calibration/coverage@25%": 0.32191505653314967,
"calibration/coverage@30%": 0.6319862988903678,
"calibration/coverage@5%": 0.04224950312801332,
"calibration/distribution_entropy_10": 0.9630968456387267,
"calibration/distribution_entropy_100": 0.9608776222329464,
"calibration/ece": 0.13256246993194612,
"calibration/mean_confidence": 0.5165659248356935,
"calibration/unique_confidence_per_question": 0.995703125,
"calibration/unique_confidences": 509.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00302734375,
"completions/max_length": 1202.2,
"completions/max_terminated_length": 1202.2,
"completions/mean_length": 204.77138671875,
"completions/mean_terminated_length": 205.40043029785156,
"completions/min_length": 0.0,
"completions/min_terminated_length": 67.4,
"epoch": 0.336,
"grad_norm": 0.013651929795742035,
"learning_rate": 1e-06,
"loss": -0.0124,
"num_tokens": 349389422.0,
"reward": 0.9386118412017822,
"reward_std": 0.08805803209543228,
"rewards/accgated_coverage_0": 0.028051501139998436,
"rewards/accgated_coverage_1": 0.028051501139998436,
"rewards/accgated_coverage_10": 0.028051501139998436,
"rewards/accgated_coverage_15": 0.028051501139998436,
"rewards/accgated_coverage_20": 0.028051501139998436,
"rewards/accgated_coverage_25": 0.028051501139998436,
"rewards/accgated_coverage_5": 0.028051501139998436,
"rewards/accuracy_reward": 0.53994140625,
"rewards/brier_reward": 0.8074901103973389,
"rewards/confidence_uniqueness_reward": 0.9470890641212464,
"rewards/format_reward": 0.996875,
"rewards/frontier_entropy_batch_reward": -0.2489030748605728,
"signal/accgated_coverage_0/centered_abs_mean": 0.037933540716767314,
"signal/accgated_coverage_0/group_std_mean": 0.05019458085298538,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06445568054914474,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0037933541927486656,
"signal/accgated_coverage_1/centered_abs_mean": 0.037933540716767314,
"signal/accgated_coverage_1/group_std_mean": 0.05019458085298538,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06445568054914474,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0037933541927486656,
"signal/accgated_coverage_10/centered_abs_mean": 0.037933540716767314,
"signal/accgated_coverage_10/group_std_mean": 0.05019458085298538,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06445568054914474,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0037933541927486656,
"signal/accgated_coverage_15/centered_abs_mean": 0.037933540716767314,
"signal/accgated_coverage_15/group_std_mean": 0.05019458085298538,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06445568054914474,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0037933541927486656,
"signal/accgated_coverage_20/centered_abs_mean": 0.037933540716767314,
"signal/accgated_coverage_20/group_std_mean": 0.05019458085298538,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.06445568054914474,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0037933541927486656,
"signal/accgated_coverage_25/centered_abs_mean": 0.037933540716767314,
"signal/accgated_coverage_25/group_std_mean": 0.05019458085298538,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.06445568054914474,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0037933541927486656,
"signal/accgated_coverage_5/centered_abs_mean": 0.037933540716767314,
"signal/accgated_coverage_5/group_std_mean": 0.05019458085298538,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06445568054914474,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0037933541927486656,
"signal/accuracy_reward/centered_abs_mean": 0.090362548828125,
"signal/accuracy_reward/group_std_mean": 0.1286356642842293,
"signal/accuracy_reward/group_zero_std_frac": 0.596875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7711039543151855,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0451812744140625,
"signal/advantage_abs_mean": 0.7434520244598388,
"signal/advantage_pre_scale_abs_mean": 0.06324905827641487,
"signal/advantage_pre_scale_std": 0.10626154541969299,
"signal/advantage_std": 0.9827974557876586,
"signal/brier_reward/centered_abs_mean": 0.10222946107387543,
"signal/brier_reward/group_std_mean": 0.13567868024110794,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17583132088184356,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010222946479916572,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01779725756496191,
"signal/confidence_uniqueness_reward/group_std_mean": 0.030765137821435928,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.030492952838540078,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0017797257518395782,
"signal/format_reward/centered_abs_mean": 0.00601806640625,
"signal/format_reward/group_std_mean": 0.016668744198977948,
"signal/format_reward/group_zero_std_frac": 0.909375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.050790101289749146,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.003009033203125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.283295339345932,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3543296158313751,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.48725026845932007,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028329535201191903,
"step": 105
},
{
"calibration/aurc": 0.25580926136676185,
"calibration/batch_distribution_entropy": 0.9227617339635762,
"calibration/batch_entropy_100bins": 0.9392732978918342,
"calibration/batch_entropy_10bins": 0.9227617339635762,
"calibration/batch_entropy_50bins": 0.9424297493514396,
"calibration/batch_uniqueness": 0.9405713721930222,
"calibration/buffer_distribution_entropy": 0.9943717800566546,
"calibration/buffer_entropy_100bins": 0.9858840653493164,
"calibration/buffer_entropy_10bins": 0.9943717800566546,
"calibration/buffer_entropy_50bins": 0.9903590636965237,
"calibration/confidence_entropy": 0.4072741298802326,
"calibration/coverage@0%": 0.10740467471437394,
"calibration/coverage@1%": 0.12196545227342903,
"calibration/coverage@10%": 0.3713839957735062,
"calibration/coverage@15%": 0.42708286291878955,
"calibration/coverage@20%": 0.47455101947274975,
"calibration/coverage@25%": 0.5345728394704339,
"calibration/coverage@30%": 0.5855858952833101,
"calibration/coverage@5%": 0.23568821647753593,
"calibration/distribution_entropy_10": 0.9227617339635762,
"calibration/distribution_entropy_100": 0.9392732978918342,
"calibration/ece": 0.1158637535934676,
"calibration/mean_confidence": 0.46497006452686873,
"calibration/unique_confidence_per_question": 0.9921875,
"calibration/unique_confidences": 508.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00419921875,
"completions/max_length": 1446.6,
"completions/max_terminated_length": 1446.6,
"completions/mean_length": 216.84267578125,
"completions/mean_terminated_length": 217.76553649902343,
"completions/min_length": 0.0,
"completions/min_terminated_length": 71.0,
"epoch": 0.352,
"grad_norm": 0.00977289117872715,
"learning_rate": 1e-06,
"loss": -0.0217,
"num_tokens": 366870307.0,
"reward": 0.9073002219200135,
"reward_std": 0.08678219318389893,
"rewards/accgated_coverage_0": 0.03460534028708935,
"rewards/accgated_coverage_1": 0.03460534028708935,
"rewards/accgated_coverage_10": 0.03460534028708935,
"rewards/accgated_coverage_15": 0.03460534028708935,
"rewards/accgated_coverage_20": 0.03460534028708935,
"rewards/accgated_coverage_25": 0.03460534028708935,
"rewards/accgated_coverage_5": 0.03460534028708935,
"rewards/accuracy_reward": 0.47412109375,
"rewards/brier_reward": 0.8070560097694397,
"rewards/confidence_uniqueness_reward": 0.9437341690063477,
"rewards/format_reward": 0.99580078125,
"rewards/frontier_entropy_batch_reward": -0.26963459253311156,
"signal/accgated_coverage_0/centered_abs_mean": 0.03781263530254364,
"signal/accgated_coverage_0/group_std_mean": 0.048241587728261946,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.068487998098135,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0037812636233866214,
"signal/accgated_coverage_1/centered_abs_mean": 0.03781263530254364,
"signal/accgated_coverage_1/group_std_mean": 0.048241587728261946,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.068487998098135,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0037812636233866214,
"signal/accgated_coverage_10/centered_abs_mean": 0.03781263530254364,
"signal/accgated_coverage_10/group_std_mean": 0.048241587728261946,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.068487998098135,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0037812636233866214,
"signal/accgated_coverage_15/centered_abs_mean": 0.03781263530254364,
"signal/accgated_coverage_15/group_std_mean": 0.048241587728261946,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.068487998098135,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0037812636233866214,
"signal/accgated_coverage_20/centered_abs_mean": 0.03781263530254364,
"signal/accgated_coverage_20/group_std_mean": 0.048241587728261946,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.068487998098135,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0037812636233866214,
"signal/accgated_coverage_25/centered_abs_mean": 0.03781263530254364,
"signal/accgated_coverage_25/group_std_mean": 0.048241587728261946,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.068487998098135,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0037812636233866214,
"signal/accgated_coverage_5/centered_abs_mean": 0.03781263530254364,
"signal/accgated_coverage_5/group_std_mean": 0.048241587728261946,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.068487998098135,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0037812636233866214,
"signal/accuracy_reward/centered_abs_mean": 0.091729736328125,
"signal/accuracy_reward/group_std_mean": 0.12285660356283187,
"signal/accuracy_reward/group_zero_std_frac": 0.64375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.8430713295936585,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0458648681640625,
"signal/advantage_abs_mean": 0.7439336776733398,
"signal/advantage_pre_scale_abs_mean": 0.06269238516688347,
"signal/advantage_pre_scale_std": 0.1061941295862198,
"signal/advantage_std": 0.9826473712921142,
"signal/brier_reward/centered_abs_mean": 0.10678147226572036,
"signal/brier_reward/group_std_mean": 0.1384610041975975,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19711463749408722,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.01067814752459526,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.02080516703426838,
"signal/confidence_uniqueness_reward/group_std_mean": 0.03538321666419506,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03896758854389191,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0020805167499929667,
"signal/format_reward/centered_abs_mean": 0.007916259765625,
"signal/format_reward/group_std_mean": 0.020008775033056737,
"signal/format_reward/group_zero_std_frac": 0.896875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.07236479222774506,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0039581298828125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2849461555480957,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3591804265975952,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5330600082874298,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028494616225361823,
"step": 110
},
{
"calibration/aurc": 0.3293896671089328,
"calibration/batch_distribution_entropy": 0.9508485949667899,
"calibration/batch_entropy_100bins": 0.9538678049286977,
"calibration/batch_entropy_10bins": 0.9508485949667899,
"calibration/batch_entropy_50bins": 0.9609589855979414,
"calibration/batch_uniqueness": 0.948158174830542,
"calibration/buffer_distribution_entropy": 0.9951915945638051,
"calibration/buffer_entropy_100bins": 0.9923308874972705,
"calibration/buffer_entropy_10bins": 0.9951915945638051,
"calibration/buffer_entropy_50bins": 0.994128764544494,
"calibration/confidence_entropy": 0.46145551338284296,
"calibration/coverage@0%": 0.041891017872838845,
"calibration/coverage@1%": 0.06498299438947289,
"calibration/coverage@10%": 0.1625035538687481,
"calibration/coverage@15%": 0.18717492899898697,
"calibration/coverage@20%": 0.3179328327202623,
"calibration/coverage@25%": 0.3853010271385415,
"calibration/coverage@30%": 0.4731552672548049,
"calibration/coverage@5%": 0.12450340777053163,
"calibration/distribution_entropy_10": 0.9508485949667899,
"calibration/distribution_entropy_100": 0.9538678049286977,
"calibration/ece": 0.11797307021881791,
"calibration/mean_confidence": 0.48178881792594136,
"calibration/unique_confidence_per_question": 0.992578125,
"calibration/unique_confidences": 508.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00673828125,
"completions/max_length": 1455.8,
"completions/max_terminated_length": 1455.8,
"completions/mean_length": 235.89619140625,
"completions/mean_terminated_length": 237.49516906738282,
"completions/min_length": 0.0,
"completions/min_terminated_length": 78.2,
"epoch": 0.368,
"grad_norm": 0.01386299915611744,
"learning_rate": 1e-06,
"loss": -0.0361,
"num_tokens": 384351356.0,
"reward": 0.9174055337905884,
"reward_std": 0.09238043874502182,
"rewards/accgated_coverage_0": 0.034235814958810805,
"rewards/accgated_coverage_1": 0.034235814958810805,
"rewards/accgated_coverage_10": 0.034235814958810805,
"rewards/accgated_coverage_15": 0.034235814958810805,
"rewards/accgated_coverage_20": 0.034235814958810805,
"rewards/accgated_coverage_25": 0.03332972377538681,
"rewards/accgated_coverage_5": 0.034235814958810805,
"rewards/accuracy_reward": 0.49736328125,
"rewards/brier_reward": 0.798619556427002,
"rewards/confidence_uniqueness_reward": 0.9427231550216675,
"rewards/format_reward": 0.9931640625,
"rewards/frontier_entropy_batch_reward": -0.25866886377334597,
"signal/accgated_coverage_0/centered_abs_mean": 0.03830415830016136,
"signal/accgated_coverage_0/group_std_mean": 0.05027427449822426,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06328124478459358,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0038304158952087164,
"signal/accgated_coverage_1/centered_abs_mean": 0.03830415830016136,
"signal/accgated_coverage_1/group_std_mean": 0.05027427449822426,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06328124478459358,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0038304158952087164,
"signal/accgated_coverage_10/centered_abs_mean": 0.03830415830016136,
"signal/accgated_coverage_10/group_std_mean": 0.05027427449822426,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06328124478459358,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0038304158952087164,
"signal/accgated_coverage_15/centered_abs_mean": 0.03830415830016136,
"signal/accgated_coverage_15/group_std_mean": 0.05027427449822426,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06328124478459358,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0038304158952087164,
"signal/accgated_coverage_20/centered_abs_mean": 0.03830415830016136,
"signal/accgated_coverage_20/group_std_mean": 0.05027427449822426,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.06328124478459358,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0038304158952087164,
"signal/accgated_coverage_25/centered_abs_mean": 0.036296261474490166,
"signal/accgated_coverage_25/group_std_mean": 0.04763664044439793,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.060014798492193225,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0036296261940151455,
"signal/accgated_coverage_5/centered_abs_mean": 0.03830415830016136,
"signal/accgated_coverage_5/group_std_mean": 0.05027427449822426,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06328124478459358,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0038304158952087164,
"signal/accuracy_reward/centered_abs_mean": 0.085137939453125,
"signal/accuracy_reward/group_std_mean": 0.12216567993164062,
"signal/accuracy_reward/group_zero_std_frac": 0.609375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.703400194644928,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0425689697265625,
"signal/advantage_abs_mean": 0.728959834575653,
"signal/advantage_pre_scale_abs_mean": 0.06334649622440339,
"signal/advantage_pre_scale_std": 0.11278722882270813,
"signal/advantage_std": 0.9828591585159302,
"signal/brier_reward/centered_abs_mean": 0.10448271930217742,
"signal/brier_reward/group_std_mean": 0.1404559552669525,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1725900113582611,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010448272339999676,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.023952210694551467,
"signal/confidence_uniqueness_reward/group_std_mean": 0.043306180834770204,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.039654774218797685,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0023952211253345014,
"signal/format_reward/centered_abs_mean": 0.01273193359375,
"signal/format_reward/group_std_mean": 0.030019040033221246,
"signal/format_reward/group_zero_std_frac": 0.853125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.10542392134666442,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.006365966796875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.29640886187553406,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3688334941864014,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.49248443841934203,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.029640886560082437,
"step": 115
},
{
"calibration/aurc": 0.3378154411514138,
"calibration/batch_distribution_entropy": 0.9487233111538966,
"calibration/batch_entropy_100bins": 0.9526827948812592,
"calibration/batch_entropy_10bins": 0.9487233111538966,
"calibration/batch_entropy_50bins": 0.9588120543508925,
"calibration/batch_uniqueness": 0.9468980138464866,
"calibration/buffer_distribution_entropy": 0.9948010183710958,
"calibration/buffer_entropy_100bins": 0.9955311325969649,
"calibration/buffer_entropy_10bins": 0.9948010183710958,
"calibration/buffer_entropy_50bins": 0.9956998186491941,
"calibration/confidence_entropy": 0.439748119843647,
"calibration/coverage@0%": 0.011041307324958463,
"calibration/coverage@1%": 0.011041307324958463,
"calibration/coverage@10%": 0.20479649081669446,
"calibration/coverage@15%": 0.2716135799171021,
"calibration/coverage@20%": 0.33228202082760394,
"calibration/coverage@25%": 0.37090030440994626,
"calibration/coverage@30%": 0.4248466235426117,
"calibration/coverage@5%": 0.0761019240169633,
"calibration/distribution_entropy_10": 0.9487233111538966,
"calibration/distribution_entropy_100": 0.9526827948812592,
"calibration/ece": 0.15366528241445918,
"calibration/mean_confidence": 0.44855419570434396,
"calibration/unique_confidence_per_question": 0.984765625,
"calibration/unique_confidences": 504.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00625,
"completions/max_length": 1482.6,
"completions/max_terminated_length": 1482.6,
"completions/mean_length": 256.2103515625,
"completions/mean_terminated_length": 257.847509765625,
"completions/min_length": 0.0,
"completions/min_terminated_length": 80.4,
"epoch": 0.384,
"grad_norm": 0.009529582224786282,
"learning_rate": 1e-06,
"loss": -0.0213,
"num_tokens": 401831462.0,
"reward": 0.933357572555542,
"reward_std": 0.09187148809432984,
"rewards/accgated_coverage_0": 0.028487255796790123,
"rewards/accgated_coverage_1": 0.028487255796790123,
"rewards/accgated_coverage_10": 0.028487255796790123,
"rewards/accgated_coverage_15": 0.028487255796790123,
"rewards/accgated_coverage_20": 0.028487255796790123,
"rewards/accgated_coverage_25": 0.02408354915678501,
"rewards/accgated_coverage_5": 0.028487255796790123,
"rewards/accuracy_reward": 0.53427734375,
"rewards/brier_reward": 0.8014801621437073,
"rewards/confidence_uniqueness_reward": 0.9445072770118713,
"rewards/format_reward": 0.99365234375,
"rewards/frontier_entropy_batch_reward": -0.24706726670265197,
"signal/accgated_coverage_0/centered_abs_mean": 0.04390236437320709,
"signal/accgated_coverage_0/group_std_mean": 0.05707613006234169,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07272210046648979,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004390236409381032,
"signal/accgated_coverage_1/centered_abs_mean": 0.04390236437320709,
"signal/accgated_coverage_1/group_std_mean": 0.05707613006234169,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07272210046648979,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004390236409381032,
"signal/accgated_coverage_10/centered_abs_mean": 0.04390236437320709,
"signal/accgated_coverage_10/group_std_mean": 0.05707613006234169,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07272210046648979,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004390236409381032,
"signal/accgated_coverage_15/centered_abs_mean": 0.04390236437320709,
"signal/accgated_coverage_15/group_std_mean": 0.05707613006234169,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.07272210046648979,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004390236409381032,
"signal/accgated_coverage_20/centered_abs_mean": 0.04390236437320709,
"signal/accgated_coverage_20/group_std_mean": 0.05707613006234169,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.07272210046648979,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.004390236409381032,
"signal/accgated_coverage_25/centered_abs_mean": 0.03483487144112587,
"signal/accgated_coverage_25/group_std_mean": 0.04554118439555168,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.057788226008415225,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0034834872465580704,
"signal/accgated_coverage_5/centered_abs_mean": 0.04390236437320709,
"signal/accgated_coverage_5/group_std_mean": 0.05707613006234169,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07272210046648979,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004390236409381032,
"signal/accuracy_reward/centered_abs_mean": 0.090350341796875,
"signal/accuracy_reward/group_std_mean": 0.12406320720911027,
"signal/accuracy_reward/group_zero_std_frac": 0.625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7498820066452027,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0451751708984375,
"signal/advantage_abs_mean": 0.7355457067489624,
"signal/advantage_pre_scale_abs_mean": 0.06460903808474541,
"signal/advantage_pre_scale_std": 0.1108278751373291,
"signal/advantage_std": 0.9828528165817261,
"signal/brier_reward/centered_abs_mean": 0.10364313423633575,
"signal/brier_reward/group_std_mean": 0.14015646576881408,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1724245071411133,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.01036431360989809,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.022669042646884918,
"signal/confidence_uniqueness_reward/group_std_mean": 0.04051032140851021,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.037741868570446965,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0022669042460620402,
"signal/format_reward/centered_abs_mean": 0.011773681640625,
"signal/format_reward/group_std_mean": 0.027581943944096566,
"signal/format_reward/group_zero_std_frac": 0.865625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.09781504720449448,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0058868408203125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.29197040796279905,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.36490694880485536,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.48627070188522337,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02919704169034958,
"step": 120
},
{
"calibration/aurc": 0.35458099000686955,
"calibration/batch_distribution_entropy": 0.9782974907022506,
"calibration/batch_entropy_100bins": 0.9709587609021237,
"calibration/batch_entropy_10bins": 0.9782974907022506,
"calibration/batch_entropy_50bins": 0.978894516984996,
"calibration/batch_uniqueness": 0.9536125976688214,
"calibration/buffer_distribution_entropy": 0.994230281802456,
"calibration/buffer_entropy_100bins": 0.99641344498817,
"calibration/buffer_entropy_10bins": 0.994230281802456,
"calibration/buffer_entropy_50bins": 0.9960766242803129,
"calibration/confidence_entropy": 0.4856963493143199,
"calibration/coverage@0%": 0.022826808972282786,
"calibration/coverage@1%": 0.022826808972282786,
"calibration/coverage@10%": 0.07332849552614941,
"calibration/coverage@15%": 0.17081343517586742,
"calibration/coverage@20%": 0.22748826411536247,
"calibration/coverage@25%": 0.2906068158242979,
"calibration/coverage@30%": 0.3947990232036716,
"calibration/coverage@5%": 0.02713209272962134,
"calibration/distribution_entropy_10": 0.9782974907022506,
"calibration/distribution_entropy_100": 0.9709587609021237,
"calibration/ece": 0.12960821365713704,
"calibration/mean_confidence": 0.5334309403028674,
"calibration/unique_confidence_per_question": 0.983203125,
"calibration/unique_confidences": 503.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0095703125,
"completions/max_length": 1478.6,
"completions/max_terminated_length": 1478.6,
"completions/mean_length": 271.75009765625,
"completions/mean_terminated_length": 274.3923858642578,
"completions/min_length": 0.0,
"completions/min_terminated_length": 86.8,
"epoch": 0.4,
"grad_norm": 0.007571675349026918,
"learning_rate": 1e-06,
"loss": -0.0412,
"num_tokens": 419650631.0,
"reward": 0.923571789264679,
"reward_std": 0.10908291041851044,
"rewards/accgated_coverage_0": 0.02695066127926111,
"rewards/accgated_coverage_1": 0.02695066127926111,
"rewards/accgated_coverage_10": 0.02695066127926111,
"rewards/accgated_coverage_15": 0.026933599449694156,
"rewards/accgated_coverage_20": 0.02603110745549202,
"rewards/accgated_coverage_25": 0.020951040834188462,
"rewards/accgated_coverage_5": 0.02695066127926111,
"rewards/accuracy_reward": 0.52734375,
"rewards/brier_reward": 0.7903018236160279,
"rewards/confidence_uniqueness_reward": 0.9398616194725037,
"rewards/format_reward": 0.9904296875,
"rewards/frontier_entropy_batch_reward": -0.2650311678647995,
"signal/accgated_coverage_0/centered_abs_mean": 0.04150879755616188,
"signal/accgated_coverage_0/group_std_mean": 0.053798314183950424,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.060856424272060394,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004150879895314574,
"signal/accgated_coverage_1/centered_abs_mean": 0.04150879755616188,
"signal/accgated_coverage_1/group_std_mean": 0.053798314183950424,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.060856424272060394,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004150879895314574,
"signal/accgated_coverage_10/centered_abs_mean": 0.04150879755616188,
"signal/accgated_coverage_10/group_std_mean": 0.053798314183950424,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.060856424272060394,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004150879895314574,
"signal/accgated_coverage_15/centered_abs_mean": 0.04148044250905514,
"signal/accgated_coverage_15/group_std_mean": 0.05376226231455803,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06081179901957512,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004148044297471643,
"signal/accgated_coverage_20/centered_abs_mean": 0.039842867851257326,
"signal/accgated_coverage_20/group_std_mean": 0.05164751708507538,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.0583540290594101,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.003984286775812506,
"signal/accgated_coverage_25/centered_abs_mean": 0.028048912808299065,
"signal/accgated_coverage_25/group_std_mean": 0.0365018580108881,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.04106144867837429,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.002804891183041036,
"signal/accgated_coverage_5/centered_abs_mean": 0.04150879755616188,
"signal/accgated_coverage_5/group_std_mean": 0.053798314183950424,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.060856424272060394,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004150879895314574,
"signal/accuracy_reward/centered_abs_mean": 0.10025634765625,
"signal/accuracy_reward/group_std_mean": 0.1399953082203865,
"signal/accuracy_reward/group_zero_std_frac": 0.565625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7302139401435852,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.050128173828125,
"signal/advantage_abs_mean": 0.7322210907936096,
"signal/advantage_pre_scale_abs_mean": 0.07584033757448197,
"signal/advantage_pre_scale_std": 0.13141691088676452,
"signal/advantage_std": 0.9830453395843506,
"signal/brier_reward/centered_abs_mean": 0.1180185854434967,
"signal/brier_reward/group_std_mean": 0.15800367891788483,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17282206416130066,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.011801858618855476,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.028158356621861457,
"signal/confidence_uniqueness_reward/group_std_mean": 0.052150913327932355,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.04124394431710243,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.002815835736691952,
"signal/format_reward/centered_abs_mean": 0.01771240234375,
"signal/format_reward/group_std_mean": 0.03994653038680553,
"signal/format_reward/group_zero_std_frac": 0.8125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.12896933853626252,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.008856201171875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2993794620037079,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.36894323825836184,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.44094533324241636,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.029937946796417238,
"step": 125
},
{
"calibration/aurc": 0.3025780177994948,
"calibration/batch_distribution_entropy": 0.979354441061426,
"calibration/batch_entropy_100bins": 0.9681388455166908,
"calibration/batch_entropy_10bins": 0.979354441061426,
"calibration/batch_entropy_50bins": 0.9770671903869058,
"calibration/batch_uniqueness": 0.953144327318293,
"calibration/buffer_distribution_entropy": 0.9949671337980046,
"calibration/buffer_entropy_100bins": 0.996895478382289,
"calibration/buffer_entropy_10bins": 0.9949671337980046,
"calibration/buffer_entropy_50bins": 0.9965874814340936,
"calibration/confidence_entropy": 0.4867560859157094,
"calibration/coverage@0%": 0.016510976495118217,
"calibration/coverage@1%": 0.016510976495118217,
"calibration/coverage@10%": 0.07892453854748714,
"calibration/coverage@15%": 0.13804742708664502,
"calibration/coverage@20%": 0.23070107105663032,
"calibration/coverage@25%": 0.3899725196732854,
"calibration/coverage@30%": 0.5141432511317298,
"calibration/coverage@5%": 0.02359921667081888,
"calibration/distribution_entropy_10": 0.979354441061426,
"calibration/distribution_entropy_100": 0.9681388455166908,
"calibration/ece": 0.0953810961460669,
"calibration/mean_confidence": 0.5386896697154562,
"calibration/unique_confidence_per_question": 0.985546875,
"calibration/unique_confidences": 504.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.009375,
"completions/max_length": 1468.2,
"completions/max_terminated_length": 1468.2,
"completions/mean_length": 294.3357421875,
"completions/mean_terminated_length": 297.1115356445313,
"completions/min_length": 0.0,
"completions/min_terminated_length": 84.6,
"epoch": 0.416,
"grad_norm": 0.0059529454447329044,
"learning_rate": 1e-06,
"loss": -0.0348,
"num_tokens": 437545813.0,
"reward": 0.9238641858100891,
"reward_std": 0.10712433755397796,
"rewards/accgated_coverage_0": 0.031001245602965355,
"rewards/accgated_coverage_1": 0.031001245602965355,
"rewards/accgated_coverage_10": 0.03100116513669491,
"rewards/accgated_coverage_15": 0.0309445146471262,
"rewards/accgated_coverage_20": 0.029189500026404858,
"rewards/accgated_coverage_25": 0.02252784688025713,
"rewards/accgated_coverage_5": 0.031001245602965355,
"rewards/accuracy_reward": 0.51220703125,
"rewards/brier_reward": 0.8075977206230164,
"rewards/confidence_uniqueness_reward": 0.9421306252479553,
"rewards/format_reward": 0.9904296875,
"rewards/frontier_entropy_batch_reward": -0.2309367150068283,
"signal/accgated_coverage_0/centered_abs_mean": 0.038562070578336716,
"signal/accgated_coverage_0/group_std_mean": 0.049818246811628344,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0587571382522583,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0038562072440981865,
"signal/accgated_coverage_1/centered_abs_mean": 0.038562070578336716,
"signal/accgated_coverage_1/group_std_mean": 0.049818246811628344,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0587571382522583,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0038562072440981865,
"signal/accgated_coverage_10/centered_abs_mean": 0.038561628013849256,
"signal/accgated_coverage_10/group_std_mean": 0.0498176708817482,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.058756455034017566,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003856162820011377,
"signal/accgated_coverage_15/centered_abs_mean": 0.03847671225667,
"signal/accgated_coverage_15/group_std_mean": 0.04971098676323891,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05862765684723854,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003847671300172806,
"signal/accgated_coverage_20/centered_abs_mean": 0.035805197432637215,
"signal/accgated_coverage_20/group_std_mean": 0.04634842798113823,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.05456043556332588,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0035805197432637213,
"signal/accgated_coverage_25/centered_abs_mean": 0.02487424500286579,
"signal/accgated_coverage_25/group_std_mean": 0.032435936480760576,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03789363726973534,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0024874245282262564,
"signal/accgated_coverage_5/centered_abs_mean": 0.038562070578336716,
"signal/accgated_coverage_5/group_std_mean": 0.049818246811628344,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.0587571382522583,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0038562072440981865,
"signal/accuracy_reward/centered_abs_mean": 0.099078369140625,
"signal/accuracy_reward/group_std_mean": 0.1307857885956764,
"signal/accuracy_reward/group_zero_std_frac": 0.61875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.750113558769226,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0495391845703125,
"signal/advantage_abs_mean": 0.739863121509552,
"signal/advantage_pre_scale_abs_mean": 0.075393907725811,
"signal/advantage_pre_scale_std": 0.13072074055671692,
"signal/advantage_std": 0.9829875588417053,
"signal/brier_reward/centered_abs_mean": 0.1089529275894165,
"signal/brier_reward/group_std_mean": 0.14743083715438843,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.16605533361434938,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010895292460918426,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.02723761834204197,
"signal/confidence_uniqueness_reward/group_std_mean": 0.05300683528184891,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.04148608818650246,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0027237618807703255,
"signal/format_reward/centered_abs_mean": 0.01790771484375,
"signal/format_reward/group_std_mean": 0.04216676875948906,
"signal/format_reward/group_zero_std_frac": 0.796875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.13594462722539902,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.008953857421875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28649061918258667,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3607529938220978,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.438063371181488,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028649061918258667,
"step": 130
},
{
"calibration/aurc": 0.20929642491548833,
"calibration/batch_distribution_entropy": 0.9666235659568363,
"calibration/batch_entropy_100bins": 0.9609003595346801,
"calibration/batch_entropy_10bins": 0.9666235659568363,
"calibration/batch_entropy_50bins": 0.9693170728795266,
"calibration/batch_uniqueness": 0.951039959386927,
"calibration/buffer_distribution_entropy": 0.9933516900987188,
"calibration/buffer_entropy_100bins": 0.9961196930898865,
"calibration/buffer_entropy_10bins": 0.9933516900987188,
"calibration/buffer_entropy_50bins": 0.9956336480526137,
"calibration/confidence_entropy": 0.45137468724276675,
"calibration/coverage@0%": 0.05803070975111001,
"calibration/coverage@1%": 0.06040225125308629,
"calibration/coverage@10%": 0.2621239390939286,
"calibration/coverage@15%": 0.3900838077965534,
"calibration/coverage@20%": 0.5375282357657016,
"calibration/coverage@25%": 0.6194343491631721,
"calibration/coverage@30%": 0.7080397314718072,
"calibration/coverage@5%": 0.18007394880617061,
"calibration/distribution_entropy_10": 0.9666235659568363,
"calibration/distribution_entropy_100": 0.9609003595346801,
"calibration/ece": 0.12470046467639159,
"calibration/mean_confidence": 0.506058883710254,
"calibration/unique_confidence_per_question": 0.9890625,
"calibration/unique_confidences": 506.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00517578125,
"completions/max_length": 1511.4,
"completions/max_terminated_length": 1511.4,
"completions/mean_length": 322.36943359375,
"completions/mean_terminated_length": 324.0555419921875,
"completions/min_length": 0.0,
"completions/min_terminated_length": 83.4,
"epoch": 0.432,
"grad_norm": 0.006484640296548605,
"learning_rate": 1e-06,
"loss": -0.014,
"num_tokens": 455861212.0,
"reward": 0.9520443677902222,
"reward_std": 0.08645348250865936,
"rewards/accgated_coverage_0": 0.038733229786157605,
"rewards/accgated_coverage_1": 0.038733229786157605,
"rewards/accgated_coverage_10": 0.038732708990573884,
"rewards/accgated_coverage_15": 0.038685943186283114,
"rewards/accgated_coverage_20": 0.03635015413165092,
"rewards/accgated_coverage_25": 0.026963535323739053,
"rewards/accgated_coverage_5": 0.038733229786157605,
"rewards/accuracy_reward": 0.5515625,
"rewards/brier_reward": 0.8254736661911011,
"rewards/confidence_uniqueness_reward": 0.945596206188202,
"rewards/format_reward": 0.99462890625,
"rewards/frontier_entropy_batch_reward": -0.23851539492607116,
"signal/accgated_coverage_0/centered_abs_mean": 0.044684822857379916,
"signal/accgated_coverage_0/group_std_mean": 0.058309643715620044,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08240354061126709,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004468482360243798,
"signal/accgated_coverage_1/centered_abs_mean": 0.044684822857379916,
"signal/accgated_coverage_1/group_std_mean": 0.058309643715620044,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08240354061126709,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004468482360243798,
"signal/accgated_coverage_10/centered_abs_mean": 0.04468408077955246,
"signal/accgated_coverage_10/group_std_mean": 0.05830869153141975,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08240217342972755,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004468408133834601,
"signal/accgated_coverage_15/centered_abs_mean": 0.04457318708300591,
"signal/accgated_coverage_15/group_std_mean": 0.05816451758146286,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.08219892829656601,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004457318875938654,
"signal/accgated_coverage_20/centered_abs_mean": 0.039794500917196274,
"signal/accgated_coverage_20/group_std_mean": 0.05199441537261009,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.07344191148877144,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0039794502779841425,
"signal/accgated_coverage_25/centered_abs_mean": 0.025528350099921226,
"signal/accgated_coverage_25/group_std_mean": 0.033343040570616725,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.04716198146343231,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0025528350844979284,
"signal/accgated_coverage_5/centered_abs_mean": 0.044684822857379916,
"signal/accgated_coverage_5/group_std_mean": 0.058309643715620044,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08240354061126709,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004468482360243798,
"signal/accuracy_reward/centered_abs_mean": 0.0884033203125,
"signal/accuracy_reward/group_std_mean": 0.11937278360128403,
"signal/accuracy_reward/group_zero_std_frac": 0.64375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.8137334942817688,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.04420166015625,
"signal/advantage_abs_mean": 0.7578792929649353,
"signal/advantage_pre_scale_abs_mean": 0.06263729482889176,
"signal/advantage_pre_scale_std": 0.10849435478448868,
"signal/advantage_std": 0.9826678991317749,
"signal/brier_reward/centered_abs_mean": 0.09573982656002045,
"signal/brier_reward/group_std_mean": 0.12826006710529328,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17727158665657045,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.00957398284226656,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.02143198624253273,
"signal/confidence_uniqueness_reward/group_std_mean": 0.0379902720451355,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.039538116008043286,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0021431986475363373,
"signal/format_reward/centered_abs_mean": 0.010125732421875,
"signal/format_reward/group_std_mean": 0.024614957720041276,
"signal/format_reward/group_zero_std_frac": 0.878125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0929967224597931,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0050628662109375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2839214622974396,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.35732280015945433,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5260580718517304,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02839214652776718,
"step": 135
},
{
"calibration/aurc": 0.24008029387452315,
"calibration/batch_distribution_entropy": 0.9582429928301884,
"calibration/batch_entropy_100bins": 0.9560383009047152,
"calibration/batch_entropy_10bins": 0.9582429928301884,
"calibration/batch_entropy_50bins": 0.9628248556163014,
"calibration/batch_uniqueness": 0.9487936878584364,
"calibration/buffer_distribution_entropy": 0.9915329987498313,
"calibration/buffer_entropy_100bins": 0.9952064630137597,
"calibration/buffer_entropy_10bins": 0.9915329987498313,
"calibration/buffer_entropy_50bins": 0.9945604913527623,
"calibration/confidence_entropy": 0.47640624400809406,
"calibration/coverage@0%": 0.02714710979633809,
"calibration/coverage@1%": 0.02714710979633809,
"calibration/coverage@10%": 0.11739827064813117,
"calibration/coverage@15%": 0.20916635049902266,
"calibration/coverage@20%": 0.393666620575441,
"calibration/coverage@25%": 0.5961379367342301,
"calibration/coverage@30%": 0.7325203249817396,
"calibration/coverage@5%": 0.07773824679421902,
"calibration/distribution_entropy_10": 0.9582429928301884,
"calibration/distribution_entropy_100": 0.9560383009047152,
"calibration/ece": 0.11184279833023521,
"calibration/mean_confidence": 0.5777567209005617,
"calibration/unique_confidence_per_question": 0.9921875,
"calibration/unique_confidences": 508.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00439453125,
"completions/max_length": 1483.4,
"completions/max_terminated_length": 1483.4,
"completions/mean_length": 300.00400390625,
"completions/mean_terminated_length": 301.3524597167969,
"completions/min_length": 0.0,
"completions/min_terminated_length": 82.8,
"epoch": 0.448,
"grad_norm": 0.006502500735223293,
"learning_rate": 1e-06,
"loss": -0.013,
"num_tokens": 473886053.0,
"reward": 0.93531174659729,
"reward_std": 0.08760718852281571,
"rewards/accgated_coverage_0": 0.03728027306497097,
"rewards/accgated_coverage_1": 0.03728027306497097,
"rewards/accgated_coverage_10": 0.03727986589074135,
"rewards/accgated_coverage_15": 0.03714171387255192,
"rewards/accgated_coverage_20": 0.03331144005060196,
"rewards/accgated_coverage_25": 0.022769278287887572,
"rewards/accgated_coverage_5": 0.03728027306497097,
"rewards/accuracy_reward": 0.523828125,
"rewards/brier_reward": 0.8183592557907104,
"rewards/confidence_uniqueness_reward": 0.946139645576477,
"rewards/format_reward": 0.9955078125,
"rewards/frontier_entropy_batch_reward": -0.25040421485900877,
"signal/accgated_coverage_0/centered_abs_mean": 0.0386271245777607,
"signal/accgated_coverage_0/group_std_mean": 0.05025699734687805,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06749763414263725,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0038627125322818756,
"signal/accgated_coverage_1/centered_abs_mean": 0.0386271245777607,
"signal/accgated_coverage_1/group_std_mean": 0.05025699734687805,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06749763414263725,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0038627125322818756,
"signal/accgated_coverage_10/centered_abs_mean": 0.03862656578421593,
"signal/accgated_coverage_10/group_std_mean": 0.05025627538561821,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06749668270349503,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0038626566529273987,
"signal/accgated_coverage_15/centered_abs_mean": 0.03826904520392418,
"signal/accgated_coverage_15/group_std_mean": 0.04980325549840927,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06686509773135185,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0038269047159701585,
"signal/accgated_coverage_20/centered_abs_mean": 0.031762827187776566,
"signal/accgated_coverage_20/group_std_mean": 0.041470900923013684,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.055422961711883545,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.003176282811909914,
"signal/accgated_coverage_25/centered_abs_mean": 0.019670241884887217,
"signal/accgated_coverage_25/group_std_mean": 0.02559336870908737,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03432374633848667,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0019670241978019475,
"signal/accgated_coverage_5/centered_abs_mean": 0.0386271245777607,
"signal/accgated_coverage_5/group_std_mean": 0.05025699734687805,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06749763414263725,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0038627125322818756,
"signal/accuracy_reward/centered_abs_mean": 0.08167724609375,
"signal/accuracy_reward/group_std_mean": 0.11164056062698365,
"signal/accuracy_reward/group_zero_std_frac": 0.6625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7124817728996277,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.040838623046875,
"signal/advantage_abs_mean": 0.7605426907539368,
"signal/advantage_pre_scale_abs_mean": 0.06395973339676857,
"signal/advantage_pre_scale_std": 0.10754972249269486,
"signal/advantage_std": 0.9827678203582764,
"signal/brier_reward/centered_abs_mean": 0.09692021906375885,
"signal/brier_reward/group_std_mean": 0.12895677536725997,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.16906891763210297,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009692022576928139,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.0202910952270031,
"signal/confidence_uniqueness_reward/group_std_mean": 0.03438720665872097,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03529815748333931,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0020291096065193416,
"signal/format_reward/centered_abs_mean": 0.008447265625,
"signal/format_reward/group_std_mean": 0.020038634724915027,
"signal/format_reward/group_zero_std_frac": 0.903125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.07201715260744095,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0042236328125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2951166212558746,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3680289924144745,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5179084181785584,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.029511663317680358,
"step": 140
},
{
"calibration/aurc": 0.3917029491229286,
"calibration/batch_distribution_entropy": 0.9642539465560755,
"calibration/batch_entropy_100bins": 0.9604995900446335,
"calibration/batch_entropy_10bins": 0.9642539465560755,
"calibration/batch_entropy_50bins": 0.9675689724428146,
"calibration/batch_uniqueness": 0.9502062758261453,
"calibration/buffer_distribution_entropy": 0.9909179299303247,
"calibration/buffer_entropy_100bins": 0.9949119515299027,
"calibration/buffer_entropy_10bins": 0.9909179299303247,
"calibration/buffer_entropy_50bins": 0.9942075736380198,
"calibration/confidence_entropy": 0.47909614412569435,
"calibration/coverage@0%": 0.004730661392515844,
"calibration/coverage@1%": 0.004730661392515844,
"calibration/coverage@10%": 0.004730661392515844,
"calibration/coverage@15%": 0.016137164693497786,
"calibration/coverage@20%": 0.05120725848846826,
"calibration/coverage@25%": 0.249298761484593,
"calibration/coverage@30%": 0.3222828686992237,
"calibration/coverage@5%": 0.004730661392515844,
"calibration/distribution_entropy_10": 0.9642539465560755,
"calibration/distribution_entropy_100": 0.9604995900446335,
"calibration/ece": 0.11787580651719938,
"calibration/mean_confidence": 0.47452307419640166,
"calibration/unique_confidence_per_question": 0.993359375,
"calibration/unique_confidences": 508.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00751953125,
"completions/max_length": 1505.0,
"completions/max_terminated_length": 1505.0,
"completions/mean_length": 268.03193359375,
"completions/mean_terminated_length": 270.0797424316406,
"completions/min_length": 0.0,
"completions/min_terminated_length": 80.8,
"epoch": 0.464,
"grad_norm": 0.02154521644115448,
"learning_rate": 1e-06,
"loss": -0.0345,
"num_tokens": 491801516.0,
"reward": 0.9017915368080139,
"reward_std": 0.09375011175870895,
"rewards/accgated_coverage_0": 0.03344345688819885,
"rewards/accgated_coverage_1": 0.03344345688819885,
"rewards/accgated_coverage_10": 0.033436056971549985,
"rewards/accgated_coverage_15": 0.03300386779010296,
"rewards/accgated_coverage_20": 0.02896622195839882,
"rewards/accgated_coverage_25": 0.0186858719214797,
"rewards/accgated_coverage_5": 0.03344345688819885,
"rewards/accuracy_reward": 0.46728515625,
"rewards/brier_reward": 0.7905396342277526,
"rewards/confidence_uniqueness_reward": 0.9437247395515442,
"rewards/format_reward": 0.99248046875,
"rewards/frontier_entropy_batch_reward": -0.22959967851638793,
"signal/accgated_coverage_0/centered_abs_mean": 0.03358997143805027,
"signal/accgated_coverage_0/group_std_mean": 0.04321310147643089,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06041149497032165,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0033589971717447042,
"signal/accgated_coverage_1/centered_abs_mean": 0.03358997143805027,
"signal/accgated_coverage_1/group_std_mean": 0.04321310147643089,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06041149497032165,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0033589971717447042,
"signal/accgated_coverage_10/centered_abs_mean": 0.03357952311635017,
"signal/accgated_coverage_10/group_std_mean": 0.04319960847496986,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06039271205663681,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003357952367514372,
"signal/accgated_coverage_15/centered_abs_mean": 0.03307505995035172,
"signal/accgated_coverage_15/group_std_mean": 0.04254415407776833,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.059490876644849776,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0033075060695409776,
"signal/accgated_coverage_20/centered_abs_mean": 0.02758634462952614,
"signal/accgated_coverage_20/group_std_mean": 0.035615530982613566,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.04958576261997223,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.002758634556084871,
"signal/accgated_coverage_25/centered_abs_mean": 0.01656769048422575,
"signal/accgated_coverage_25/group_std_mean": 0.02142725735902786,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.02984500527381897,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.001656769053079188,
"signal/accgated_coverage_5/centered_abs_mean": 0.03358997143805027,
"signal/accgated_coverage_5/group_std_mean": 0.04321310147643089,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06041149497032165,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0033589971717447042,
"signal/accuracy_reward/centered_abs_mean": 0.086663818359375,
"signal/accuracy_reward/group_std_mean": 0.11796283423900604,
"signal/accuracy_reward/group_zero_std_frac": 0.64375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7730632901191712,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0433319091796875,
"signal/advantage_abs_mean": 0.7327277064323425,
"signal/advantage_pre_scale_abs_mean": 0.06540881618857383,
"signal/advantage_pre_scale_std": 0.11837582737207412,
"signal/advantage_std": 0.9827141165733337,
"signal/brier_reward/centered_abs_mean": 0.10903198122978211,
"signal/brier_reward/group_std_mean": 0.14466988742351533,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19576336443424225,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010903198458254337,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.024211350455880166,
"signal/confidence_uniqueness_reward/group_std_mean": 0.04441752508282661,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.04333461448550224,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.002421135036274791,
"signal/format_reward/centered_abs_mean": 0.014019775390625,
"signal/format_reward/group_std_mean": 0.03256259262561798,
"signal/format_reward/group_zero_std_frac": 0.84375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.12432230561971665,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0070098876953125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.27115010023117064,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.34502353668212893,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.4899654030799866,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02711501009762287,
"step": 145
},
{
"calibration/aurc": 0.23882732228418374,
"calibration/batch_distribution_entropy": 0.9522994811617709,
"calibration/batch_entropy_100bins": 0.9549520672524052,
"calibration/batch_entropy_10bins": 0.9522994811617709,
"calibration/batch_entropy_50bins": 0.9608661896347657,
"calibration/batch_uniqueness": 0.947633446280596,
"calibration/buffer_distribution_entropy": 0.9903671098075589,
"calibration/buffer_entropy_100bins": 0.9946231764020024,
"calibration/buffer_entropy_10bins": 0.9903671098075589,
"calibration/buffer_entropy_50bins": 0.9938679155448907,
"calibration/confidence_entropy": 0.4420387559249096,
"calibration/coverage@0%": 0.01608321149214828,
"calibration/coverage@1%": 0.01608321149214828,
"calibration/coverage@10%": 0.3075598043932112,
"calibration/coverage@15%": 0.37025884404002973,
"calibration/coverage@20%": 0.468727859351924,
"calibration/coverage@25%": 0.5899137669598538,
"calibration/coverage@30%": 0.6943019461659274,
"calibration/coverage@5%": 0.024316203319016017,
"calibration/distribution_entropy_10": 0.9522994811617709,
"calibration/distribution_entropy_100": 0.9549520672524052,
"calibration/ece": 0.12366753272232198,
"calibration/mean_confidence": 0.4882041866925267,
"calibration/unique_confidence_per_question": 0.994140625,
"calibration/unique_confidences": 509.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00703125,
"completions/max_length": 1501.4,
"completions/max_terminated_length": 1501.4,
"completions/mean_length": 267.16162109375,
"completions/mean_terminated_length": 269.0512451171875,
"completions/min_length": 0.0,
"completions/min_terminated_length": 92.4,
"epoch": 0.48,
"grad_norm": 0.008620921522378922,
"learning_rate": 1e-06,
"loss": -0.0291,
"num_tokens": 509585283.0,
"reward": 0.9257105112075805,
"reward_std": 0.0899645060300827,
"rewards/accgated_coverage_0": 0.03870713487267494,
"rewards/accgated_coverage_1": 0.03870713487267494,
"rewards/accgated_coverage_10": 0.03870380818843842,
"rewards/accgated_coverage_15": 0.03801336362957954,
"rewards/accgated_coverage_20": 0.03206237107515335,
"rewards/accgated_coverage_25": 0.021657981723546983,
"rewards/accgated_coverage_5": 0.038706672936677934,
"rewards/accuracy_reward": 0.5154296875,
"rewards/brier_reward": 0.8048705458641052,
"rewards/confidence_uniqueness_reward": 0.9407513856887817,
"rewards/format_reward": 0.99287109375,
"rewards/frontier_entropy_batch_reward": -0.2765791267156601,
"signal/accgated_coverage_0/centered_abs_mean": 0.04596189185976982,
"signal/accgated_coverage_0/group_std_mean": 0.05914460569620132,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08689215183258056,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004596189223229885,
"signal/accgated_coverage_1/centered_abs_mean": 0.04596189185976982,
"signal/accgated_coverage_1/group_std_mean": 0.05914460569620132,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08689215183258056,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004596189223229885,
"signal/accgated_coverage_10/centered_abs_mean": 0.045957055687904355,
"signal/accgated_coverage_10/group_std_mean": 0.05913829058408737,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08688297122716904,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0045957055408507586,
"signal/accgated_coverage_15/centered_abs_mean": 0.04415743127465248,
"signal/accgated_coverage_15/group_std_mean": 0.05682581588625908,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.08346445932984352,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004415743285790086,
"signal/accgated_coverage_20/centered_abs_mean": 0.03405577689409256,
"signal/accgated_coverage_20/group_std_mean": 0.043708188086748125,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.06430306360125541,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0034055779222399,
"signal/accgated_coverage_25/centered_abs_mean": 0.019960780069231986,
"signal/accgated_coverage_25/group_std_mean": 0.025570009648799897,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03764847815036774,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.001996078062802553,
"signal/accgated_coverage_5/centered_abs_mean": 0.04596079960465431,
"signal/accgated_coverage_5/group_std_mean": 0.0591432049870491,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08689013719558716,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0045960799790918825,
"signal/accuracy_reward/centered_abs_mean": 0.101123046875,
"signal/accuracy_reward/group_std_mean": 0.13275287449359893,
"signal/accuracy_reward/group_zero_std_frac": 0.621875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9607232213020325,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0505615234375,
"signal/advantage_abs_mean": 0.7425110936164856,
"signal/advantage_pre_scale_abs_mean": 0.06363283097743988,
"signal/advantage_pre_scale_std": 0.11245516985654831,
"signal/advantage_std": 0.9826167941093444,
"signal/brier_reward/centered_abs_mean": 0.10832853019237518,
"signal/brier_reward/group_std_mean": 0.14394364058971404,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.20565287172794341,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010832853615283966,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.025344723463058473,
"signal/confidence_uniqueness_reward/group_std_mean": 0.0453621007502079,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.048137671500444415,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0025344724301248787,
"signal/format_reward/centered_abs_mean": 0.013360595703125,
"signal/format_reward/group_std_mean": 0.03160409592092037,
"signal/format_reward/group_zero_std_frac": 0.846875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.12691295444965361,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0066802978515625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.29284334778785703,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3625558912754059,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.556717038154602,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02928433530032635,
"step": 150
},
{
"epoch": 0.48,
"eval_completions/clipped_ratio": 0.001953125,
"eval_completions/max_length": 1247.25,
"eval_completions/max_terminated_length": 1247.25,
"eval_completions/mean_length": 268.54673767089844,
"eval_completions/mean_terminated_length": 269.0869369506836,
"eval_completions/min_length": 83.0,
"eval_completions/min_terminated_length": 111.25,
"eval_loss": 0.0,
"eval_num_tokens": 509585283.0,
"eval_reward": 0.7974581271409988,
"eval_reward_std": 0.22370443120598793,
"eval_rewards/accgated_coverage_0": 0.036959752440452576,
"eval_rewards/accgated_coverage_1": 0.036959752440452576,
"eval_rewards/accgated_coverage_10": 0.03695661947131157,
"eval_rewards/accgated_coverage_15": 0.03556866850703955,
"eval_rewards/accgated_coverage_20": 0.02814620081335306,
"eval_rewards/accgated_coverage_25": 0.015634644776582718,
"eval_rewards/accgated_coverage_5": 0.03695772774517536,
"eval_rewards/accuracy_reward": 0.41015625,
"eval_rewards/brier_reward": 0.8109452575445175,
"eval_rewards/confidence_uniqueness_reward": 0.8934838175773621,
"eval_rewards/format_reward": 0.998046875,
"eval_rewards/frontier_entropy_batch_reward": -0.998046875,
"eval_runtime": 49.8231,
"eval_samples_per_second": 10.036,
"eval_signal/accgated_coverage_0/centered_abs_mean": 0.06472484860569239,
"eval_signal/accgated_coverage_0/group_std_mean": 0.07637850567698479,
"eval_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.029184915591031313,
"eval_signal/accgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.006472484674304724,
"eval_signal/accgated_coverage_1/centered_abs_mean": 0.06472484860569239,
"eval_signal/accgated_coverage_1/group_std_mean": 0.07637850567698479,
"eval_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.029184915591031313,
"eval_signal/accgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.006472484674304724,
"eval_signal/accgated_coverage_10/centered_abs_mean": 0.06471595913171768,
"eval_signal/accgated_coverage_10/group_std_mean": 0.076368298381567,
"eval_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.029180883895605803,
"eval_signal/accgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.006471596076153219,
"eval_signal/accgated_coverage_15/centered_abs_mean": 0.06202875729650259,
"eval_signal/accgated_coverage_15/group_std_mean": 0.07324669323861599,
"eval_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.027967726346105337,
"eval_signal/accgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.006202876102179289,
"eval_signal/accgated_coverage_20/centered_abs_mean": 0.047151570208370686,
"eval_signal/accgated_coverage_20/group_std_mean": 0.056151374243199825,
"eval_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.021246890537440777,
"eval_signal/accgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.004715157090686262,
"eval_signal/accgated_coverage_25/centered_abs_mean": 0.022495068609714508,
"eval_signal/accgated_coverage_25/group_std_mean": 0.0276912827976048,
"eval_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.01014216081239283,
"eval_signal/accgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.002249506942462176,
"eval_signal/accgated_coverage_5/centered_abs_mean": 0.06471883039921522,
"eval_signal/accgated_coverage_5/group_std_mean": 0.07637158688157797,
"eval_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.02918218309059739,
"eval_signal/accgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.006471883156336844,
"eval_signal/accuracy_reward/centered_abs_mean": 0.471435546875,
"eval_signal/accuracy_reward/group_std_mean": 0.4930347874760628,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.0589872151613235,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.2357177734375,
"eval_signal/advantage_abs_mean": 0.9201153963804245,
"eval_signal/advantage_pre_scale_abs_mean": 0.2057698667049408,
"eval_signal/advantage_pre_scale_std": 0.22152478992938995,
"eval_signal/advantage_std": 0.9876745939254761,
"eval_signal/brier_reward/centered_abs_mean": 0.1955561675131321,
"eval_signal/brier_reward/group_std_mean": 0.2507976330816746,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.08785480260848999,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.019555617589503527,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.04666137509047985,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.06006394140422344,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0208881963044405,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004666137741878629,
"eval_signal/format_reward/centered_abs_mean": 0.0037841796875,
"eval_signal/format_reward/group_std_mean": 0.011048543266952038,
"eval_signal/format_reward/group_zero_std_frac": 0.9375,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.008115135133266449,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.00189208984375,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0037841796875,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.011048543266952038,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.9375,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.001623026910237968,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0003784179862122983,
"eval_steps_per_second": 0.08,
"step": 150
},
{
"epoch": 0.48,
"step": 150,
"train_probe_completions/clipped_ratio": 0.002155172413793094,
"train_probe_completions/max_length": 1190.0,
"train_probe_completions/max_terminated_length": 1190.0,
"train_probe_completions/mean_length": 259.09496307373047,
"train_probe_completions/mean_terminated_length": 259.7217025756836,
"train_probe_completions/min_length": 75.5,
"train_probe_completions/min_terminated_length": 102.5,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 509585283.0,
"train_probe_reward": 0.8882113993167877,
"train_probe_reward_std": 0.2017120160162449,
"train_probe_rewards/accgated_coverage_0": 0.03797450661659241,
"train_probe_rewards/accgated_coverage_1": 0.03797450661659241,
"train_probe_rewards/accgated_coverage_10": 0.03797140344977379,
"train_probe_rewards/accgated_coverage_15": 0.037906466983258724,
"train_probe_rewards/accgated_coverage_20": 0.03250941447913647,
"train_probe_rewards/accgated_coverage_25": 0.024213303346186876,
"train_probe_rewards/accgated_coverage_5": 0.03797234874218702,
"train_probe_rewards/accuracy_reward": 0.583984375,
"train_probe_rewards/brier_reward": 0.8307550251483917,
"train_probe_rewards/confidence_uniqueness_reward": 0.8927277624607086,
"train_probe_rewards/format_reward": 0.998046875,
"train_probe_rewards/frontier_entropy_batch_reward": -0.998046875,
"train_probe_runtime": 47.8525,
"train_probe_samples_per_second": 10.449,
"train_probe_signal/accgated_coverage_0/centered_abs_mean": 0.11417684331536293,
"train_probe_signal/accgated_coverage_0/group_std_mean": 0.14883660897612572,
"train_probe_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.05820417031645775,
"train_probe_signal/accgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.011417684610933065,
"train_probe_signal/accgated_coverage_1/centered_abs_mean": 0.11417684331536293,
"train_probe_signal/accgated_coverage_1/group_std_mean": 0.14883660897612572,
"train_probe_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.05820417031645775,
"train_probe_signal/accgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.011417684610933065,
"train_probe_signal/accgated_coverage_10/centered_abs_mean": 0.11416159942746162,
"train_probe_signal/accgated_coverage_10/group_std_mean": 0.1488186288625002,
"train_probe_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.058196413330733776,
"train_probe_signal/accgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.011416160501539707,
"train_probe_signal/accgated_coverage_15/centered_abs_mean": 0.10755505412817001,
"train_probe_signal/accgated_coverage_15/group_std_mean": 0.1405606660991907,
"train_probe_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05483449716120958,
"train_probe_signal/accgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.010755505179986358,
"train_probe_signal/accgated_coverage_20/centered_abs_mean": 0.07564054802060127,
"train_probe_signal/accgated_coverage_20/group_std_mean": 0.10105748660862446,
"train_probe_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.038530619349330664,
"train_probe_signal/accgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.007564054802060127,
"train_probe_signal/accgated_coverage_25/centered_abs_mean": 0.037670995108783245,
"train_probe_signal/accgated_coverage_25/group_std_mean": 0.049820009618997574,
"train_probe_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.019213061314076185,
"train_probe_signal/accgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0037670995807275176,
"train_probe_signal/accgated_coverage_5/centered_abs_mean": 0.11416670307517052,
"train_probe_signal/accgated_coverage_5/group_std_mean": 0.14882470294833183,
"train_probe_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.058199009858071804,
"train_probe_signal/accgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.011416670633479953,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.4678955078125,
"train_probe_signal/accuracy_reward/group_std_mean": 0.4908546209335327,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.1728087663650513,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.23394775390625,
"train_probe_signal/advantage_abs_mean": 0.9004213958978653,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.1819896325469017,
"train_probe_signal/advantage_pre_scale_std": 0.20005683600902557,
"train_probe_signal/advantage_std": 0.9876216948032379,
"train_probe_signal/brier_reward/centered_abs_mean": 0.17260217666625977,
"train_probe_signal/brier_reward/group_std_mean": 0.2330055944621563,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.08651809394359589,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.017260216875001788,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.04319123178720474,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.05528763961046934,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02162163658067584,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004319122992455959,
"train_probe_signal/format_reward/centered_abs_mean": 0.0037841796875,
"train_probe_signal/format_reward/group_std_mean": 0.011048543266952038,
"train_probe_signal/format_reward/group_zero_std_frac": 0.9375,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.008350024931132793,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.00189208984375,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0037841796875,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.011048543266952038,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.9375,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0016700050327926874,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0003784179862122983,
"train_probe_steps_per_second": 0.084
},
{
"calibration/aurc": 0.3420323303490845,
"calibration/batch_distribution_entropy": 0.953856268261801,
"calibration/batch_entropy_100bins": 0.9555289676131504,
"calibration/batch_entropy_10bins": 0.953856268261801,
"calibration/batch_entropy_50bins": 0.9637568598653241,
"calibration/batch_uniqueness": 0.9488766278319645,
"calibration/buffer_distribution_entropy": 0.9888314725440429,
"calibration/buffer_entropy_100bins": 0.9938779189301894,
"calibration/buffer_entropy_10bins": 0.9888314725440429,
"calibration/buffer_entropy_50bins": 0.9929675610242462,
"calibration/confidence_entropy": 0.44061714569681687,
"calibration/coverage@0%": 0.05644603046698131,
"calibration/coverage@1%": 0.06389701085913817,
"calibration/coverage@10%": 0.20849866772294073,
"calibration/coverage@15%": 0.23865875815526838,
"calibration/coverage@20%": 0.27314336049515264,
"calibration/coverage@25%": 0.3745824915488253,
"calibration/coverage@30%": 0.43732419329842276,
"calibration/coverage@5%": 0.14542189478531137,
"calibration/distribution_entropy_10": 0.953856268261801,
"calibration/distribution_entropy_100": 0.9555289676131504,
"calibration/ece": 0.16718015766331334,
"calibration/mean_confidence": 0.5286337853535039,
"calibration/unique_confidence_per_question": 0.996875,
"calibration/unique_confidences": 510.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.003515625,
"completions/max_length": 1479.2,
"completions/max_terminated_length": 1479.2,
"completions/mean_length": 254.3328125,
"completions/mean_terminated_length": 255.25278015136718,
"completions/min_length": 0.0,
"completions/min_terminated_length": 94.2,
"epoch": 0.496,
"grad_norm": 0.010015669278800488,
"learning_rate": 1e-06,
"loss": -0.0189,
"num_tokens": 527497491.0,
"reward": 0.9453690052032471,
"reward_std": 0.08472145050764084,
"rewards/accgated_coverage_0": 0.029031409695744515,
"rewards/accgated_coverage_1": 0.029031409695744515,
"rewards/accgated_coverage_10": 0.029028695449233054,
"rewards/accgated_coverage_15": 0.027484452724456786,
"rewards/accgated_coverage_20": 0.022288791462779046,
"rewards/accgated_coverage_25": 0.018118551559746267,
"rewards/accgated_coverage_5": 0.0290294848382473,
"rewards/accuracy_reward": 0.5580078125,
"rewards/brier_reward": 0.8064975500106811,
"rewards/confidence_uniqueness_reward": 0.9461641788482666,
"rewards/format_reward": 0.99638671875,
"rewards/frontier_entropy_batch_reward": -0.25495721101760865,
"signal/accgated_coverage_0/centered_abs_mean": 0.03917410671710968,
"signal/accgated_coverage_0/group_std_mean": 0.05181092619895935,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0706573635339737,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.003917410597205162,
"signal/accgated_coverage_1/centered_abs_mean": 0.03917410671710968,
"signal/accgated_coverage_1/group_std_mean": 0.05181092619895935,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0706573635339737,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.003917410597205162,
"signal/accgated_coverage_10/centered_abs_mean": 0.03917007222771644,
"signal/accgated_coverage_10/group_std_mean": 0.051805579662323,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07065014839172364,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003917007287964225,
"signal/accgated_coverage_15/centered_abs_mean": 0.037576368823647496,
"signal/accgated_coverage_15/group_std_mean": 0.04971002787351608,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06780253425240516,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0037576370406895874,
"signal/accgated_coverage_20/centered_abs_mean": 0.02778107412159443,
"signal/accgated_coverage_20/group_std_mean": 0.03678085245192051,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.05017025843262672,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.002778107486665249,
"signal/accgated_coverage_25/centered_abs_mean": 0.017353014089167118,
"signal/accgated_coverage_25/group_std_mean": 0.022508595883846284,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03130726628005505,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.001735301362350583,
"signal/accgated_coverage_5/centered_abs_mean": 0.03917149640619755,
"signal/accgated_coverage_5/group_std_mean": 0.05180745795369148,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.0706527054309845,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.003917149687185884,
"signal/accuracy_reward/centered_abs_mean": 0.0716552734375,
"signal/accuracy_reward/group_std_mean": 0.0997350975871086,
"signal/accuracy_reward/group_zero_std_frac": 0.69375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6486071467399597,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.03582763671875,
"signal/advantage_abs_mean": 0.7581081986427307,
"signal/advantage_pre_scale_abs_mean": 0.06170215085148811,
"signal/advantage_pre_scale_std": 0.1044373482465744,
"signal/advantage_std": 0.9827094554901123,
"signal/brier_reward/centered_abs_mean": 0.09506244212388992,
"signal/brier_reward/group_std_mean": 0.12750503271818162,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17145202159881592,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009506244026124478,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.019138817116618155,
"signal/confidence_uniqueness_reward/group_std_mean": 0.03300557024776936,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03471752479672432,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0019138817442581058,
"signal/format_reward/centered_abs_mean": 0.006890869140625,
"signal/format_reward/group_std_mean": 0.018271204456686972,
"signal/format_reward/group_zero_std_frac": 0.903125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.06173182912170887,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0034454345703125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28996633291244506,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3605659306049347,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5249950170516968,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028996633365750313,
"step": 155
},
{
"calibration/aurc": 0.24336511158588384,
"calibration/batch_distribution_entropy": 0.9741255238083613,
"calibration/batch_entropy_100bins": 0.9672183961460645,
"calibration/batch_entropy_10bins": 0.9741255238083613,
"calibration/batch_entropy_50bins": 0.9751571454574985,
"calibration/batch_uniqueness": 0.9524256199384771,
"calibration/buffer_distribution_entropy": 0.9884854106419265,
"calibration/buffer_entropy_100bins": 0.9937200263994566,
"calibration/buffer_entropy_10bins": 0.9884854106419265,
"calibration/buffer_entropy_50bins": 0.9927770596117339,
"calibration/confidence_entropy": 0.5050990715096699,
"calibration/coverage@0%": 0.015240490459882582,
"calibration/coverage@1%": 0.015240490459882582,
"calibration/coverage@10%": 0.296508072407045,
"calibration/coverage@15%": 0.43369083292563604,
"calibration/coverage@20%": 0.53027917074364,
"calibration/coverage@25%": 0.6354069838551859,
"calibration/coverage@30%": 0.6928494985322896,
"calibration/coverage@5%": 0.11211549045988259,
"calibration/distribution_entropy_10": 0.9741255238083613,
"calibration/distribution_entropy_100": 0.9672183961460645,
"calibration/ece": 0.12253826250374203,
"calibration/mean_confidence": 0.5244593642068389,
"calibration/unique_confidence_per_question": 0.998828125,
"calibration/unique_confidences": 511.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00068359375,
"completions/max_length": 1252.6,
"completions/max_terminated_length": 1252.6,
"completions/mean_length": 222.87978515625,
"completions/mean_terminated_length": 223.04269409179688,
"completions/min_length": 62.8,
"completions/min_terminated_length": 101.4,
"epoch": 0.512,
"grad_norm": 0.009343143552541733,
"learning_rate": 1e-06,
"loss": -0.0038,
"num_tokens": 544925444.0,
"reward": 0.948420524597168,
"reward_std": 0.07811777293682098,
"rewards/accgated_coverage_0": 0.031681787222623825,
"rewards/accgated_coverage_1": 0.031681787222623825,
"rewards/accgated_coverage_10": 0.03168613165616989,
"rewards/accgated_coverage_15": 0.03075209353119135,
"rewards/accgated_coverage_20": 0.02438853606581688,
"rewards/accgated_coverage_25": 0.01804626304656267,
"rewards/accgated_coverage_5": 0.03167982473969459,
"rewards/accuracy_reward": 0.5533203125,
"rewards/brier_reward": 0.8238914132118225,
"rewards/confidence_uniqueness_reward": 0.9494869828224182,
"rewards/format_reward": 0.99921875,
"rewards/frontier_entropy_batch_reward": -0.2517848014831543,
"signal/accgated_coverage_0/centered_abs_mean": 0.03790315687656402,
"signal/accgated_coverage_0/group_std_mean": 0.049896400421857834,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06948045268654823,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0037903156131505965,
"signal/accgated_coverage_1/centered_abs_mean": 0.03790315687656402,
"signal/accgated_coverage_1/group_std_mean": 0.049896400421857834,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06948045268654823,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0037903156131505965,
"signal/accgated_coverage_10/centered_abs_mean": 0.037872648984193805,
"signal/accgated_coverage_10/group_std_mean": 0.04985624402761459,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06942275539040565,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0037872649263590574,
"signal/accgated_coverage_15/centered_abs_mean": 0.035352344065904616,
"signal/accgated_coverage_15/group_std_mean": 0.04655801430344582,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06478032171726227,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0035352344159036873,
"signal/accgated_coverage_20/centered_abs_mean": 0.0254873551428318,
"signal/accgated_coverage_20/group_std_mean": 0.033710374310612676,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.04670252650976181,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.002548735588788986,
"signal/accgated_coverage_25/centered_abs_mean": 0.01565339956432581,
"signal/accgated_coverage_25/group_std_mean": 0.020219568535685538,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.028708093985915183,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0015653399517759682,
"signal/accgated_coverage_5/centered_abs_mean": 0.037899602204561234,
"signal/accgated_coverage_5/group_std_mean": 0.04989180192351341,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06947392150759697,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0037899603135883807,
"signal/accuracy_reward/centered_abs_mean": 0.07874755859375,
"signal/accuracy_reward/group_std_mean": 0.10795319825410843,
"signal/accuracy_reward/group_zero_std_frac": 0.675,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.717556095123291,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.039373779296875,
"signal/advantage_abs_mean": 0.7793641209602356,
"signal/advantage_pre_scale_abs_mean": 0.06046253964304924,
"signal/advantage_pre_scale_std": 0.09541922360658646,
"signal/advantage_std": 0.9826901078224182,
"signal/brier_reward/centered_abs_mean": 0.09056967496871948,
"signal/brier_reward/group_std_mean": 0.11818202584981918,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.16516581773757935,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009056967869400977,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.014374598115682601,
"signal/confidence_uniqueness_reward/group_std_mean": 0.020020603016018868,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02630560249090195,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014374598627910017,
"signal/format_reward/centered_abs_mean": 0.00150146484375,
"signal/format_reward/group_std_mean": 0.0040831089485436674,
"signal/format_reward/group_zero_std_frac": 0.978125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.013917620107531548,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000750732421875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.302550220489502,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.37201085686683655,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.552851402759552,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.030255021899938582,
"step": 160
},
{
"calibration/aurc": 0.20161084735568863,
"calibration/batch_distribution_entropy": 0.9800647239236904,
"calibration/batch_entropy_100bins": 0.967589528148735,
"calibration/batch_entropy_10bins": 0.9800647239236904,
"calibration/batch_entropy_50bins": 0.9757939371199746,
"calibration/batch_uniqueness": 0.9532220959217362,
"calibration/buffer_distribution_entropy": 0.9891358174136526,
"calibration/buffer_entropy_100bins": 0.9940499510075647,
"calibration/buffer_entropy_10bins": 0.9891358174136526,
"calibration/buffer_entropy_50bins": 0.9931725423071013,
"calibration/confidence_entropy": 0.4836396897314044,
"calibration/coverage@0%": 0.08956855430528375,
"calibration/coverage@1%": 0.11422608855185909,
"calibration/coverage@10%": 0.3523169948630137,
"calibration/coverage@15%": 0.44181675024461836,
"calibration/coverage@20%": 0.5270769630626223,
"calibration/coverage@25%": 0.6647955907534246,
"calibration/coverage@30%": 0.746929274706458,
"calibration/coverage@5%": 0.20727663282778863,
"calibration/distribution_entropy_10": 0.9800647239236904,
"calibration/distribution_entropy_100": 0.967589528148735,
"calibration/ece": 0.1378452124678064,
"calibration/mean_confidence": 0.4713559625503825,
"calibration/unique_confidence_per_question": 0.997265625,
"calibration/unique_confidences": 510.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00078125,
"completions/max_length": 1198.0,
"completions/max_terminated_length": 1198.0,
"completions/mean_length": 216.287109375,
"completions/mean_terminated_length": 216.45720520019532,
"completions/min_length": 16.0,
"completions/min_terminated_length": 89.0,
"epoch": 0.528,
"grad_norm": 0.2967407703399658,
"learning_rate": 1e-06,
"loss": 0.0028,
"num_tokens": 562169760.0,
"reward": 0.9424612522125244,
"reward_std": 0.07506856471300125,
"rewards/accgated_coverage_0": 0.038681335374712945,
"rewards/accgated_coverage_1": 0.038681335374712945,
"rewards/accgated_coverage_10": 0.038645824417471886,
"rewards/accgated_coverage_15": 0.03622375801205635,
"rewards/accgated_coverage_20": 0.027649413608014584,
"rewards/accgated_coverage_25": 0.019064302928745745,
"rewards/accgated_coverage_5": 0.03867810592055321,
"rewards/accuracy_reward": 0.5322265625,
"rewards/brier_reward": 0.8208467364311218,
"rewards/confidence_uniqueness_reward": 0.9486340403556823,
"rewards/format_reward": 0.99921875,
"rewards/frontier_entropy_batch_reward": -0.23971877098083497,
"signal/accgated_coverage_0/centered_abs_mean": 0.04236748963594437,
"signal/accgated_coverage_0/group_std_mean": 0.05560220032930374,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08717814385890961,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004236749000847339,
"signal/accgated_coverage_1/centered_abs_mean": 0.04236748963594437,
"signal/accgated_coverage_1/group_std_mean": 0.05560220032930374,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08717814385890961,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004236749000847339,
"signal/accgated_coverage_10/centered_abs_mean": 0.042280571162700654,
"signal/accgated_coverage_10/group_std_mean": 0.055491919815540316,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08699899464845658,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004228057293221355,
"signal/accgated_coverage_15/centered_abs_mean": 0.038330917805433275,
"signal/accgated_coverage_15/group_std_mean": 0.05043485909700394,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.07886097878217697,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0038330917246639728,
"signal/accgated_coverage_20/centered_abs_mean": 0.02628464326262474,
"signal/accgated_coverage_20/group_std_mean": 0.03487337827682495,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.0540402777493,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.002628464251756668,
"signal/accgated_coverage_25/centered_abs_mean": 0.015496164374053479,
"signal/accgated_coverage_25/group_std_mean": 0.020068111643195153,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03188837841153145,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.001549616502597928,
"signal/accgated_coverage_5/centered_abs_mean": 0.04236375838518143,
"signal/accgated_coverage_5/group_std_mean": 0.05559727922081947,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08717051595449447,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.00423637586645782,
"signal/accuracy_reward/centered_abs_mean": 0.08587646484375,
"signal/accuracy_reward/group_std_mean": 0.11777384281158447,
"signal/accuracy_reward/group_zero_std_frac": 0.65,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.87207111120224,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.042938232421875,
"signal/advantage_abs_mean": 0.7578014612197876,
"signal/advantage_pre_scale_abs_mean": 0.05674182251095772,
"signal/advantage_pre_scale_std": 0.09196056127548217,
"signal/advantage_std": 0.9824815988540649,
"signal/brier_reward/centered_abs_mean": 0.09073255360126495,
"signal/brier_reward/group_std_mean": 0.11953855156898499,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18493298590183258,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009073255769908427,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01484372466802597,
"signal/confidence_uniqueness_reward/group_std_mean": 0.020842796191573143,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.030377311632037163,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014843725133687259,
"signal/format_reward/centered_abs_mean": 0.001513671875,
"signal/format_reward/group_std_mean": 0.004419417260214687,
"signal/format_reward/group_zero_std_frac": 0.975,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.015100087225437164,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0007568359375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2723666787147522,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.34593590497970583,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.003125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5558390021324158,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02723666876554489,
"step": 165
},
{
"calibration/aurc": 0.2127012981546347,
"calibration/batch_distribution_entropy": 0.8756573466140865,
"calibration/batch_entropy_100bins": 0.917205109215071,
"calibration/batch_entropy_10bins": 0.8756573466140865,
"calibration/batch_entropy_50bins": 0.9156912434049069,
"calibration/batch_uniqueness": 0.9328559763846872,
"calibration/buffer_distribution_entropy": 0.988454367532832,
"calibration/buffer_entropy_100bins": 0.9937202994058179,
"calibration/buffer_entropy_10bins": 0.988454367532832,
"calibration/buffer_entropy_50bins": 0.9927845605107096,
"calibration/confidence_entropy": 0.3698845979442108,
"calibration/coverage@0%": 0.014069379892367905,
"calibration/coverage@1%": 0.014069379892367905,
"calibration/coverage@10%": 0.16329959637964775,
"calibration/coverage@15%": 0.33439563967710373,
"calibration/coverage@20%": 0.587691872553816,
"calibration/coverage@25%": 0.7267864787181997,
"calibration/coverage@30%": 0.7912602433953033,
"calibration/coverage@5%": 0.06681216364970646,
"calibration/distribution_entropy_10": 0.8756573466140865,
"calibration/distribution_entropy_100": 0.917205109215071,
"calibration/ece": 0.1055862900073389,
"calibration/mean_confidence": 0.57320109373254,
"calibration/unique_confidence_per_question": 0.997265625,
"calibration/unique_confidences": 510.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.000390625,
"completions/max_length": 998.0,
"completions/max_terminated_length": 998.0,
"completions/mean_length": 214.7228515625,
"completions/mean_terminated_length": 214.8075439453125,
"completions/min_length": 18.0,
"completions/min_terminated_length": 92.2,
"epoch": 0.544,
"grad_norm": 0.012686481699347496,
"learning_rate": 1e-06,
"loss": -0.0063,
"num_tokens": 579532106.0,
"reward": 0.950640869140625,
"reward_std": 0.07806460708379745,
"rewards/accgated_coverage_0": 0.030770195834338664,
"rewards/accgated_coverage_1": 0.030770195834338664,
"rewards/accgated_coverage_10": 0.030660450272262097,
"rewards/accgated_coverage_15": 0.028487682063132524,
"rewards/accgated_coverage_20": 0.023728859797120094,
"rewards/accgated_coverage_25": 0.02452710084617138,
"rewards/accgated_coverage_5": 0.030770359840244055,
"rewards/accuracy_reward": 0.58134765625,
"rewards/brier_reward": 0.8025527000427246,
"rewards/confidence_uniqueness_reward": 0.9405021071434021,
"rewards/format_reward": 0.999609375,
"rewards/frontier_entropy_batch_reward": -0.34114604592323305,
"signal/accgated_coverage_0/centered_abs_mean": 0.04897341579198837,
"signal/accgated_coverage_0/group_std_mean": 0.06379004120826721,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.10184406340122223,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004897341597825289,
"signal/accgated_coverage_1/centered_abs_mean": 0.04897341579198837,
"signal/accgated_coverage_1/group_std_mean": 0.06379004120826721,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.10184406340122223,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004897341597825289,
"signal/accgated_coverage_10/centered_abs_mean": 0.048729841411113736,
"signal/accgated_coverage_10/group_std_mean": 0.06347204670310021,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.10132041871547699,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004872984066605568,
"signal/accgated_coverage_15/centered_abs_mean": 0.043547382950782774,
"signal/accgated_coverage_15/group_std_mean": 0.056818865984678266,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.09032139182090759,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004354738490656018,
"signal/accgated_coverage_20/centered_abs_mean": 0.030738670006394388,
"signal/accgated_coverage_20/group_std_mean": 0.0399206280708313,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.06399924382567405,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0030738670378923417,
"signal/accgated_coverage_25/centered_abs_mean": 0.020795797929167747,
"signal/accgated_coverage_25/group_std_mean": 0.026206348463892935,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.04349189549684525,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0020795798394829036,
"signal/accgated_coverage_5/centered_abs_mean": 0.048969811946153644,
"signal/accgated_coverage_5/group_std_mean": 0.06378556117415428,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.10183656066656113,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004896981548517943,
"signal/accuracy_reward/centered_abs_mean": 0.093328857421875,
"signal/accuracy_reward/group_std_mean": 0.12614899277687072,
"signal/accuracy_reward/group_zero_std_frac": 0.625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9720237970352172,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0466644287109375,
"signal/advantage_abs_mean": 0.7704769253730774,
"signal/advantage_pre_scale_abs_mean": 0.059611483663320544,
"signal/advantage_pre_scale_std": 0.09445669502019882,
"signal/advantage_std": 0.9824301958084106,
"signal/brier_reward/centered_abs_mean": 0.10469038486480713,
"signal/brier_reward/group_std_mean": 0.1372752457857132,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.2182036131620407,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010469038598239421,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.019409450143575667,
"signal/confidence_uniqueness_reward/group_std_mean": 0.025501700490713118,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.04110885038971901,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001940944977104664,
"signal/format_reward/centered_abs_mean": 0.0007568359375,
"signal/format_reward/group_std_mean": 0.0022097086533904076,
"signal/format_reward/group_zero_std_frac": 0.9875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.007568595930933952,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.00037841796875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.31217177510261535,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3844032406806946,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.003125,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.6534318447113037,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.031217177212238312,
"step": 170
},
{
"calibration/aurc": 0.21616299076577192,
"calibration/batch_distribution_entropy": 0.9499892353191359,
"calibration/batch_entropy_100bins": 0.9528780554219113,
"calibration/batch_entropy_10bins": 0.9499892353191359,
"calibration/batch_entropy_50bins": 0.9601596215921351,
"calibration/batch_uniqueness": 0.9480644962847558,
"calibration/buffer_distribution_entropy": 0.9864459040025011,
"calibration/buffer_entropy_100bins": 0.9927299176719748,
"calibration/buffer_entropy_10bins": 0.9864459040025011,
"calibration/buffer_entropy_50bins": 0.9916101081105726,
"calibration/confidence_entropy": 0.4707825918320333,
"calibration/coverage@0%": 0.12826895013142242,
"calibration/coverage@1%": 0.23100332513142244,
"calibration/coverage@10%": 0.37683735994397755,
"calibration/coverage@15%": 0.42612423436840496,
"calibration/coverage@20%": 0.4781914666551552,
"calibration/coverage@25%": 0.6086977248091017,
"calibration/coverage@30%": 0.661576061931622,
"calibration/coverage@5%": 0.3326416148651241,
"calibration/distribution_entropy_10": 0.9499892353191359,
"calibration/distribution_entropy_100": 0.9528780554219113,
"calibration/ece": 0.11514462672916644,
"calibration/mean_confidence": 0.5047624252345548,
"calibration/unique_confidence_per_question": 0.997265625,
"calibration/unique_confidences": 510.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00068359375,
"completions/max_length": 960.8,
"completions/max_terminated_length": 960.8,
"completions/mean_length": 223.94697265625,
"completions/mean_terminated_length": 224.10359802246094,
"completions/min_length": 34.8,
"completions/min_terminated_length": 94.2,
"epoch": 0.56,
"grad_norm": 0.007545601110905409,
"learning_rate": 1e-06,
"loss": -0.0075,
"num_tokens": 596646731.0,
"reward": 0.9411831617355346,
"reward_std": 0.0767708495259285,
"rewards/accgated_coverage_0": 0.036232196539640424,
"rewards/accgated_coverage_1": 0.036232196539640424,
"rewards/accgated_coverage_10": 0.03612459674477577,
"rewards/accgated_coverage_15": 0.03392649590969086,
"rewards/accgated_coverage_20": 0.025726568326354025,
"rewards/accgated_coverage_25": 0.01997104436159134,
"rewards/accgated_coverage_5": 0.03623101748526096,
"rewards/accuracy_reward": 0.53486328125,
"rewards/brier_reward": 0.8268492817878723,
"rewards/confidence_uniqueness_reward": 0.9489155530929565,
"rewards/format_reward": 0.99931640625,
"rewards/frontier_entropy_batch_reward": -0.25927582681179046,
"signal/accgated_coverage_0/centered_abs_mean": 0.04069398865103722,
"signal/accgated_coverage_0/group_std_mean": 0.052661196887493135,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07653182074427604,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004069398902356624,
"signal/accgated_coverage_1/centered_abs_mean": 0.04069398865103722,
"signal/accgated_coverage_1/group_std_mean": 0.052661196887493135,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07653182074427604,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004069398902356624,
"signal/accgated_coverage_10/centered_abs_mean": 0.04049623943865299,
"signal/accgated_coverage_10/group_std_mean": 0.05240870788693428,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07615845426917076,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.00404962389729917,
"signal/accgated_coverage_15/centered_abs_mean": 0.03667354956269264,
"signal/accgated_coverage_15/group_std_mean": 0.047530582547187804,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06888345256447792,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003667354956269264,
"signal/accgated_coverage_20/centered_abs_mean": 0.025723686814308165,
"signal/accgated_coverage_20/group_std_mean": 0.03343113139271736,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.04828597232699394,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0025723687140271068,
"signal/accgated_coverage_25/centered_abs_mean": 0.015581303276121616,
"signal/accgated_coverage_25/group_std_mean": 0.0199411753565073,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.029290680214762686,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0015581303741782904,
"signal/accgated_coverage_5/centered_abs_mean": 0.04069224148988724,
"signal/accgated_coverage_5/group_std_mean": 0.0526589035987854,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07652856186032295,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0040692242328077555,
"signal/accuracy_reward/centered_abs_mean": 0.078973388671875,
"signal/accuracy_reward/group_std_mean": 0.11022768467664719,
"signal/accuracy_reward/group_zero_std_frac": 0.659375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7338344335556031,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0394866943359375,
"signal/advantage_abs_mean": 0.7597652792930603,
"signal/advantage_pre_scale_abs_mean": 0.05810219869017601,
"signal/advantage_pre_scale_std": 0.09265348017215728,
"signal/advantage_std": 0.9826419591903687,
"signal/brier_reward/centered_abs_mean": 0.09374683797359466,
"signal/brier_reward/group_std_mean": 0.12306535094976426,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17533251345157624,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.00937468409538269,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.014428452216088771,
"signal/confidence_uniqueness_reward/group_std_mean": 0.02015237547457218,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.027135471254587172,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014428452355787157,
"signal/format_reward/centered_abs_mean": 0.001324462890625,
"signal/format_reward/group_std_mean": 0.0038669900968670845,
"signal/format_reward/group_zero_std_frac": 0.978125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.012127935513854026,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0006622314453125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.29531155824661254,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3678881824016571,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5555107235908509,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02953115701675415,
"step": 175
},
{
"calibration/aurc": 0.24870554033073483,
"calibration/batch_distribution_entropy": 0.9564601874156471,
"calibration/batch_entropy_100bins": 0.9554654061443909,
"calibration/batch_entropy_10bins": 0.9564601874156471,
"calibration/batch_entropy_50bins": 0.9633422689417068,
"calibration/batch_uniqueness": 0.9486245733931928,
"calibration/buffer_distribution_entropy": 0.9876485060186908,
"calibration/buffer_entropy_100bins": 0.9933432140055386,
"calibration/buffer_entropy_10bins": 0.9876485060186908,
"calibration/buffer_entropy_50bins": 0.992327366462923,
"calibration/confidence_entropy": 0.49152514158059724,
"calibration/coverage@0%": 0.050390625,
"calibration/coverage@1%": 0.061328125,
"calibration/coverage@10%": 0.280078125,
"calibration/coverage@15%": 0.337109375,
"calibration/coverage@20%": 0.40547257216242655,
"calibration/coverage@25%": 0.5622630259295499,
"calibration/coverage@30%": 0.6650501467710372,
"calibration/coverage@5%": 0.160546875,
"calibration/distribution_entropy_10": 0.9564601874156471,
"calibration/distribution_entropy_100": 0.9554654061443909,
"calibration/ece": 0.10753129139544775,
"calibration/mean_confidence": 0.47240811602168564,
"calibration/unique_confidence_per_question": 0.996875,
"calibration/unique_confidences": 510.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.000390625,
"completions/max_length": 1143.6,
"completions/max_terminated_length": 1143.6,
"completions/mean_length": 238.07197265625,
"completions/mean_terminated_length": 238.16708679199218,
"completions/min_length": 59.6,
"completions/min_terminated_length": 94.4,
"epoch": 0.576,
"grad_norm": 0.029306583106517792,
"learning_rate": 1e-06,
"loss": -0.0042,
"num_tokens": 614271212.0,
"reward": 0.9423859000205994,
"reward_std": 0.07297937572002411,
"rewards/accgated_coverage_0": 0.03318969458341599,
"rewards/accgated_coverage_1": 0.03318969458341599,
"rewards/accgated_coverage_10": 0.033165304362773894,
"rewards/accgated_coverage_15": 0.03151619657874107,
"rewards/accgated_coverage_20": 0.022509010136127473,
"rewards/accgated_coverage_25": 0.017029831930994988,
"rewards/accgated_coverage_5": 0.03318969458341599,
"rewards/accuracy_reward": 0.53720703125,
"rewards/brier_reward": 0.8193418741226196,
"rewards/confidence_uniqueness_reward": 0.9501522421836853,
"rewards/format_reward": 0.999609375,
"rewards/frontier_entropy_batch_reward": -0.23350661993026733,
"signal/accgated_coverage_0/centered_abs_mean": 0.038569864630699155,
"signal/accgated_coverage_0/group_std_mean": 0.050657518208026886,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07474226653575897,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0038569864351302384,
"signal/accgated_coverage_1/centered_abs_mean": 0.038569864630699155,
"signal/accgated_coverage_1/group_std_mean": 0.050657518208026886,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07474226653575897,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0038569864351302384,
"signal/accgated_coverage_10/centered_abs_mean": 0.038270081579685214,
"signal/accgated_coverage_10/group_std_mean": 0.0502681314945221,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.0741748906672001,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0038270081859081984,
"signal/accgated_coverage_15/centered_abs_mean": 0.033872129023075105,
"signal/accgated_coverage_15/group_std_mean": 0.044599755108356474,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06561982929706574,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003387212986126542,
"signal/accgated_coverage_20/centered_abs_mean": 0.02085312847048044,
"signal/accgated_coverage_20/group_std_mean": 0.027447084337472914,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.04035593904554844,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0020853129215538503,
"signal/accgated_coverage_25/centered_abs_mean": 0.013619895465672016,
"signal/accgated_coverage_25/group_std_mean": 0.017637683637440206,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.026398248597979547,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0013619895791634918,
"signal/accgated_coverage_5/centered_abs_mean": 0.038569864630699155,
"signal/accgated_coverage_5/group_std_mean": 0.050657518208026886,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07474226653575897,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0038569864351302384,
"signal/accuracy_reward/centered_abs_mean": 0.073394775390625,
"signal/accuracy_reward/group_std_mean": 0.10116819739341736,
"signal/accuracy_reward/group_zero_std_frac": 0.69375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7273806869983673,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0366973876953125,
"signal/advantage_abs_mean": 0.7706441640853882,
"signal/advantage_pre_scale_abs_mean": 0.05648205280303955,
"signal/advantage_pre_scale_std": 0.08870679885149002,
"signal/advantage_std": 0.982547378540039,
"signal/brier_reward/centered_abs_mean": 0.08554486632347107,
"signal/brier_reward/group_std_mean": 0.1119070053100586,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1688222885131836,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008554487023502588,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.013720016553997993,
"signal/confidence_uniqueness_reward/group_std_mean": 0.018226930871605873,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.027218982949852942,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001372001669369638,
"signal/format_reward/centered_abs_mean": 0.0007568359375,
"signal/format_reward/group_std_mean": 0.0022097086533904076,
"signal/format_reward/group_zero_std_frac": 0.9875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.007409945875406265,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.00037841796875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28034502267837524,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3513504147529602,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5526053249835968,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028034502640366554,
"step": 180
},
{
"calibration/aurc": 0.26805516402016777,
"calibration/batch_distribution_entropy": 0.96327666834478,
"calibration/batch_entropy_100bins": 0.9604539043833341,
"calibration/batch_entropy_10bins": 0.96327666834478,
"calibration/batch_entropy_50bins": 0.9674448570395444,
"calibration/batch_uniqueness": 0.9498058592527059,
"calibration/buffer_distribution_entropy": 0.9877242838500356,
"calibration/buffer_entropy_100bins": 0.9933788809639716,
"calibration/buffer_entropy_10bins": 0.9877242838500356,
"calibration/buffer_entropy_50bins": 0.9923705422872408,
"calibration/confidence_entropy": 0.4465089418818532,
"calibration/coverage@0%": 0.03755886130136986,
"calibration/coverage@1%": 0.052823049168297456,
"calibration/coverage@10%": 0.28158329256360076,
"calibration/coverage@15%": 0.32186964897260273,
"calibration/coverage@20%": 0.5048694349315068,
"calibration/coverage@25%": 0.5748692820450099,
"calibration/coverage@30%": 0.6385900807240704,
"calibration/coverage@5%": 0.2170506665851272,
"calibration/distribution_entropy_10": 0.96327666834478,
"calibration/distribution_entropy_100": 0.9604539043833341,
"calibration/ece": 0.13900513626529432,
"calibration/mean_confidence": 0.5118142436275825,
"calibration/unique_confidence_per_question": 0.994140625,
"calibration/unique_confidences": 509.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00068359375,
"completions/max_length": 1146.6,
"completions/max_terminated_length": 1146.6,
"completions/mean_length": 257.3923828125,
"completions/mean_terminated_length": 257.570166015625,
"completions/min_length": 22.4,
"completions/min_terminated_length": 105.4,
"epoch": 0.592,
"grad_norm": 0.010327347554266453,
"learning_rate": 1e-06,
"loss": -0.0026,
"num_tokens": 632074622.0,
"reward": 0.9404775619506835,
"reward_std": 0.07204383015632629,
"rewards/accgated_coverage_0": 0.03652774021029472,
"rewards/accgated_coverage_1": 0.03652774021029472,
"rewards/accgated_coverage_10": 0.036451513320207594,
"rewards/accgated_coverage_15": 0.033471403270959856,
"rewards/accgated_coverage_20": 0.023897471278905867,
"rewards/accgated_coverage_25": 0.020095057785511017,
"rewards/accgated_coverage_5": 0.03652774021029472,
"rewards/accuracy_reward": 0.53427734375,
"rewards/brier_reward": 0.8124344825744629,
"rewards/confidence_uniqueness_reward": 0.9494822144508361,
"rewards/format_reward": 0.99921875,
"rewards/frontier_entropy_batch_reward": -0.2481201171875,
"signal/accgated_coverage_0/centered_abs_mean": 0.037816737592220304,
"signal/accgated_coverage_0/group_std_mean": 0.05005468130111694,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07453058958053589,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.003781673824414611,
"signal/accgated_coverage_1/centered_abs_mean": 0.037816737592220304,
"signal/accgated_coverage_1/group_std_mean": 0.05005468130111694,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07453058958053589,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.003781673824414611,
"signal/accgated_coverage_10/centered_abs_mean": 0.037637120485305785,
"signal/accgated_coverage_10/group_std_mean": 0.04981823265552521,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07417668029665947,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003763712057843804,
"signal/accgated_coverage_15/centered_abs_mean": 0.033142223954200745,
"signal/accgated_coverage_15/group_std_mean": 0.04393702447414398,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06528697684407234,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0033142223954200746,
"signal/accgated_coverage_20/centered_abs_mean": 0.020543180406093597,
"signal/accgated_coverage_20/group_std_mean": 0.02713068202137947,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.040526320785284044,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0020543180871754885,
"signal/accgated_coverage_25/centered_abs_mean": 0.014620135352015496,
"signal/accgated_coverage_25/group_std_mean": 0.018793880194425582,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.029032538086175917,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0014620135305449366,
"signal/accgated_coverage_5/centered_abs_mean": 0.037816737592220304,
"signal/accgated_coverage_5/group_std_mean": 0.05005468130111694,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07453058958053589,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.003781673824414611,
"signal/accuracy_reward/centered_abs_mean": 0.065631103515625,
"signal/accuracy_reward/group_std_mean": 0.09536090344190598,
"signal/accuracy_reward/group_zero_std_frac": 0.69375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6428999900817871,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0328155517578125,
"signal/advantage_abs_mean": 0.769312858581543,
"signal/advantage_pre_scale_abs_mean": 0.05415288209915161,
"signal/advantage_pre_scale_std": 0.08658980876207352,
"signal/advantage_std": 0.9825261235237122,
"signal/brier_reward/centered_abs_mean": 0.09160646945238113,
"signal/brier_reward/group_std_mean": 0.12202910035848617,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1822609782218933,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009160646796226501,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.014031932502985001,
"signal/confidence_uniqueness_reward/group_std_mean": 0.019931669533252715,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02816827893257141,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014031932922080159,
"signal/format_reward/centered_abs_mean": 0.001513671875,
"signal/format_reward/group_std_mean": 0.004419417306780815,
"signal/format_reward/group_zero_std_frac": 0.975,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.015054125711321831,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0007568359375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2796932965517044,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3528794288635254,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.561053317785263,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027969330549240112,
"step": 185
},
{
"calibration/aurc": 0.22854390079993597,
"calibration/batch_distribution_entropy": 0.9342445936984319,
"calibration/batch_entropy_100bins": 0.9443006583452396,
"calibration/batch_entropy_10bins": 0.9342445936984319,
"calibration/batch_entropy_50bins": 0.9486951461901765,
"calibration/batch_uniqueness": 0.9434102059227779,
"calibration/buffer_distribution_entropy": 0.9871109769194184,
"calibration/buffer_entropy_100bins": 0.9930631463222463,
"calibration/buffer_entropy_10bins": 0.9871109769194184,
"calibration/buffer_entropy_50bins": 0.9919856453547234,
"calibration/confidence_entropy": 0.43308417544458766,
"calibration/coverage@0%": 0.09429750486833859,
"calibration/coverage@1%": 0.11695757703076519,
"calibration/coverage@10%": 0.3041011652654374,
"calibration/coverage@15%": 0.36283683989077237,
"calibration/coverage@20%": 0.5412957271798431,
"calibration/coverage@25%": 0.6133638868036786,
"calibration/coverage@30%": 0.6999827139137405,
"calibration/coverage@5%": 0.23044426262932577,
"calibration/distribution_entropy_10": 0.9342445936984319,
"calibration/distribution_entropy_100": 0.9443006583452396,
"calibration/ece": 0.10073084611089396,
"calibration/mean_confidence": 0.4644423418707818,
"calibration/unique_confidence_per_question": 0.996484375,
"calibration/unique_confidences": 510.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00107421875,
"completions/max_length": 1095.2,
"completions/max_terminated_length": 1095.2,
"completions/mean_length": 264.25986328125,
"completions/mean_terminated_length": 264.5446441650391,
"completions/min_length": 0.0,
"completions/min_terminated_length": 98.6,
"epoch": 0.608,
"grad_norm": 0.0073462496511638165,
"learning_rate": 1e-06,
"loss": -0.0079,
"num_tokens": 649780131.0,
"reward": 0.9384020328521728,
"reward_std": 0.07450624108314514,
"rewards/accgated_coverage_0": 0.04424070194363594,
"rewards/accgated_coverage_1": 0.04424070194363594,
"rewards/accgated_coverage_10": 0.04398571848869324,
"rewards/accgated_coverage_15": 0.04071062356233597,
"rewards/accgated_coverage_20": 0.02770284190773964,
"rewards/accgated_coverage_25": 0.022501731663942336,
"rewards/accgated_coverage_5": 0.04424070194363594,
"rewards/accuracy_reward": 0.528125,
"rewards/brier_reward": 0.8340995788574219,
"rewards/confidence_uniqueness_reward": 0.9445814251899719,
"rewards/format_reward": 0.998828125,
"rewards/frontier_entropy_batch_reward": -0.29704912900924685,
"signal/accgated_coverage_0/centered_abs_mean": 0.04013014137744904,
"signal/accgated_coverage_0/group_std_mean": 0.052305794507265094,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08512404710054397,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004013014072552323,
"signal/accgated_coverage_1/centered_abs_mean": 0.04013014137744904,
"signal/accgated_coverage_1/group_std_mean": 0.052305794507265094,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08512404710054397,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004013014072552323,
"signal/accgated_coverage_10/centered_abs_mean": 0.0398762971162796,
"signal/accgated_coverage_10/group_std_mean": 0.051977039128541944,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08457908630371094,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.00398762971162796,
"signal/accgated_coverage_15/centered_abs_mean": 0.03536115810275078,
"signal/accgated_coverage_15/group_std_mean": 0.04630323871970177,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.07503833025693893,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003536115912720561,
"signal/accgated_coverage_20/centered_abs_mean": 0.021407318487763404,
"signal/accgated_coverage_20/group_std_mean": 0.02791901268064976,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.04545978829264641,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0021407319232821464,
"signal/accgated_coverage_25/centered_abs_mean": 0.0154871279373765,
"signal/accgated_coverage_25/group_std_mean": 0.01972283162176609,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.032853809744119646,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.001548712863586843,
"signal/accgated_coverage_5/centered_abs_mean": 0.04013014137744904,
"signal/accgated_coverage_5/group_std_mean": 0.052305794507265094,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08512404710054397,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004013014072552323,
"signal/accuracy_reward/centered_abs_mean": 0.0779296875,
"signal/accuracy_reward/group_std_mean": 0.10503357201814652,
"signal/accuracy_reward/group_zero_std_frac": 0.69375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.8200716495513916,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.03896484375,
"signal/advantage_abs_mean": 0.7692615270614624,
"signal/advantage_pre_scale_abs_mean": 0.05708964392542839,
"signal/advantage_pre_scale_std": 0.09408938735723496,
"signal/advantage_std": 0.9823990583419799,
"signal/brier_reward/centered_abs_mean": 0.09090130180120468,
"signal/brier_reward/group_std_mean": 0.11966802328824996,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19230906069278716,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009090129844844341,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01815231665968895,
"signal/confidence_uniqueness_reward/group_std_mean": 0.025292839109897613,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03859374597668648,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0018152316566556691,
"signal/format_reward/centered_abs_mean": 0.00224609375,
"signal/format_reward/group_std_mean": 0.0059565090574324134,
"signal/format_reward/group_zero_std_frac": 0.96875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.02406892254948616,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.001123046875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.3105348527431488,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3823462545871735,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.6582759857177735,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.031053485348820686,
"step": 190
},
{
"calibration/aurc": 0.18840231553484604,
"calibration/batch_distribution_entropy": 0.9705119601228167,
"calibration/batch_entropy_100bins": 0.9647742565852673,
"calibration/batch_entropy_10bins": 0.9705119601228167,
"calibration/batch_entropy_50bins": 0.9719550324649265,
"calibration/batch_uniqueness": 0.951206409857362,
"calibration/buffer_distribution_entropy": 0.9871774413269341,
"calibration/buffer_entropy_100bins": 0.9931078939496111,
"calibration/buffer_entropy_10bins": 0.9871774413269341,
"calibration/buffer_entropy_50bins": 0.9920370349760832,
"calibration/confidence_entropy": 0.4626274673323495,
"calibration/coverage@0%": 0.09730384662426614,
"calibration/coverage@1%": 0.11253822162426615,
"calibration/coverage@10%": 0.38591380259295505,
"calibration/coverage@15%": 0.49302378913894324,
"calibration/coverage@20%": 0.5634005014677104,
"calibration/coverage@25%": 0.6341563723091976,
"calibration/coverage@30%": 0.7252545560176126,
"calibration/coverage@5%": 0.23495443982387476,
"calibration/distribution_entropy_10": 0.9705119601228167,
"calibration/distribution_entropy_100": 0.9647742565852673,
"calibration/ece": 0.1295659536901283,
"calibration/mean_confidence": 0.49497178734278596,
"calibration/unique_confidence_per_question": 0.996875,
"calibration/unique_confidences": 510.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00205078125,
"completions/max_length": 1279.4,
"completions/max_terminated_length": 1279.4,
"completions/mean_length": 275.11123046875,
"completions/mean_terminated_length": 275.6785827636719,
"completions/min_length": 0.0,
"completions/min_terminated_length": 94.6,
"epoch": 0.624,
"grad_norm": 0.006845767609775066,
"learning_rate": 1e-06,
"loss": -0.0045,
"num_tokens": 667941174.0,
"reward": 0.9511773705482482,
"reward_std": 0.07531674206256866,
"rewards/accgated_coverage_0": 0.0394110344350338,
"rewards/accgated_coverage_1": 0.0394110344350338,
"rewards/accgated_coverage_10": 0.039174669235944745,
"rewards/accgated_coverage_15": 0.035135726630687716,
"rewards/accgated_coverage_20": 0.023504067212343216,
"rewards/accgated_coverage_25": 0.021448821201920508,
"rewards/accgated_coverage_5": 0.0394110344350338,
"rewards/accuracy_reward": 0.549609375,
"rewards/brier_reward": 0.8249976396560669,
"rewards/confidence_uniqueness_reward": 0.948533046245575,
"rewards/format_reward": 0.9978515625,
"rewards/frontier_entropy_batch_reward": -0.2365578830242157,
"signal/accgated_coverage_0/centered_abs_mean": 0.04684974066913128,
"signal/accgated_coverage_0/group_std_mean": 0.05947878286242485,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.09336906671524048,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004684974253177643,
"signal/accgated_coverage_1/centered_abs_mean": 0.04684974066913128,
"signal/accgated_coverage_1/group_std_mean": 0.05947878286242485,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.09336906671524048,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004684974253177643,
"signal/accgated_coverage_10/centered_abs_mean": 0.04639555364847183,
"signal/accgated_coverage_10/group_std_mean": 0.05891382470726967,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.0924602597951889,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004639555606991053,
"signal/accgated_coverage_15/centered_abs_mean": 0.040813561901450156,
"signal/accgated_coverage_15/group_std_mean": 0.05178205445408821,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.08130657076835632,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.004081356199458241,
"signal/accgated_coverage_20/centered_abs_mean": 0.02181070726364851,
"signal/accgated_coverage_20/group_std_mean": 0.02770060859620571,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.04348419830203056,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0021810708567500114,
"signal/accgated_coverage_25/centered_abs_mean": 0.015771867148578166,
"signal/accgated_coverage_25/group_std_mean": 0.02001376114785671,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03133785650134087,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0015771867241710425,
"signal/accgated_coverage_5/centered_abs_mean": 0.04684974066913128,
"signal/accgated_coverage_5/group_std_mean": 0.05947878286242485,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.09336906671524048,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004684974253177643,
"signal/accuracy_reward/centered_abs_mean": 0.08671875,
"signal/accuracy_reward/group_std_mean": 0.11369689106941223,
"signal/accuracy_reward/group_zero_std_frac": 0.675,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.8666375637054443,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.043359375,
"signal/advantage_abs_mean": 0.76714586019516,
"signal/advantage_pre_scale_abs_mean": 0.05644790306687355,
"signal/advantage_pre_scale_std": 0.09328747689723968,
"signal/advantage_std": 0.9825296521186828,
"signal/brier_reward/centered_abs_mean": 0.09611079096794128,
"signal/brier_reward/group_std_mean": 0.12497009932994843,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19165875315666198,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009611079469323158,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.015913079679012298,
"signal/confidence_uniqueness_reward/group_std_mean": 0.02492523044347763,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.031683918088674545,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001591307995840907,
"signal/format_reward/centered_abs_mean": 0.00406494140625,
"signal/format_reward/group_std_mean": 0.01032110583037138,
"signal/format_reward/group_zero_std_frac": 0.946875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.040427832305431365,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.002032470703125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2693305671215057,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.34054588079452514,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5360336899757385,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026933058723807336,
"step": 195
},
{
"calibration/aurc": 0.2340914682252441,
"calibration/batch_distribution_entropy": 0.9646448354613091,
"calibration/batch_entropy_100bins": 0.962165901556174,
"calibration/batch_entropy_10bins": 0.9646448354613091,
"calibration/batch_entropy_50bins": 0.9691434203607667,
"calibration/batch_uniqueness": 0.9501688893282754,
"calibration/buffer_distribution_entropy": 0.9872129038417563,
"calibration/buffer_entropy_100bins": 0.9931008577857959,
"calibration/buffer_entropy_10bins": 0.9872129038417563,
"calibration/buffer_entropy_50bins": 0.9920397890667705,
"calibration/confidence_entropy": 0.48001664005114114,
"calibration/coverage@0%": 0.014468437631902076,
"calibration/coverage@1%": 0.014468437631902076,
"calibration/coverage@10%": 0.3223428566632132,
"calibration/coverage@15%": 0.37549002520528757,
"calibration/coverage@20%": 0.43458157663750435,
"calibration/coverage@25%": 0.5995355038419478,
"calibration/coverage@30%": 0.7253013662656843,
"calibration/coverage@5%": 0.2547242167414911,
"calibration/distribution_entropy_10": 0.9646448354613091,
"calibration/distribution_entropy_100": 0.962165901556174,
"calibration/ece": 0.1429748956807023,
"calibration/mean_confidence": 0.5688173059771868,
"calibration/unique_confidence_per_question": 0.996875,
"calibration/unique_confidences": 510.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00185546875,
"completions/max_length": 1434.6,
"completions/max_terminated_length": 1434.6,
"completions/mean_length": 270.08671875,
"completions/mean_terminated_length": 270.585498046875,
"completions/min_length": 0.0,
"completions/min_terminated_length": 91.0,
"epoch": 0.64,
"grad_norm": 0.009514262899756432,
"learning_rate": 1e-06,
"loss": -0.0075,
"num_tokens": 686049550.0,
"reward": 0.9561322808265686,
"reward_std": 0.07826831936836243,
"rewards/accgated_coverage_0": 0.03302889429032803,
"rewards/accgated_coverage_1": 0.03302889429032803,
"rewards/accgated_coverage_10": 0.03289430364966393,
"rewards/accgated_coverage_15": 0.029243571683764458,
"rewards/accgated_coverage_20": 0.02063581645488739,
"rewards/accgated_coverage_25": 0.024958676472306252,
"rewards/accgated_coverage_5": 0.03302889429032803,
"rewards/accuracy_reward": 0.57392578125,
"rewards/brier_reward": 0.8312363505363465,
"rewards/confidence_uniqueness_reward": 0.9454058289527894,
"rewards/format_reward": 0.9978515625,
"rewards/frontier_entropy_batch_reward": -0.2810252606868744,
"signal/accgated_coverage_0/centered_abs_mean": 0.035560768097639084,
"signal/accgated_coverage_0/group_std_mean": 0.04652189090847969,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06885926574468612,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.003556076902896166,
"signal/accgated_coverage_1/centered_abs_mean": 0.035560768097639084,
"signal/accgated_coverage_1/group_std_mean": 0.04652189090847969,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06885926574468612,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.003556076902896166,
"signal/accgated_coverage_10/centered_abs_mean": 0.03517423123121262,
"signal/accgated_coverage_10/group_std_mean": 0.046020156145095824,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06810760721564293,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0035174232441931965,
"signal/accgated_coverage_15/centered_abs_mean": 0.029718888550996782,
"signal/accgated_coverage_15/group_std_mean": 0.03901107758283615,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.057539229094982144,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.002971888892352581,
"signal/accgated_coverage_20/centered_abs_mean": 0.016794146597385408,
"signal/accgated_coverage_20/group_std_mean": 0.021738207340240477,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03249258175492287,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0016794146737083792,
"signal/accgated_coverage_25/centered_abs_mean": 0.01692484375089407,
"signal/accgated_coverage_25/group_std_mean": 0.02153026219457388,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03264769427478313,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0016924844589084388,
"signal/accgated_coverage_5/centered_abs_mean": 0.035560768097639084,
"signal/accgated_coverage_5/group_std_mean": 0.04652189090847969,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06885926574468612,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.003556076902896166,
"signal/accuracy_reward/centered_abs_mean": 0.068438720703125,
"signal/accuracy_reward/group_std_mean": 0.09378217607736587,
"signal/accuracy_reward/group_zero_std_frac": 0.71875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6637552976608276,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0342193603515625,
"signal/advantage_abs_mean": 0.76838698387146,
"signal/advantage_pre_scale_abs_mean": 0.05800668299198151,
"signal/advantage_pre_scale_std": 0.09820467084646226,
"signal/advantage_std": 0.9825891971588134,
"signal/brier_reward/centered_abs_mean": 0.08307508230209351,
"signal/brier_reward/group_std_mean": 0.11090421378612518,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.16045838296413423,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008307508379220962,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01777377799153328,
"signal/confidence_uniqueness_reward/group_std_mean": 0.02854650840163231,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.034318356215953826,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0017773778177797793,
"signal/format_reward/centered_abs_mean": 0.004150390625,
"signal/format_reward/group_std_mean": 0.011817088723182679,
"signal/format_reward/group_zero_std_frac": 0.934375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.03976389355957508,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0020751953125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.29591967463493346,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.36354907751083376,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5716009616851807,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.029591968283057213,
"step": 200
},
{
"epoch": 0.64,
"eval_completions/clipped_ratio": 0.0078125,
"eval_completions/max_length": 703.0,
"eval_completions/max_terminated_length": 703.0,
"eval_completions/mean_length": 264.0889663696289,
"eval_completions/mean_terminated_length": 266.20446014404297,
"eval_completions/min_length": 45.75,
"eval_completions/min_terminated_length": 100.75,
"eval_loss": 0.0,
"eval_num_tokens": 686049550.0,
"eval_reward": 0.7884643077850342,
"eval_reward_std": 0.23599497228860855,
"eval_rewards/accgated_coverage_0": 0.0322989858686924,
"eval_rewards/accgated_coverage_1": 0.0322989858686924,
"eval_rewards/accgated_coverage_10": 0.03210703004151583,
"eval_rewards/accgated_coverage_15": 0.028336460702121258,
"eval_rewards/accgated_coverage_20": 0.015293212374672294,
"eval_rewards/accgated_coverage_25": 0.009063742589205503,
"eval_rewards/accgated_coverage_5": 0.0322989858686924,
"eval_rewards/accuracy_reward": 0.41015625,
"eval_rewards/brier_reward": 0.7990112900733948,
"eval_rewards/confidence_uniqueness_reward": 0.8844033628702164,
"eval_rewards/format_reward": 0.9921875,
"eval_rewards/frontier_entropy_batch_reward": -0.9921875,
"eval_runtime": 43.9583,
"eval_samples_per_second": 11.374,
"eval_signal/accgated_coverage_0/centered_abs_mean": 0.05813842825591564,
"eval_signal/accgated_coverage_0/group_std_mean": 0.06973426975309849,
"eval_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.02499791793525219,
"eval_signal/accgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.005813842872157693,
"eval_signal/accgated_coverage_1/centered_abs_mean": 0.05813842825591564,
"eval_signal/accgated_coverage_1/group_std_mean": 0.06973426975309849,
"eval_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.02499791793525219,
"eval_signal/accgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.005813842872157693,
"eval_signal/accgated_coverage_10/centered_abs_mean": 0.05784142296761274,
"eval_signal/accgated_coverage_10/group_std_mean": 0.06938402354717255,
"eval_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.024872958660125732,
"eval_signal/accgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.005784142529591918,
"eval_signal/accgated_coverage_15/centered_abs_mean": 0.05046619940549135,
"eval_signal/accgated_coverage_15/group_std_mean": 0.06078723073005676,
"eval_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.021718178410083055,
"eval_signal/accgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.005046619800850749,
"eval_signal/accgated_coverage_20/centered_abs_mean": 0.024308386724442244,
"eval_signal/accgated_coverage_20/group_std_mean": 0.03040266875177622,
"eval_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.010437258519232273,
"eval_signal/accgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.00243083864916116,
"eval_signal/accgated_coverage_25/centered_abs_mean": 0.01377779245376587,
"eval_signal/accgated_coverage_25/group_std_mean": 0.017504670657217503,
"eval_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.005877090850844979,
"eval_signal/accgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0013777792628388852,
"eval_signal/accgated_coverage_5/centered_abs_mean": 0.05813842825591564,
"eval_signal/accgated_coverage_5/group_std_mean": 0.06973426975309849,
"eval_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.02499791793525219,
"eval_signal/accgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.005813842872157693,
"eval_signal/accuracy_reward/centered_abs_mean": 0.468505859375,
"eval_signal/accuracy_reward/group_std_mean": 0.49142079055309296,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.0008344203233719,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.2342529296875,
"eval_signal/advantage_abs_mean": 0.9115358293056488,
"eval_signal/advantage_pre_scale_abs_mean": 0.214738629758358,
"eval_signal/advantage_pre_scale_std": 0.23366694524884224,
"eval_signal/advantage_std": 0.9876959919929504,
"eval_signal/brier_reward/centered_abs_mean": 0.18382024392485619,
"eval_signal/brier_reward/group_std_mean": 0.23682136088609695,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.07804572582244873,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.018382024951279163,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.05135123524814844,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.08175851963460445,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.021677627228200436,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.00513512350153178,
"eval_signal/format_reward/centered_abs_mean": 0.01513671875,
"eval_signal/format_reward/group_std_mean": 0.04419417213648558,
"eval_signal/format_reward/group_zero_std_frac": 0.75,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.029725436121225357,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.007568359375,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.01513671875,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.04419417213648558,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.75,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.005945087177678943,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0015136717993300408,
"eval_steps_per_second": 0.091,
"step": 200
},
{
"epoch": 0.64,
"step": 200,
"train_probe_completions/clipped_ratio": 0.00390625,
"train_probe_completions/max_length": 971.25,
"train_probe_completions/max_terminated_length": 971.25,
"train_probe_completions/mean_length": 268.5261993408203,
"train_probe_completions/mean_terminated_length": 269.5601577758789,
"train_probe_completions/min_length": 53.75,
"train_probe_completions/min_terminated_length": 105.75,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 686049550.0,
"train_probe_reward": 0.8882567435503006,
"train_probe_reward_std": 0.21442579850554466,
"train_probe_rewards/accgated_coverage_0": 0.034551781602203846,
"train_probe_rewards/accgated_coverage_1": 0.034551781602203846,
"train_probe_rewards/accgated_coverage_10": 0.034383486956357956,
"train_probe_rewards/accgated_coverage_15": 0.0312191816046834,
"train_probe_rewards/accgated_coverage_20": 0.021827260963618755,
"train_probe_rewards/accgated_coverage_25": 0.023835682310163975,
"train_probe_rewards/accgated_coverage_5": 0.034551781602203846,
"train_probe_rewards/accuracy_reward": 0.591796875,
"train_probe_rewards/brier_reward": 0.8259834200143814,
"train_probe_rewards/confidence_uniqueness_reward": 0.8983036279678345,
"train_probe_rewards/format_reward": 0.99609375,
"train_probe_rewards/frontier_entropy_batch_reward": -0.99609375,
"train_probe_runtime": 50.9436,
"train_probe_samples_per_second": 9.815,
"train_probe_signal/accgated_coverage_0/centered_abs_mean": 0.10451133362948895,
"train_probe_signal/accgated_coverage_0/group_std_mean": 0.13702973537147045,
"train_probe_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.04985595401376486,
"train_probe_signal/accgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.01045113313011825,
"train_probe_signal/accgated_coverage_1/centered_abs_mean": 0.10451133362948895,
"train_probe_signal/accgated_coverage_1/group_std_mean": 0.13702973537147045,
"train_probe_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.04985595401376486,
"train_probe_signal/accgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.01045113313011825,
"train_probe_signal/accgated_coverage_10/centered_abs_mean": 0.10347560793161392,
"train_probe_signal/accgated_coverage_10/group_std_mean": 0.13582053780555725,
"train_probe_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.049361344426870346,
"train_probe_signal/accgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.010347560979425907,
"train_probe_signal/accgated_coverage_15/centered_abs_mean": 0.08562582544982433,
"train_probe_signal/accgated_coverage_15/group_std_mean": 0.11436908133327961,
"train_probe_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.040853239595890045,
"train_probe_signal/accgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.008562582195736468,
"train_probe_signal/accgated_coverage_20/centered_abs_mean": 0.03912342991679907,
"train_probe_signal/accgated_coverage_20/group_std_mean": 0.054525068029761314,
"train_probe_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.018668745644390583,
"train_probe_signal/accgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.003912343003321439,
"train_probe_signal/accgated_coverage_25/centered_abs_mean": 0.029649582225829363,
"train_probe_signal/accgated_coverage_25/group_std_mean": 0.037494669668376446,
"train_probe_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.014144881395623088,
"train_probe_signal/accgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0029649583157151937,
"train_probe_signal/accgated_coverage_5/centered_abs_mean": 0.10451133362948895,
"train_probe_signal/accgated_coverage_5/group_std_mean": 0.13702973537147045,
"train_probe_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.04985595401376486,
"train_probe_signal/accgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.01045113313011825,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.4691162109375,
"train_probe_signal/accuracy_reward/group_std_mean": 0.49178970605134964,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.1073245704174042,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.23455810546875,
"train_probe_signal/advantage_abs_mean": 0.8932997584342957,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.19187427684664726,
"train_probe_signal/advantage_pre_scale_std": 0.2131635658442974,
"train_probe_signal/advantage_std": 0.987650915980339,
"train_probe_signal/brier_reward/centered_abs_mean": 0.17040055245161057,
"train_probe_signal/brier_reward/group_std_mean": 0.22702116146683693,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.08043165691196918,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.01704005664214492,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.040967877954244614,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.05691860616207123,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.019406195729970932,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004096788004972041,
"train_probe_signal/format_reward/centered_abs_mean": 0.007568359375,
"train_probe_signal/format_reward/group_std_mean": 0.022097086533904076,
"train_probe_signal/format_reward/group_zero_std_frac": 0.875,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0181626807898283,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.0037841796875,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.007568359375,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.022097086533904076,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.875,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0036325360415503383,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0007568359724245965,
"train_probe_steps_per_second": 0.079
},
{
"calibration/aurc": 0.39216359278767826,
"calibration/batch_distribution_entropy": 0.9830481977996305,
"calibration/batch_entropy_100bins": 0.970614406607854,
"calibration/batch_entropy_10bins": 0.9830481977996305,
"calibration/batch_entropy_50bins": 0.9792514367058693,
"calibration/batch_uniqueness": 0.9539253263419575,
"calibration/buffer_distribution_entropy": 0.9879784074998682,
"calibration/buffer_entropy_100bins": 0.9934953291914577,
"calibration/buffer_entropy_10bins": 0.9879784074998682,
"calibration/buffer_entropy_50bins": 0.9925171708499076,
"calibration/confidence_entropy": 0.5075616523225377,
"calibration/coverage@0%": 0.005117440302438047,
"calibration/coverage@1%": 0.005117440302438047,
"calibration/coverage@10%": 0.02510592884854142,
"calibration/coverage@15%": 0.08624306332765116,
"calibration/coverage@20%": 0.14123458291994853,
"calibration/coverage@25%": 0.20605559194514073,
"calibration/coverage@30%": 0.2840726153559724,
"calibration/coverage@5%": 0.005117440302438047,
"calibration/distribution_entropy_10": 0.9830481977996305,
"calibration/distribution_entropy_100": 0.970614406607854,
"calibration/ece": 0.13875104500898902,
"calibration/mean_confidence": 0.48783545205216117,
"calibration/unique_confidence_per_question": 0.99140625,
"calibration/unique_confidences": 507.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.005859375,
"completions/max_length": 1400.4,
"completions/max_terminated_length": 1400.4,
"completions/mean_length": 257.24794921875,
"completions/mean_terminated_length": 258.78313903808595,
"completions/min_length": 0.0,
"completions/min_terminated_length": 84.8,
"epoch": 0.656,
"grad_norm": 0.014073879458010197,
"learning_rate": 1e-06,
"loss": -0.0279,
"num_tokens": 703540313.0,
"reward": 0.9258555769920349,
"reward_std": 0.08865600377321244,
"rewards/accgated_coverage_0": 0.029099652916193007,
"rewards/accgated_coverage_1": 0.029099652916193007,
"rewards/accgated_coverage_10": 0.028934436663985253,
"rewards/accgated_coverage_15": 0.0259669441729784,
"rewards/accgated_coverage_20": 0.016663705557584764,
"rewards/accgated_coverage_25": 0.015503963083028793,
"rewards/accgated_coverage_5": 0.029099652916193007,
"rewards/accuracy_reward": 0.51484375,
"rewards/brier_reward": 0.8014851450920105,
"rewards/confidence_uniqueness_reward": 0.9465346932411194,
"rewards/format_reward": 0.99404296875,
"rewards/frontier_entropy_batch_reward": -0.20826578736305237,
"signal/accgated_coverage_0/centered_abs_mean": 0.03515791557729244,
"signal/accgated_coverage_0/group_std_mean": 0.045750449597835544,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06359007135033608,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0035157914739102126,
"signal/accgated_coverage_1/centered_abs_mean": 0.03515791557729244,
"signal/accgated_coverage_1/group_std_mean": 0.045750449597835544,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06359007135033608,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0035157914739102126,
"signal/accgated_coverage_10/centered_abs_mean": 0.03485213294625282,
"signal/accgated_coverage_10/group_std_mean": 0.04534183442592621,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06302980110049247,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003485213406383991,
"signal/accgated_coverage_15/centered_abs_mean": 0.029917076230049133,
"signal/accgated_coverage_15/group_std_mean": 0.038870205730199815,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05418357625603676,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.002991707669571042,
"signal/accgated_coverage_20/centered_abs_mean": 0.015889767184853555,
"signal/accgated_coverage_20/group_std_mean": 0.020615510642528534,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.02875935472548008,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0015889767557382584,
"signal/accgated_coverage_25/centered_abs_mean": 0.01298975683748722,
"signal/accgated_coverage_25/group_std_mean": 0.01674633491784334,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.023474621400237084,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0012989756651222706,
"signal/accgated_coverage_5/centered_abs_mean": 0.03515791557729244,
"signal/accgated_coverage_5/group_std_mean": 0.045750449597835544,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06359007135033608,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0035157914739102126,
"signal/accuracy_reward/centered_abs_mean": 0.074365234375,
"signal/accuracy_reward/group_std_mean": 0.10158015489578247,
"signal/accuracy_reward/group_zero_std_frac": 0.69375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.670850682258606,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0371826171875,
"signal/advantage_abs_mean": 0.7407027244567871,
"signal/advantage_pre_scale_abs_mean": 0.06300503239035607,
"signal/advantage_pre_scale_std": 0.10972063392400741,
"signal/advantage_std": 0.9827126502990723,
"signal/brier_reward/centered_abs_mean": 0.09832997173070908,
"signal/brier_reward/group_std_mean": 0.13315942287445068,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1776996672153473,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009832997247576714,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.021487113460898398,
"signal/confidence_uniqueness_reward/group_std_mean": 0.039552245289087296,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.038640718534588815,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0021487114019691942,
"signal/format_reward/centered_abs_mean": 0.011260986328125,
"signal/format_reward/group_std_mean": 0.027375328913331033,
"signal/format_reward/group_zero_std_frac": 0.865625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.10070117712020873,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0056304931640625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.27000904083251953,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.34318543672561647,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.4870367467403412,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02700090445578098,
"step": 205
},
{
"calibration/aurc": 0.28315405589518716,
"calibration/batch_distribution_entropy": 0.9781284476144634,
"calibration/batch_entropy_100bins": 0.9673603746484707,
"calibration/batch_entropy_10bins": 0.9781284476144634,
"calibration/batch_entropy_50bins": 0.9773774525356034,
"calibration/batch_uniqueness": 0.9523873457718846,
"calibration/buffer_distribution_entropy": 0.9898232782845453,
"calibration/buffer_entropy_100bins": 0.994444942287959,
"calibration/buffer_entropy_10bins": 0.9898232782845453,
"calibration/buffer_entropy_50bins": 0.9936545538949251,
"calibration/confidence_entropy": 0.4695703422433656,
"calibration/coverage@0%": 0.040067847662757804,
"calibration/coverage@1%": 0.06476904288188132,
"calibration/coverage@10%": 0.18758377099260143,
"calibration/coverage@15%": 0.24193346539578053,
"calibration/coverage@20%": 0.32764118627871197,
"calibration/coverage@25%": 0.41888253560636735,
"calibration/coverage@30%": 0.49131108355286457,
"calibration/coverage@5%": 0.15161141697525143,
"calibration/distribution_entropy_10": 0.9781284476144634,
"calibration/distribution_entropy_100": 0.9673603746484707,
"calibration/ece": 0.12865128981918725,
"calibration/mean_confidence": 0.4981986236110421,
"calibration/unique_confidence_per_question": 0.991796875,
"calibration/unique_confidences": 507.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00458984375,
"completions/max_length": 1228.0,
"completions/max_terminated_length": 1228.0,
"completions/mean_length": 237.980859375,
"completions/mean_terminated_length": 239.06908569335937,
"completions/min_length": 0.0,
"completions/min_terminated_length": 88.2,
"epoch": 0.672,
"grad_norm": 0.006821173243224621,
"learning_rate": 1e-06,
"loss": -0.0222,
"num_tokens": 720890677.0,
"reward": 0.9358129620552063,
"reward_std": 0.0787442997097969,
"rewards/accgated_coverage_0": 0.04184006974101066,
"rewards/accgated_coverage_1": 0.04184006974101066,
"rewards/accgated_coverage_10": 0.041683945804834366,
"rewards/accgated_coverage_15": 0.03728590272367001,
"rewards/accgated_coverage_20": 0.02388199679553509,
"rewards/accgated_coverage_25": 0.02002035602927208,
"rewards/accgated_coverage_5": 0.04184006974101066,
"rewards/accuracy_reward": 0.522265625,
"rewards/brier_reward": 0.8177942276000977,
"rewards/confidence_uniqueness_reward": 0.9441993474960327,
"rewards/format_reward": 0.99541015625,
"rewards/frontier_entropy_batch_reward": -0.2406352460384369,
"signal/accgated_coverage_0/centered_abs_mean": 0.040953823179006574,
"signal/accgated_coverage_0/group_std_mean": 0.05248353555798531,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08236506283283233,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0040953824762254955,
"signal/accgated_coverage_1/centered_abs_mean": 0.040953823179006574,
"signal/accgated_coverage_1/group_std_mean": 0.05248353555798531,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08236506283283233,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0040953824762254955,
"signal/accgated_coverage_10/centered_abs_mean": 0.04038825705647468,
"signal/accgated_coverage_10/group_std_mean": 0.05175998955965042,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08127994686365128,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004038825817406178,
"signal/accgated_coverage_15/centered_abs_mean": 0.03313384279608726,
"signal/accgated_coverage_15/group_std_mean": 0.04257287010550499,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06671571135520935,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003313384298235178,
"signal/accgated_coverage_20/centered_abs_mean": 0.018710185959935188,
"signal/accgated_coverage_20/group_std_mean": 0.023820379748940468,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03774301260709763,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.001871018554084003,
"signal/accgated_coverage_25/centered_abs_mean": 0.014438809268176556,
"signal/accgated_coverage_25/group_std_mean": 0.01827642824500799,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.02910972833633423,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.001443880912847817,
"signal/accgated_coverage_5/centered_abs_mean": 0.040953823179006574,
"signal/accgated_coverage_5/group_std_mean": 0.05248353555798531,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08236506283283233,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0040953824762254955,
"signal/accuracy_reward/centered_abs_mean": 0.0782958984375,
"signal/accuracy_reward/group_std_mean": 0.10683944523334503,
"signal/accuracy_reward/group_zero_std_frac": 0.678125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7833632588386535,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.03914794921875,
"signal/advantage_abs_mean": 0.739534056186676,
"signal/advantage_pre_scale_abs_mean": 0.05598316714167595,
"signal/advantage_pre_scale_std": 0.09874502718448638,
"signal/advantage_std": 0.9824853301048279,
"signal/brier_reward/centered_abs_mean": 0.09665140360593796,
"signal/brier_reward/group_std_mean": 0.12781293392181398,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19586214125156404,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009665140882134438,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.020561978220939636,
"signal/confidence_uniqueness_reward/group_std_mean": 0.03474752940237522,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.04224793538451195,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.002056197845377028,
"signal/format_reward/centered_abs_mean": 0.008343505859375,
"signal/format_reward/group_std_mean": 0.020251142606139184,
"signal/format_reward/group_zero_std_frac": 0.896875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.08657302036881447,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0041717529296875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2655516296625137,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.34094269275665284,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5395448684692383,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026555164158344267,
"step": 210
},
{
"calibration/aurc": 0.2792192523604893,
"calibration/batch_distribution_entropy": 0.9566927004044243,
"calibration/batch_entropy_100bins": 0.956963159495429,
"calibration/batch_entropy_10bins": 0.9566927004044243,
"calibration/batch_entropy_50bins": 0.9636204768382731,
"calibration/batch_uniqueness": 0.948669977502167,
"calibration/buffer_distribution_entropy": 0.9898413579362421,
"calibration/buffer_entropy_100bins": 0.9944850364761466,
"calibration/buffer_entropy_10bins": 0.9898413579362421,
"calibration/buffer_entropy_50bins": 0.9936959965033839,
"calibration/confidence_entropy": 0.4771618609638734,
"calibration/coverage@0%": 0.03061039868001995,
"calibration/coverage@1%": 0.03061039868001995,
"calibration/coverage@10%": 0.22824731207551513,
"calibration/coverage@15%": 0.3005039100571736,
"calibration/coverage@20%": 0.43389756340892516,
"calibration/coverage@25%": 0.5344393077779056,
"calibration/coverage@30%": 0.5989174628755611,
"calibration/coverage@5%": 0.11696333985649054,
"calibration/distribution_entropy_10": 0.9566927004044243,
"calibration/distribution_entropy_100": 0.956963159495429,
"calibration/ece": 0.12260640395380626,
"calibration/mean_confidence": 0.5166461235681139,
"calibration/unique_confidence_per_question": 0.99140625,
"calibration/unique_confidences": 507.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0046875,
"completions/max_length": 1315.6,
"completions/max_terminated_length": 1315.6,
"completions/mean_length": 239.36162109375,
"completions/mean_terminated_length": 240.49996337890624,
"completions/min_length": 0.0,
"completions/min_terminated_length": 85.4,
"epoch": 0.688,
"grad_norm": 0.0074991402216255665,
"learning_rate": 1e-06,
"loss": -0.0239,
"num_tokens": 738295660.0,
"reward": 0.9411136507987976,
"reward_std": 0.08507151305675506,
"rewards/accgated_coverage_0": 0.03155128005892038,
"rewards/accgated_coverage_1": 0.03155128005892038,
"rewards/accgated_coverage_10": 0.03145138081163168,
"rewards/accgated_coverage_15": 0.02701001465320587,
"rewards/accgated_coverage_20": 0.019650851003825666,
"rewards/accgated_coverage_25": 0.019739412888884544,
"rewards/accgated_coverage_5": 0.03155111595988273,
"rewards/accuracy_reward": 0.54736328125,
"rewards/brier_reward": 0.8122114896774292,
"rewards/confidence_uniqueness_reward": 0.9460292339324952,
"rewards/format_reward": 0.9953125,
"rewards/frontier_entropy_batch_reward": -0.25298852026462554,
"signal/accgated_coverage_0/centered_abs_mean": 0.043075624480843545,
"signal/accgated_coverage_0/group_std_mean": 0.05640400499105454,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07659767717123031,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0043075625784695145,
"signal/accgated_coverage_1/centered_abs_mean": 0.043075624480843545,
"signal/accgated_coverage_1/group_std_mean": 0.05640400499105454,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07659767717123031,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0043075625784695145,
"signal/accgated_coverage_10/centered_abs_mean": 0.04276303648948669,
"signal/accgated_coverage_10/group_std_mean": 0.055990930646657944,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07605005651712418,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0042763037607073786,
"signal/accgated_coverage_15/centered_abs_mean": 0.03298235461115837,
"signal/accgated_coverage_15/group_std_mean": 0.04331007674336433,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05889697000384331,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0032982355682179333,
"signal/accgated_coverage_20/centered_abs_mean": 0.02059024143964052,
"signal/accgated_coverage_20/group_std_mean": 0.026587991788983346,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03684881329536438,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0020590241765603423,
"signal/accgated_coverage_25/centered_abs_mean": 0.01691297572106123,
"signal/accgated_coverage_25/group_std_mean": 0.021273587271571158,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03004997633397579,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0016912976047024132,
"signal/accgated_coverage_5/centered_abs_mean": 0.043075176700949666,
"signal/accgated_coverage_5/group_std_mean": 0.05640345141291618,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07659696787595749,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004307517874985933,
"signal/accuracy_reward/centered_abs_mean": 0.078936767578125,
"signal/accuracy_reward/group_std_mean": 0.11089948713779449,
"signal/accuracy_reward/group_zero_std_frac": 0.65625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.70556138753891,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0394683837890625,
"signal/advantage_abs_mean": 0.7577574968338012,
"signal/advantage_pre_scale_abs_mean": 0.06174611151218414,
"signal/advantage_pre_scale_std": 0.10472253262996674,
"signal/advantage_std": 0.9827207207679749,
"signal/brier_reward/centered_abs_mean": 0.10126348733901977,
"signal/brier_reward/group_std_mean": 0.1337401956319809,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18201247751712799,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010126348584890366,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.020276128128170968,
"signal/confidence_uniqueness_reward/group_std_mean": 0.03399667665362358,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03674969710409641,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.002027612901292741,
"signal/format_reward/centered_abs_mean": 0.008642578125,
"signal/format_reward/group_std_mean": 0.01992634441703558,
"signal/format_reward/group_zero_std_frac": 0.903125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.07950785346329212,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0043212890625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2949717164039612,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.36637176275253297,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5286233246326446,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02949717193841934,
"step": 215
},
{
"calibration/aurc": 0.22692466464313932,
"calibration/batch_distribution_entropy": 0.9641940416180546,
"calibration/batch_entropy_100bins": 0.9619821754672346,
"calibration/batch_entropy_10bins": 0.9641940416180546,
"calibration/batch_entropy_50bins": 0.9684448233035102,
"calibration/batch_uniqueness": 0.9504850823404295,
"calibration/buffer_distribution_entropy": 0.9903376968487738,
"calibration/buffer_entropy_100bins": 0.9947417458502505,
"calibration/buffer_entropy_10bins": 0.9903376968487738,
"calibration/buffer_entropy_50bins": 0.993989233910378,
"calibration/confidence_entropy": 0.46354835983236925,
"calibration/coverage@0%": 0.067679608532548,
"calibration/coverage@1%": 0.0724040179813669,
"calibration/coverage@10%": 0.20556670159541665,
"calibration/coverage@15%": 0.3414797580925503,
"calibration/coverage@20%": 0.4624978872932092,
"calibration/coverage@25%": 0.5962122581892076,
"calibration/coverage@30%": 0.6998552811309904,
"calibration/coverage@5%": 0.109016711091575,
"calibration/distribution_entropy_10": 0.9641940416180546,
"calibration/distribution_entropy_100": 0.9619821754672346,
"calibration/ece": 0.0914957362264657,
"calibration/mean_confidence": 0.5411261651146633,
"calibration/unique_confidence_per_question": 0.9953125,
"calibration/unique_confidences": 509.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.003515625,
"completions/max_length": 1379.6,
"completions/max_terminated_length": 1379.6,
"completions/mean_length": 231.32412109375,
"completions/mean_terminated_length": 232.14586791992187,
"completions/min_length": 0.0,
"completions/min_terminated_length": 81.2,
"epoch": 0.704,
"grad_norm": 0.0077322134748101234,
"learning_rate": 1e-06,
"loss": -0.0222,
"num_tokens": 755530563.0,
"reward": 0.9389920711517334,
"reward_std": 0.09092791378498077,
"rewards/accgated_coverage_0": 0.035602913424372676,
"rewards/accgated_coverage_1": 0.035602913424372676,
"rewards/accgated_coverage_10": 0.035379021614789966,
"rewards/accgated_coverage_15": 0.029057538881897927,
"rewards/accgated_coverage_20": 0.020715073123574256,
"rewards/accgated_coverage_25": 0.01995195373892784,
"rewards/accgated_coverage_5": 0.03560251221060753,
"rewards/accuracy_reward": 0.532421875,
"rewards/brier_reward": 0.8180535912513733,
"rewards/confidence_uniqueness_reward": 0.947830855846405,
"rewards/format_reward": 0.996484375,
"rewards/frontier_entropy_batch_reward": -0.23240697681903838,
"signal/accgated_coverage_0/centered_abs_mean": 0.040877356007695195,
"signal/accgated_coverage_0/group_std_mean": 0.05286289900541306,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0696209765970707,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004087735712528229,
"signal/accgated_coverage_1/centered_abs_mean": 0.040877356007695195,
"signal/accgated_coverage_1/group_std_mean": 0.05286289900541306,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0696209765970707,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004087735712528229,
"signal/accgated_coverage_10/centered_abs_mean": 0.04057118110358715,
"signal/accgated_coverage_10/group_std_mean": 0.052476315200328826,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06909865662455558,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004057118156924844,
"signal/accgated_coverage_15/centered_abs_mean": 0.029750393703579903,
"signal/accgated_coverage_15/group_std_mean": 0.038633452355861665,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.050678076595067977,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.002975039416924119,
"signal/accgated_coverage_20/centered_abs_mean": 0.019006701558828352,
"signal/accgated_coverage_20/group_std_mean": 0.024423104897141456,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.032375164702534674,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0019006703048944473,
"signal/accgated_coverage_25/centered_abs_mean": 0.016715809144079684,
"signal/accgated_coverage_25/group_std_mean": 0.021140381693840027,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.0284941665828228,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0016715808771550656,
"signal/accgated_coverage_5/centered_abs_mean": 0.04087658077478409,
"signal/accgated_coverage_5/group_std_mean": 0.052861882746219634,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06961965933442116,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004087658366188407,
"signal/accuracy_reward/centered_abs_mean": 0.0920654296875,
"signal/accuracy_reward/group_std_mean": 0.12160763740539551,
"signal/accuracy_reward/group_zero_std_frac": 0.65,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.786667776107788,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.04603271484375,
"signal/advantage_abs_mean": 0.7658281564712525,
"signal/advantage_pre_scale_abs_mean": 0.06736339032649993,
"signal/advantage_pre_scale_std": 0.11175605654716492,
"signal/advantage_std": 0.9828084111213684,
"signal/brier_reward/centered_abs_mean": 0.10543047487735749,
"signal/brier_reward/group_std_mean": 0.13940512239933014,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18008263111114503,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.010543047264218331,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.018053193762898444,
"signal/confidence_uniqueness_reward/group_std_mean": 0.031151178106665612,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03084605559706688,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0018053193809464574,
"signal/format_reward/centered_abs_mean": 0.0067138671875,
"signal/format_reward/group_std_mean": 0.017474005930125713,
"signal/format_reward/group_zero_std_frac": 0.909375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.05733720734715462,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.00335693359375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28432013988494875,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3522124171257019,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.4862986385822296,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02843201532959938,
"step": 220
},
{
"calibration/aurc": 0.2321254343006741,
"calibration/batch_distribution_entropy": 0.9724462303987096,
"calibration/batch_entropy_100bins": 0.9680618544609747,
"calibration/batch_entropy_10bins": 0.9724462303987096,
"calibration/batch_entropy_50bins": 0.9734966261652783,
"calibration/batch_uniqueness": 0.9525465110283473,
"calibration/buffer_distribution_entropy": 0.9905084817479126,
"calibration/buffer_entropy_100bins": 0.9948056135036689,
"calibration/buffer_entropy_10bins": 0.9905084817479126,
"calibration/buffer_entropy_50bins": 0.9940673557175665,
"calibration/confidence_entropy": 0.4713699075847605,
"calibration/coverage@0%": 0.04295034429212284,
"calibration/coverage@1%": 0.07482285425228222,
"calibration/coverage@10%": 0.2571288664240675,
"calibration/coverage@15%": 0.42293772743670993,
"calibration/coverage@20%": 0.5053176687933306,
"calibration/coverage@25%": 0.579075432195104,
"calibration/coverage@30%": 0.7083562674479633,
"calibration/coverage@5%": 0.14088468958810427,
"calibration/distribution_entropy_10": 0.9724462303987096,
"calibration/distribution_entropy_100": 0.9680618544609747,
"calibration/ece": 0.13794344563035618,
"calibration/mean_confidence": 0.45736570422651823,
"calibration/unique_confidence_per_question": 0.9921875,
"calibration/unique_confidences": 508.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.003125,
"completions/max_length": 1181.8,
"completions/max_terminated_length": 1181.8,
"completions/mean_length": 236.59169921875,
"completions/mean_terminated_length": 237.33655700683593,
"completions/min_length": 0.0,
"completions/min_terminated_length": 100.4,
"epoch": 0.72,
"grad_norm": 0.007477594539523125,
"learning_rate": 1e-06,
"loss": -0.0031,
"num_tokens": 772963118.0,
"reward": 0.9516620635986328,
"reward_std": 0.07770900875329971,
"rewards/accgated_coverage_0": 0.032990981638431546,
"rewards/accgated_coverage_1": 0.032990981638431546,
"rewards/accgated_coverage_10": 0.03284294828772545,
"rewards/accgated_coverage_15": 0.02697284035384655,
"rewards/accgated_coverage_20": 0.020624830573797225,
"rewards/accgated_coverage_25": 0.021644752472639084,
"rewards/accgated_coverage_5": 0.03299048468470574,
"rewards/accuracy_reward": 0.55947265625,
"rewards/brier_reward": 0.8174762964248657,
"rewards/confidence_uniqueness_reward": 0.9474241018295289,
"rewards/format_reward": 0.996875,
"rewards/frontier_entropy_batch_reward": -0.23107567429542542,
"signal/accgated_coverage_0/centered_abs_mean": 0.046656031906604764,
"signal/accgated_coverage_0/group_std_mean": 0.060182292759418485,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.09277255088090897,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004665603395551443,
"signal/accgated_coverage_1/centered_abs_mean": 0.046656031906604764,
"signal/accgated_coverage_1/group_std_mean": 0.060182292759418485,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.09277255088090897,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004665603395551443,
"signal/accgated_coverage_10/centered_abs_mean": 0.04628909006714821,
"signal/accgated_coverage_10/group_std_mean": 0.05971537530422211,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.09204467386007309,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004628909006714821,
"signal/accgated_coverage_15/centered_abs_mean": 0.0334114033728838,
"signal/accgated_coverage_15/group_std_mean": 0.04316308423876762,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06641377955675125,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003341140411794186,
"signal/accgated_coverage_20/centered_abs_mean": 0.020335762202739714,
"signal/accgated_coverage_20/group_std_mean": 0.026082827895879745,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.04038082361221314,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0020335762295871974,
"signal/accgated_coverage_25/centered_abs_mean": 0.01695841308683157,
"signal/accgated_coverage_25/group_std_mean": 0.021627087518572806,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03363958969712257,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.001695841271430254,
"signal/accgated_coverage_5/centered_abs_mean": 0.046655257046222684,
"signal/accgated_coverage_5/group_std_mean": 0.06018129289150238,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.09277101010084152,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.00466552572324872,
"signal/accuracy_reward/centered_abs_mean": 0.076776123046875,
"signal/accuracy_reward/group_std_mean": 0.10414892733097077,
"signal/accuracy_reward/group_zero_std_frac": 0.69375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7590139627456665,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0383880615234375,
"signal/advantage_abs_mean": 0.7622187852859497,
"signal/advantage_pre_scale_abs_mean": 0.057844098657369614,
"signal/advantage_pre_scale_std": 0.09815582484006882,
"signal/advantage_std": 0.982541561126709,
"signal/brier_reward/centered_abs_mean": 0.09804008901119232,
"signal/brier_reward/group_std_mean": 0.12837443351745606,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19414441883563996,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009804009087383747,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.017306267842650415,
"signal/confidence_uniqueness_reward/group_std_mean": 0.02669024094939232,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03402584083378315,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.00173062679823488,
"signal/format_reward/centered_abs_mean": 0.00560302734375,
"signal/format_reward/group_std_mean": 0.012464468041434884,
"signal/format_reward/group_zero_std_frac": 0.940625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.05448856502771378,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.002801513671875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2716618120670319,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.34504327178001404,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5372507333755493,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027166182547807692,
"step": 225
},
{
"calibration/aurc": 0.22857583013561342,
"calibration/batch_distribution_entropy": 0.9561877111180614,
"calibration/batch_entropy_100bins": 0.9536644987816254,
"calibration/batch_entropy_10bins": 0.9561877111180614,
"calibration/batch_entropy_50bins": 0.9610272392371939,
"calibration/batch_uniqueness": 0.9485505723561689,
"calibration/buffer_distribution_entropy": 0.9904877895919147,
"calibration/buffer_entropy_100bins": 0.9947859128414864,
"calibration/buffer_entropy_10bins": 0.9904877895919147,
"calibration/buffer_entropy_50bins": 0.9940575187564692,
"calibration/confidence_entropy": 0.44142700770924925,
"calibration/coverage@0%": 0.05607079914480527,
"calibration/coverage@1%": 0.06273746581147194,
"calibration/coverage@10%": 0.1529274167918641,
"calibration/coverage@15%": 0.3825168866048384,
"calibration/coverage@20%": 0.4886995207586963,
"calibration/coverage@25%": 0.636531245786625,
"calibration/coverage@30%": 0.7164133555558765,
"calibration/coverage@5%": 0.09136491679186409,
"calibration/distribution_entropy_10": 0.9561877111180614,
"calibration/distribution_entropy_100": 0.9536644987816254,
"calibration/ece": 0.17047584453687753,
"calibration/mean_confidence": 0.547597776455292,
"calibration/unique_confidence_per_question": 0.9953125,
"calibration/unique_confidences": 509.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.001171875,
"completions/max_length": 1092.8,
"completions/max_terminated_length": 1092.8,
"completions/mean_length": 231.80146484375,
"completions/mean_terminated_length": 232.07601928710938,
"completions/min_length": 39.8,
"completions/min_terminated_length": 99.8,
"epoch": 0.736,
"grad_norm": 0.011304637417197227,
"learning_rate": 1e-06,
"loss": -0.0075,
"num_tokens": 790276349.0,
"reward": 0.9554084539413452,
"reward_std": 0.07698090374469757,
"rewards/accgated_coverage_0": 0.030976343713700772,
"rewards/accgated_coverage_1": 0.030976343713700772,
"rewards/accgated_coverage_10": 0.031016813311725854,
"rewards/accgated_coverage_15": 0.027573382295668126,
"rewards/accgated_coverage_20": 0.02284515053033829,
"rewards/accgated_coverage_25": 0.025892441719770433,
"rewards/accgated_coverage_5": 0.030976308975368738,
"rewards/accuracy_reward": 0.574609375,
"rewards/brier_reward": 0.8094848275184632,
"rewards/confidence_uniqueness_reward": 0.946896755695343,
"rewards/format_reward": 0.99873046875,
"rewards/frontier_entropy_batch_reward": -0.2692530870437622,
"signal/accgated_coverage_0/centered_abs_mean": 0.047133274376392365,
"signal/accgated_coverage_0/group_std_mean": 0.06041975915431976,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.09118882566690445,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.00471332767046988,
"signal/accgated_coverage_1/centered_abs_mean": 0.047133274376392365,
"signal/accgated_coverage_1/group_std_mean": 0.06041975915431976,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.09118882566690445,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.00471332767046988,
"signal/accgated_coverage_10/centered_abs_mean": 0.04686759561300278,
"signal/accgated_coverage_10/group_std_mean": 0.06007424667477608,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.09067109674215316,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0046867595985531805,
"signal/accgated_coverage_15/centered_abs_mean": 0.03247784599661827,
"signal/accgated_coverage_15/group_std_mean": 0.04144761227071285,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06295121684670449,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0032477845903486013,
"signal/accgated_coverage_20/centered_abs_mean": 0.021346032992005347,
"signal/accgated_coverage_20/group_std_mean": 0.0269421249628067,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.04133254289627075,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.002134603355079889,
"signal/accgated_coverage_25/centered_abs_mean": 0.019845048524439336,
"signal/accgated_coverage_25/group_std_mean": 0.025061707571148874,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.038311416655778883,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.001984504843130708,
"signal/accgated_coverage_5/centered_abs_mean": 0.04713238552212715,
"signal/accgated_coverage_5/group_std_mean": 0.06041858121752739,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.09118715524673462,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0047132385894656185,
"signal/accuracy_reward/centered_abs_mean": 0.0810546875,
"signal/accuracy_reward/group_std_mean": 0.10810347646474838,
"signal/accuracy_reward/group_zero_std_frac": 0.684375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7821998298168182,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.04052734375,
"signal/advantage_abs_mean": 0.7715708971023559,
"signal/advantage_pre_scale_abs_mean": 0.05839451476931572,
"signal/advantage_pre_scale_std": 0.09450199007987976,
"signal/advantage_std": 0.9825576782226563,
"signal/brier_reward/centered_abs_mean": 0.10095490515232086,
"signal/brier_reward/group_std_mean": 0.13110905587673188,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.196344730257988,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.0100954907014966,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01719374507665634,
"signal/confidence_uniqueness_reward/group_std_mean": 0.024562888219952582,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03373674005270004,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0017193744657561182,
"signal/format_reward/centered_abs_mean": 0.002435302734375,
"signal/format_reward/group_std_mean": 0.006508936360478401,
"signal/format_reward/group_zero_std_frac": 0.965625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.022646866738796234,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0012176513671875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.3013079404830933,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.37587220668792726,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5925543427467346,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.030130793154239655,
"step": 230
},
{
"calibration/aurc": 0.2515122508201165,
"calibration/batch_distribution_entropy": 0.9794667232408376,
"calibration/batch_entropy_100bins": 0.9692927140006175,
"calibration/batch_entropy_10bins": 0.9794667232408376,
"calibration/batch_entropy_50bins": 0.9771586457703514,
"calibration/batch_uniqueness": 0.9531182494024673,
"calibration/buffer_distribution_entropy": 0.9903217587522288,
"calibration/buffer_entropy_100bins": 0.9946750395801901,
"calibration/buffer_entropy_10bins": 0.9903217587522288,
"calibration/buffer_entropy_50bins": 0.9939247551954591,
"calibration/confidence_entropy": 0.4605606238831365,
"calibration/coverage@0%": 0.048442086594911934,
"calibration/coverage@1%": 0.048442086594911934,
"calibration/coverage@10%": 0.23829959637964776,
"calibration/coverage@15%": 0.33637399094911935,
"calibration/coverage@20%": 0.4570893468688846,
"calibration/coverage@25%": 0.5372011068982387,
"calibration/coverage@30%": 0.6294145976027397,
"calibration/coverage@5%": 0.12773896159491194,
"calibration/distribution_entropy_10": 0.9794667232408376,
"calibration/distribution_entropy_100": 0.9692927140006175,
"calibration/ece": 0.10564030415042205,
"calibration/mean_confidence": 0.4995648915637366,
"calibration/unique_confidence_per_question": 0.99921875,
"calibration/unique_confidences": 511.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00048828125,
"completions/max_length": 886.2,
"completions/max_terminated_length": 886.2,
"completions/mean_length": 224.879296875,
"completions/mean_terminated_length": 224.988232421875,
"completions/min_length": 22.0,
"completions/min_terminated_length": 109.0,
"epoch": 0.752,
"grad_norm": 0.007847636938095093,
"learning_rate": 1e-06,
"loss": -0.0022,
"num_tokens": 807806313.0,
"reward": 0.9517120122909546,
"reward_std": 0.07299316674470901,
"rewards/accgated_coverage_0": 0.0342946071876213,
"rewards/accgated_coverage_1": 0.0342946071876213,
"rewards/accgated_coverage_10": 0.03421574032399803,
"rewards/accgated_coverage_15": 0.0260434458963573,
"rewards/accgated_coverage_20": 0.018571365531533955,
"rewards/accgated_coverage_25": 0.021483676321804524,
"rewards/accgated_coverage_5": 0.03429428436793387,
"rewards/accuracy_reward": 0.5576171875,
"rewards/brier_reward": 0.8174810171127319,
"rewards/confidence_uniqueness_reward": 0.9508139848709106,
"rewards/format_reward": 0.99951171875,
"rewards/frontier_entropy_batch_reward": -0.24001707434654235,
"signal/accgated_coverage_0/centered_abs_mean": 0.04164009690284729,
"signal/accgated_coverage_0/group_std_mean": 0.05486035719513893,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08213849514722824,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004164009727537632,
"signal/accgated_coverage_1/centered_abs_mean": 0.04164009690284729,
"signal/accgated_coverage_1/group_std_mean": 0.05486035719513893,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08213849514722824,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004164009727537632,
"signal/accgated_coverage_10/centered_abs_mean": 0.041509507969021794,
"signal/accgated_coverage_10/group_std_mean": 0.05469439253211021,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08188254684209824,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004150950908660888,
"signal/accgated_coverage_15/centered_abs_mean": 0.031168679520487786,
"signal/accgated_coverage_15/group_std_mean": 0.041431061923503876,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06170603632926941,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0031168680638074876,
"signal/accgated_coverage_20/centered_abs_mean": 0.01970433034002781,
"signal/accgated_coverage_20/group_std_mean": 0.026028184592723845,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.038947059959173205,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.001970433117821813,
"signal/accgated_coverage_25/centered_abs_mean": 0.017557105235755443,
"signal/accgated_coverage_25/group_std_mean": 0.022410409152507783,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03476561158895493,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0017557105980813502,
"signal/accgated_coverage_5/centered_abs_mean": 0.04163953140377998,
"signal/accgated_coverage_5/group_std_mean": 0.054859652370214465,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08213742077350616,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004163953196257353,
"signal/accuracy_reward/centered_abs_mean": 0.0763671875,
"signal/accuracy_reward/group_std_mean": 0.10212709307670594,
"signal/accuracy_reward/group_zero_std_frac": 0.703125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7433548927307129,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.03818359375,
"signal/advantage_abs_mean": 0.7848060369491577,
"signal/advantage_pre_scale_abs_mean": 0.05702820718288422,
"signal/advantage_pre_scale_std": 0.09056562930345535,
"signal/advantage_std": 0.9825245618820191,
"signal/brier_reward/centered_abs_mean": 0.0946608766913414,
"signal/brier_reward/group_std_mean": 0.12278099656105042,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1882014036178589,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009466087631881236,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.013663103990256786,
"signal/confidence_uniqueness_reward/group_std_mean": 0.018707066029310226,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.027282388880848885,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013663104502484203,
"signal/format_reward/centered_abs_mean": 0.000946044921875,
"signal/format_reward/group_std_mean": 0.0027621358167380095,
"signal/format_reward/group_zero_std_frac": 0.984375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.009747405909001828,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0004730224609375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28806787729263306,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.35946682691574094,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5752003192901611,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028806789591908456,
"step": 235
},
{
"calibration/aurc": 0.2360991059768581,
"calibration/batch_distribution_entropy": 0.9796125365845967,
"calibration/batch_entropy_100bins": 0.9679281734145213,
"calibration/batch_entropy_10bins": 0.9796125365845967,
"calibration/batch_entropy_50bins": 0.9773396249825288,
"calibration/batch_uniqueness": 0.9525554124121859,
"calibration/buffer_distribution_entropy": 0.9905937254901949,
"calibration/buffer_entropy_100bins": 0.9948069080114852,
"calibration/buffer_entropy_10bins": 0.9905937254901949,
"calibration/buffer_entropy_50bins": 0.9940814467636867,
"calibration/confidence_entropy": 0.4958575384227478,
"calibration/coverage@0%": 0.13365949119373774,
"calibration/coverage@1%": 0.15162824119373775,
"calibration/coverage@10%": 0.2897053877201566,
"calibration/coverage@15%": 0.34369419642857146,
"calibration/coverage@20%": 0.40550697162426613,
"calibration/coverage@25%": 0.5847358121330724,
"calibration/coverage@30%": 0.6767123287671233,
"calibration/coverage@5%": 0.2365123532289628,
"calibration/distribution_entropy_10": 0.9796125365845967,
"calibration/distribution_entropy_100": 0.9679281734145213,
"calibration/ece": 0.15638391811279378,
"calibration/mean_confidence": 0.49494100795368656,
"calibration/unique_confidence_per_question": 0.99765625,
"calibration/unique_confidences": 510.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00087890625,
"completions/max_length": 897.4,
"completions/max_terminated_length": 897.4,
"completions/mean_length": 237.68349609375,
"completions/mean_terminated_length": 237.89142456054688,
"completions/min_length": 0.0,
"completions/min_terminated_length": 117.4,
"epoch": 0.768,
"grad_norm": 0.0094685647636652,
"learning_rate": 1e-06,
"loss": -0.0038,
"num_tokens": 825172896.0,
"reward": 0.9281659960746765,
"reward_std": 0.07511979043483734,
"rewards/accgated_coverage_0": 0.03205515444278717,
"rewards/accgated_coverage_1": 0.03205515444278717,
"rewards/accgated_coverage_10": 0.03200423792004585,
"rewards/accgated_coverage_15": 0.025837481021881104,
"rewards/accgated_coverage_20": 0.016886461153626443,
"rewards/accgated_coverage_25": 0.017518024519085885,
"rewards/accgated_coverage_5": 0.03205499015748501,
"rewards/accuracy_reward": 0.51015625,
"rewards/brier_reward": 0.8242544651031494,
"rewards/confidence_uniqueness_reward": 0.9506318688392639,
"rewards/format_reward": 0.99912109375,
"rewards/frontier_entropy_batch_reward": -0.22802455723285675,
"signal/accgated_coverage_0/centered_abs_mean": 0.03561207354068756,
"signal/accgated_coverage_0/group_std_mean": 0.046975846588611606,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0703155055642128,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.003561207326129079,
"signal/accgated_coverage_1/centered_abs_mean": 0.03561207354068756,
"signal/accgated_coverage_1/group_std_mean": 0.046975846588611606,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0703155055642128,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.003561207326129079,
"signal/accgated_coverage_10/centered_abs_mean": 0.035506021603941916,
"signal/accgated_coverage_10/group_std_mean": 0.04683723300695419,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.0701051376760006,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0035506021231412888,
"signal/accgated_coverage_15/centered_abs_mean": 0.025805173441767693,
"signal/accgated_coverage_15/group_std_mean": 0.03390178754925728,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05101533234119415,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0025805174373090267,
"signal/accgated_coverage_20/centered_abs_mean": 0.015104633755981922,
"signal/accgated_coverage_20/group_std_mean": 0.019619957730174064,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.029813244193792342,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0015104633755981922,
"signal/accgated_coverage_25/centered_abs_mean": 0.013850444182753562,
"signal/accgated_coverage_25/group_std_mean": 0.017736049741506575,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.027254897728562356,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0013850443996489049,
"signal/accgated_coverage_5/centered_abs_mean": 0.03561191745102406,
"signal/accgated_coverage_5/group_std_mean": 0.04697564542293549,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07031520903110504,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0035611918196082116,
"signal/accuracy_reward/centered_abs_mean": 0.07623291015625,
"signal/accuracy_reward/group_std_mean": 0.10346686244010925,
"signal/accuracy_reward/group_zero_std_frac": 0.690625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7489666223526001,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.038116455078125,
"signal/advantage_abs_mean": 0.7693462014198303,
"signal/advantage_pre_scale_abs_mean": 0.05771623626351356,
"signal/advantage_pre_scale_std": 0.09313028007745743,
"signal/advantage_std": 0.9825564384460449,
"signal/brier_reward/centered_abs_mean": 0.0894073262810707,
"signal/brier_reward/group_std_mean": 0.11743923127651215,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17550100684165953,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008940733037889004,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.013525043800473213,
"signal/confidence_uniqueness_reward/group_std_mean": 0.01963619813323021,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.026574809476733206,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001352504384703934,
"signal/format_reward/centered_abs_mean": 0.001702880859375,
"signal/format_reward/group_std_mean": 0.004971844470128417,
"signal/format_reward/group_zero_std_frac": 0.971875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.016506462730467318,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0008514404296875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.27548631429672243,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.34585945010185243,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5404228210449219,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0275486309081316,
"step": 240
},
{
"calibration/aurc": 0.27954099279776556,
"calibration/batch_distribution_entropy": 0.9616386407545183,
"calibration/batch_entropy_100bins": 0.9588806945818666,
"calibration/batch_entropy_10bins": 0.9616386407545183,
"calibration/batch_entropy_50bins": 0.9660992837119222,
"calibration/batch_uniqueness": 0.9503832631883938,
"calibration/buffer_distribution_entropy": 0.9907114072492103,
"calibration/buffer_entropy_100bins": 0.9948650155575696,
"calibration/buffer_entropy_10bins": 0.9907114072492103,
"calibration/buffer_entropy_50bins": 0.994171840633628,
"calibration/confidence_entropy": 0.4479261259770905,
"calibration/coverage@0%": 0.10199899094911938,
"calibration/coverage@1%": 0.10863961594911938,
"calibration/coverage@10%": 0.2407679488747554,
"calibration/coverage@15%": 0.3244557240704501,
"calibration/coverage@20%": 0.3952222969667319,
"calibration/coverage@25%": 0.4867753180039139,
"calibration/coverage@30%": 0.5559671599804306,
"calibration/coverage@5%": 0.18722098214285715,
"calibration/distribution_entropy_10": 0.9616386407545183,
"calibration/distribution_entropy_100": 0.9588806945818666,
"calibration/ece": 0.1577776661245444,
"calibration/mean_confidence": 0.5146126241922342,
"calibration/unique_confidence_per_question": 0.997265625,
"calibration/unique_confidences": 510.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.000390625,
"completions/max_length": 716.4,
"completions/max_terminated_length": 716.4,
"completions/mean_length": 241.2662109375,
"completions/mean_terminated_length": 241.36029663085938,
"completions/min_length": 41.0,
"completions/min_terminated_length": 110.4,
"epoch": 0.784,
"grad_norm": 0.007176261395215988,
"learning_rate": 1e-06,
"loss": -0.0003,
"num_tokens": 842817830.0,
"reward": 0.9434802532196045,
"reward_std": 0.07407928258180618,
"rewards/accgated_coverage_0": 0.021479576639831067,
"rewards/accgated_coverage_1": 0.021479576639831067,
"rewards/accgated_coverage_10": 0.02146818898618221,
"rewards/accgated_coverage_15": 0.02032339256256819,
"rewards/accgated_coverage_20": 0.017040536180138587,
"rewards/accgated_coverage_25": 0.022098255157470704,
"rewards/accgated_coverage_5": 0.021477173455059527,
"rewards/accuracy_reward": 0.55634765625,
"rewards/brier_reward": 0.8033675789833069,
"rewards/confidence_uniqueness_reward": 0.9504077434539795,
"rewards/format_reward": 0.9994140625,
"rewards/frontier_entropy_batch_reward": -0.24314816892147065,
"signal/accgated_coverage_0/centered_abs_mean": 0.042506469041109086,
"signal/accgated_coverage_0/group_std_mean": 0.057106205448508265,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07912889569997787,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004250647011213004,
"signal/accgated_coverage_1/centered_abs_mean": 0.042506469041109086,
"signal/accgated_coverage_1/group_std_mean": 0.057106205448508265,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07912889569997787,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004250647011213004,
"signal/accgated_coverage_10/centered_abs_mean": 0.04226944372057915,
"signal/accgated_coverage_10/group_std_mean": 0.05678433850407601,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07866270318627358,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0042269445257261395,
"signal/accgated_coverage_15/centered_abs_mean": 0.02873615864664316,
"signal/accgated_coverage_15/group_std_mean": 0.03854703418910503,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05351117104291916,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.00287361575756222,
"signal/accgated_coverage_20/centered_abs_mean": 0.017688192985951902,
"signal/accgated_coverage_20/group_std_mean": 0.02309804279357195,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.033285561949014664,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0017688193591311574,
"signal/accgated_coverage_25/centered_abs_mean": 0.017029119841754436,
"signal/accgated_coverage_25/group_std_mean": 0.021728209033608437,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03228543736040592,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0017029120470397174,
"signal/accgated_coverage_5/centered_abs_mean": 0.04250315241515636,
"signal/accgated_coverage_5/group_std_mean": 0.0571018286049366,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07912281155586243,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0042503152275457975,
"signal/accuracy_reward/centered_abs_mean": 0.076116943359375,
"signal/accuracy_reward/group_std_mean": 0.1046619787812233,
"signal/accuracy_reward/group_zero_std_frac": 0.690625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6926234304904938,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0380584716796875,
"signal/advantage_abs_mean": 0.7796943783760071,
"signal/advantage_pre_scale_abs_mean": 0.05701693370938301,
"signal/advantage_pre_scale_std": 0.08956358879804611,
"signal/advantage_std": 0.9826038718223572,
"signal/brier_reward/centered_abs_mean": 0.09299083799123764,
"signal/brier_reward/group_std_mean": 0.12319995164871216,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17471539080142975,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009299084078520536,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.013640527613461018,
"signal/confidence_uniqueness_reward/group_std_mean": 0.018771480396389963,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.026319539174437523,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013640527613461018,
"signal/format_reward/centered_abs_mean": 0.00113525390625,
"signal/format_reward/group_std_mean": 0.0033145629800856113,
"signal/format_reward/group_zero_std_frac": 0.98125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.011480103805661201,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000567626953125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.298565137386322,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.37063016891479494,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5731562972068787,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02985651567578316,
"step": 245
},
{
"calibration/aurc": 0.17158607906792486,
"calibration/batch_distribution_entropy": 0.9481941043489982,
"calibration/batch_entropy_100bins": 0.9532288194354577,
"calibration/batch_entropy_10bins": 0.9481941043489982,
"calibration/batch_entropy_50bins": 0.9597535817989004,
"calibration/batch_uniqueness": 0.9485763609935025,
"calibration/buffer_distribution_entropy": 0.9906086356700137,
"calibration/buffer_entropy_100bins": 0.9948159049639553,
"calibration/buffer_entropy_10bins": 0.9906086356700137,
"calibration/buffer_entropy_50bins": 0.9941139291763476,
"calibration/confidence_entropy": 0.44870173660148194,
"calibration/coverage@0%": 0.18137842465753423,
"calibration/coverage@1%": 0.2103358916340509,
"calibration/coverage@10%": 0.442847969667319,
"calibration/coverage@15%": 0.5460004892367907,
"calibration/coverage@20%": 0.6069586289138943,
"calibration/coverage@25%": 0.6616682974559687,
"calibration/coverage@30%": 0.732788802592955,
"calibration/coverage@5%": 0.30374724804305286,
"calibration/distribution_entropy_10": 0.9481941043489982,
"calibration/distribution_entropy_100": 0.9532288194354577,
"calibration/ece": 0.1259540474702078,
"calibration/mean_confidence": 0.5011604939398695,
"calibration/unique_confidence_per_question": 0.996484375,
"calibration/unique_confidences": 510.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 574.2,
"completions/max_terminated_length": 574.2,
"completions/mean_length": 239.2634765625,
"completions/mean_terminated_length": 239.3091033935547,
"completions/min_length": 66.6,
"completions/min_terminated_length": 108.2,
"epoch": 0.8,
"grad_norm": 0.01166082825511694,
"learning_rate": 1e-06,
"loss": -0.0027,
"num_tokens": 860278448.0,
"reward": 0.9679004430770874,
"reward_std": 0.0692349448800087,
"rewards/accgated_coverage_0": 0.03927905634045601,
"rewards/accgated_coverage_1": 0.03927905634045601,
"rewards/accgated_coverage_10": 0.039093124121427535,
"rewards/accgated_coverage_15": 0.029265206307172775,
"rewards/accgated_coverage_20": 0.022285545244812965,
"rewards/accgated_coverage_25": 0.0278712198138237,
"rewards/accgated_coverage_5": 0.03927643671631813,
"rewards/accuracy_reward": 0.59169921875,
"rewards/brier_reward": 0.8310788750648499,
"rewards/confidence_uniqueness_reward": 0.9477739930152893,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.29371761679649355,
"signal/accgated_coverage_0/centered_abs_mean": 0.04012473821640015,
"signal/accgated_coverage_0/group_std_mean": 0.05344737246632576,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08500743508338929,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004012473858892918,
"signal/accgated_coverage_1/centered_abs_mean": 0.04012473821640015,
"signal/accgated_coverage_1/group_std_mean": 0.05344737246632576,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08500743508338929,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004012473858892918,
"signal/accgated_coverage_10/centered_abs_mean": 0.03984549641609192,
"signal/accgated_coverage_10/group_std_mean": 0.05309554561972618,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08441805690526963,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003984549688175321,
"signal/accgated_coverage_15/centered_abs_mean": 0.028042495995759965,
"signal/accgated_coverage_15/group_std_mean": 0.03761502355337143,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.059543855488300323,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.002804249618202448,
"signal/accgated_coverage_20/centered_abs_mean": 0.01801654864102602,
"signal/accgated_coverage_20/group_std_mean": 0.023348598182201384,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03828248009085655,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0018016548361629247,
"signal/accgated_coverage_25/centered_abs_mean": 0.019567286223173143,
"signal/accgated_coverage_25/group_std_mean": 0.024642554670572282,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.04154387414455414,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0019567286828532813,
"signal/accgated_coverage_5/centered_abs_mean": 0.04012225195765495,
"signal/accgated_coverage_5/group_std_mean": 0.05344438999891281,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08500203192234039,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.00401222500950098,
"signal/accuracy_reward/centered_abs_mean": 0.070562744140625,
"signal/accuracy_reward/group_std_mean": 0.09432597011327744,
"signal/accuracy_reward/group_zero_std_frac": 0.721875,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7422341227531433,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0352813720703125,
"signal/advantage_abs_mean": 0.7890103697776795,
"signal/advantage_pre_scale_abs_mean": 0.05488408431410789,
"signal/advantage_pre_scale_std": 0.08801894187927246,
"signal/advantage_std": 0.982406497001648,
"signal/brier_reward/centered_abs_mean": 0.08436839580535889,
"signal/brier_reward/group_std_mean": 0.11052148044109344,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.17804842293262482,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008436839748173951,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.013637678325176239,
"signal/confidence_uniqueness_reward/group_std_mean": 0.017684570327401162,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.028901634737849236,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013637678930535913,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.003997116722166539,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.29651297330856324,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.36457338333129885,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.627856433391571,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.029651298001408578,
"step": 250
},
{
"epoch": 0.8,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 411.5,
"eval_completions/max_terminated_length": 411.5,
"eval_completions/mean_length": 233.9325828552246,
"eval_completions/mean_terminated_length": 233.9325828552246,
"eval_completions/min_length": 127.0,
"eval_completions/min_terminated_length": 127.0,
"eval_loss": 0.0,
"eval_num_tokens": 860278448.0,
"eval_reward": 0.8073989003896713,
"eval_reward_std": 0.22174905985593796,
"eval_rewards/accgated_coverage_0": 0.037435878068208694,
"eval_rewards/accgated_coverage_1": 0.037435878068208694,
"eval_rewards/accgated_coverage_10": 0.03717117663472891,
"eval_rewards/accgated_coverage_15": 0.02733432548120618,
"eval_rewards/accgated_coverage_20": 0.015584731008857489,
"eval_rewards/accgated_coverage_25": 0.011484242277219892,
"eval_rewards/accgated_coverage_5": 0.03742485959082842,
"eval_rewards/accuracy_reward": 0.431640625,
"eval_rewards/brier_reward": 0.8120125830173492,
"eval_rewards/confidence_uniqueness_reward": 0.89990234375,
"eval_rewards/format_reward": 1.0,
"eval_rewards/frontier_entropy_batch_reward": -1.0,
"eval_runtime": 21.6991,
"eval_samples_per_second": 23.042,
"eval_signal/accgated_coverage_0/centered_abs_mean": 0.07356817275285721,
"eval_signal/accgated_coverage_0/group_std_mean": 0.08836030773818493,
"eval_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.03338225232437253,
"eval_signal/accgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.007356817368417978,
"eval_signal/accgated_coverage_1/centered_abs_mean": 0.07356817275285721,
"eval_signal/accgated_coverage_1/group_std_mean": 0.08836030773818493,
"eval_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.03338225232437253,
"eval_signal/accgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.007356817368417978,
"eval_signal/accgated_coverage_10/centered_abs_mean": 0.07308768387883902,
"eval_signal/accgated_coverage_10/group_std_mean": 0.08781414665281773,
"eval_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.033165073953568935,
"eval_signal/accgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.007308768457733095,
"eval_signal/accgated_coverage_15/centered_abs_mean": 0.05316424090415239,
"eval_signal/accgated_coverage_15/group_std_mean": 0.06487995572388172,
"eval_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.02413960499688983,
"eval_signal/accgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.005316423950716853,
"eval_signal/accgated_coverage_20/centered_abs_mean": 0.023750313092023134,
"eval_signal/accgated_coverage_20/group_std_mean": 0.030248934403061867,
"eval_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.010784113546833396,
"eval_signal/accgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0023750313266646117,
"eval_signal/accgated_coverage_25/centered_abs_mean": 0.016915465239435434,
"eval_signal/accgated_coverage_25/group_std_mean": 0.021362427156418562,
"eval_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.007680113427340984,
"eval_signal/accgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0016915465239435434,
"eval_signal/accgated_coverage_5/centered_abs_mean": 0.0735478987917304,
"eval_signal/accgated_coverage_5/group_std_mean": 0.08833653852343559,
"eval_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.03337304340675473,
"eval_signal/accgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.007354790228419006,
"eval_signal/accuracy_reward/centered_abs_mean": 0.4737548828125,
"eval_signal/accuracy_reward/group_std_mean": 0.4942636787891388,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.0715261101722717,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.23687744140625,
"eval_signal/advantage_abs_mean": 0.9273225963115692,
"eval_signal/advantage_pre_scale_abs_mean": 0.2058747410774231,
"eval_signal/advantage_pre_scale_std": 0.21947148814797401,
"eval_signal/advantage_std": 0.9876712262630463,
"eval_signal/brier_reward/centered_abs_mean": 0.18244916945695877,
"eval_signal/brier_reward/group_std_mean": 0.2339242845773697,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.0826017502695322,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.018244916573166847,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.0406341552734375,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.047885650768876076,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.018425464630126953,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004063415399286896,
"eval_signal/format_reward/centered_abs_mean": 0.0,
"eval_signal/format_reward/group_std_mean": 0.0,
"eval_signal/format_reward/group_zero_std_frac": 1.0,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"eval_steps_per_second": 0.184,
"step": 250
},
{
"epoch": 0.8,
"step": 250,
"train_probe_completions/clipped_ratio": 0.0,
"train_probe_completions/max_length": 420.25,
"train_probe_completions/max_terminated_length": 420.25,
"train_probe_completions/mean_length": 234.53279876708984,
"train_probe_completions/mean_terminated_length": 234.53279876708984,
"train_probe_completions/min_length": 138.0,
"train_probe_completions/min_terminated_length": 138.0,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 860278448.0,
"train_probe_reward": 0.895091250538826,
"train_probe_reward_std": 0.19868780300021172,
"train_probe_rewards/accgated_coverage_0": 0.03365300549194217,
"train_probe_rewards/accgated_coverage_1": 0.03365300549194217,
"train_probe_rewards/accgated_coverage_10": 0.03356103505939245,
"train_probe_rewards/accgated_coverage_15": 0.028990946244448423,
"train_probe_rewards/accgated_coverage_20": 0.02319516707211733,
"train_probe_rewards/accgated_coverage_25": 0.028358858078718185,
"train_probe_rewards/accgated_coverage_5": 0.033650572411715984,
"train_probe_rewards/accuracy_reward": 0.603515625,
"train_probe_rewards/brier_reward": 0.8244728147983551,
"train_probe_rewards/confidence_uniqueness_reward": 0.893798828125,
"train_probe_rewards/format_reward": 1.0,
"train_probe_rewards/frontier_entropy_batch_reward": -1.0,
"train_probe_runtime": 21.3569,
"train_probe_samples_per_second": 23.412,
"train_probe_signal/accgated_coverage_0/centered_abs_mean": 0.12618132680654526,
"train_probe_signal/accgated_coverage_0/group_std_mean": 0.16332727298140526,
"train_probe_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.06476442236453295,
"train_probe_signal/accgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.012618132401257753,
"train_probe_signal/accgated_coverage_1/centered_abs_mean": 0.12618132680654526,
"train_probe_signal/accgated_coverage_1/group_std_mean": 0.16332727298140526,
"train_probe_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.06476442236453295,
"train_probe_signal/accgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.012618132401257753,
"train_probe_signal/accgated_coverage_10/centered_abs_mean": 0.12497550994157791,
"train_probe_signal/accgated_coverage_10/group_std_mean": 0.1618593893945217,
"train_probe_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06414960976690054,
"train_probe_signal/accgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.012497550807893276,
"train_probe_signal/accgated_coverage_15/centered_abs_mean": 0.08380677178502083,
"train_probe_signal/accgated_coverage_15/group_std_mean": 0.11126936413347721,
"train_probe_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.04305774066597223,
"train_probe_signal/accgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.008380677318200469,
"train_probe_signal/accgated_coverage_20/centered_abs_mean": 0.04052241565659642,
"train_probe_signal/accgated_coverage_20/group_std_mean": 0.053040504455566406,
"train_probe_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.02083801245316863,
"train_probe_signal/accgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.004052241565659642,
"train_probe_signal/accgated_coverage_25/centered_abs_mean": 0.03542446158826351,
"train_probe_signal/accgated_coverage_25/group_std_mean": 0.045083730481565,
"train_probe_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.01815139828249812,
"train_probe_signal/accgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0035424461821094155,
"train_probe_signal/accgated_coverage_5/centered_abs_mean": 0.12614190205931664,
"train_probe_signal/accgated_coverage_5/group_std_mean": 0.16327836737036705,
"train_probe_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06474409252405167,
"train_probe_signal/accgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.012614190811291337,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.4598388671875,
"train_probe_signal/accuracy_reward/group_std_mean": 0.4867246150970459,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.1634121239185333,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.22991943359375,
"train_probe_signal/advantage_abs_mean": 0.8950022608041763,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.1786213181912899,
"train_probe_signal/advantage_pre_scale_std": 0.19667648151516914,
"train_probe_signal/advantage_std": 0.987617090344429,
"train_probe_signal/brier_reward/centered_abs_mean": 0.17699800431728363,
"train_probe_signal/brier_reward/group_std_mean": 0.2345779426395893,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.08973932452499866,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.017699801363050938,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.039459228515625,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.04700522497296333,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02006423007696867,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0039459228864870965,
"train_probe_signal/format_reward/centered_abs_mean": 0.0,
"train_probe_signal/format_reward/group_std_mean": 0.0,
"train_probe_signal/format_reward/group_zero_std_frac": 1.0,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"train_probe_steps_per_second": 0.187
},
{
"calibration/aurc": 0.19669236415494645,
"calibration/batch_distribution_entropy": 0.9293557273743327,
"calibration/batch_entropy_100bins": 0.9426278337010487,
"calibration/batch_entropy_10bins": 0.9293557273743327,
"calibration/batch_entropy_50bins": 0.9460355451615525,
"calibration/batch_uniqueness": 0.9431339051312321,
"calibration/buffer_distribution_entropy": 0.990752630231395,
"calibration/buffer_entropy_100bins": 0.9948754982489417,
"calibration/buffer_entropy_10bins": 0.990752630231395,
"calibration/buffer_entropy_50bins": 0.9941852174510581,
"calibration/confidence_entropy": 0.44474018505780144,
"calibration/coverage@0%": 0.026194043542074365,
"calibration/coverage@1%": 0.026194043542074365,
"calibration/coverage@10%": 0.28303647871819965,
"calibration/coverage@15%": 0.4092771526418787,
"calibration/coverage@20%": 0.6315848214285714,
"calibration/coverage@25%": 0.7034842832681018,
"calibration/coverage@30%": 0.8335914567025441,
"calibration/coverage@5%": 0.14388530454990214,
"calibration/distribution_entropy_10": 0.9293557273743327,
"calibration/distribution_entropy_100": 0.9426278337010487,
"calibration/ece": 0.14496823728398528,
"calibration/mean_confidence": 0.5126159137012547,
"calibration/unique_confidence_per_question": 0.996875,
"calibration/unique_confidences": 510.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 609.0,
"completions/max_terminated_length": 609.0,
"completions/mean_length": 232.4162109375,
"completions/mean_terminated_length": 232.46085510253906,
"completions/min_length": 66.8,
"completions/min_terminated_length": 109.0,
"epoch": 0.816,
"grad_norm": 0.0076979706063866615,
"learning_rate": 1e-06,
"loss": -0.0027,
"num_tokens": 877757558.0,
"reward": 0.9612357497215271,
"reward_std": 0.0741512656211853,
"rewards/accgated_coverage_0": 0.035010688751935956,
"rewards/accgated_coverage_1": 0.035010688751935956,
"rewards/accgated_coverage_10": 0.034884200990200044,
"rewards/accgated_coverage_15": 0.0289004173129797,
"rewards/accgated_coverage_20": 0.021493689343333244,
"rewards/accgated_coverage_25": 0.024621672928333282,
"rewards/accgated_coverage_5": 0.03501078113913536,
"rewards/accuracy_reward": 0.5759765625,
"rewards/brier_reward": 0.8196485042572021,
"rewards/confidence_uniqueness_reward": 0.9491020560264587,
"rewards/format_reward": 0.99951171875,
"rewards/frontier_entropy_batch_reward": -0.24876676201820375,
"signal/accgated_coverage_0/centered_abs_mean": 0.04417483434081078,
"signal/accgated_coverage_0/group_std_mean": 0.05687285587191582,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0901245966553688,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004417483601719141,
"signal/accgated_coverage_1/centered_abs_mean": 0.04417483434081078,
"signal/accgated_coverage_1/group_std_mean": 0.05687285587191582,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0901245966553688,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004417483601719141,
"signal/accgated_coverage_10/centered_abs_mean": 0.04383779391646385,
"signal/accgated_coverage_10/group_std_mean": 0.05643892139196396,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08943924754858017,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004383779456838965,
"signal/accgated_coverage_15/centered_abs_mean": 0.029114584624767303,
"signal/accgated_coverage_15/group_std_mean": 0.03757238090038299,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.059433210641145706,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0029114584904164076,
"signal/accgated_coverage_20/centered_abs_mean": 0.018224692344665526,
"signal/accgated_coverage_20/group_std_mean": 0.023159447684884073,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03722381070256233,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0018224692903459072,
"signal/accgated_coverage_25/centered_abs_mean": 0.01875607669353485,
"signal/accgated_coverage_25/group_std_mean": 0.023673447966575622,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03820987939834595,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0018756077392026783,
"signal/accgated_coverage_5/centered_abs_mean": 0.04415859580039978,
"signal/accgated_coverage_5/group_std_mean": 0.05685246363282204,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.0900917574763298,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004415859654545784,
"signal/accuracy_reward/centered_abs_mean": 0.0822998046875,
"signal/accuracy_reward/group_std_mean": 0.11028514206409454,
"signal/accuracy_reward/group_zero_std_frac": 0.678125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.837281334400177,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.04114990234375,
"signal/advantage_abs_mean": 0.7725938320159912,
"signal/advantage_pre_scale_abs_mean": 0.056962195038795474,
"signal/advantage_pre_scale_std": 0.09214915037155151,
"signal/advantage_std": 0.9824849009513855,
"signal/brier_reward/centered_abs_mean": 0.09328299462795257,
"signal/brier_reward/group_std_mean": 0.12126787453889847,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19003214538097382,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009328299760818481,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.013943759351968765,
"signal/confidence_uniqueness_reward/group_std_mean": 0.01892264299094677,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.028514381870627402,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001394375949166715,
"signal/format_reward/centered_abs_mean": 0.000946044921875,
"signal/format_reward/group_std_mean": 0.0027621358167380095,
"signal/format_reward/group_zero_std_frac": 0.984375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.009643366187810897,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0004730224609375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2811394721269608,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3532793700695038,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5730468273162842,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028113946691155432,
"step": 255
},
{
"calibration/aurc": 0.25976579962016283,
"calibration/batch_distribution_entropy": 0.9647766678355831,
"calibration/batch_entropy_100bins": 0.9614017294190125,
"calibration/batch_entropy_10bins": 0.9647766678355831,
"calibration/batch_entropy_50bins": 0.9679454638202383,
"calibration/batch_uniqueness": 0.9500096634770546,
"calibration/buffer_distribution_entropy": 0.9897827405105482,
"calibration/buffer_entropy_100bins": 0.9943560503138507,
"calibration/buffer_entropy_10bins": 0.9897827405105482,
"calibration/buffer_entropy_50bins": 0.993598669820295,
"calibration/confidence_entropy": 0.48892402531686197,
"calibration/coverage@0%": 0.10710004892367905,
"calibration/coverage@1%": 0.12235353473581215,
"calibration/coverage@10%": 0.27214331580234835,
"calibration/coverage@15%": 0.33037013820939337,
"calibration/coverage@20%": 0.41131818737769077,
"calibration/coverage@25%": 0.46448752446183955,
"calibration/coverage@30%": 0.531363900440313,
"calibration/coverage@5%": 0.22016343566536203,
"calibration/distribution_entropy_10": 0.9647766678355831,
"calibration/distribution_entropy_100": 0.9614017294190125,
"calibration/ece": 0.1208211307724828,
"calibration/mean_confidence": 0.48189860114884225,
"calibration/unique_confidence_per_question": 0.994140625,
"calibration/unique_confidences": 509.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 737.8,
"completions/max_terminated_length": 737.8,
"completions/mean_length": 248.1353515625,
"completions/mean_terminated_length": 248.16002502441407,
"completions/min_length": 89.4,
"completions/min_terminated_length": 115.0,
"epoch": 0.832,
"grad_norm": 0.006655729841440916,
"learning_rate": 1e-06,
"loss": 0.0002,
"num_tokens": 895306816.0,
"reward": 0.9543121933937073,
"reward_std": 0.07087772786617279,
"rewards/accgated_coverage_0": 0.03877972103655338,
"rewards/accgated_coverage_1": 0.03877972103655338,
"rewards/accgated_coverage_10": 0.038498851284384725,
"rewards/accgated_coverage_15": 0.029690840840339662,
"rewards/accgated_coverage_20": 0.021420810744166373,
"rewards/accgated_coverage_25": 0.025324767082929613,
"rewards/accgated_coverage_5": 0.03877520188689232,
"rewards/accuracy_reward": 0.55634765625,
"rewards/brier_reward": 0.8314706444740295,
"rewards/confidence_uniqueness_reward": 0.9494145393371582,
"rewards/format_reward": 0.99970703125,
"rewards/frontier_entropy_batch_reward": -0.24930644035339355,
"signal/accgated_coverage_0/centered_abs_mean": 0.03758116811513901,
"signal/accgated_coverage_0/group_std_mean": 0.04910837560892105,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07910573855042458,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0037581168580800295,
"signal/accgated_coverage_1/centered_abs_mean": 0.03758116811513901,
"signal/accgated_coverage_1/group_std_mean": 0.04910837560892105,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07910573855042458,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0037581168580800295,
"signal/accgated_coverage_10/centered_abs_mean": 0.03725464642047882,
"signal/accgated_coverage_10/group_std_mean": 0.04869186207652092,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07842497527599335,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0037254647817462684,
"signal/accgated_coverage_15/centered_abs_mean": 0.025605784356594087,
"signal/accgated_coverage_15/group_std_mean": 0.033432379364967346,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.053861670196056366,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0025605784263461827,
"signal/accgated_coverage_20/centered_abs_mean": 0.016141471080482007,
"signal/accgated_coverage_20/group_std_mean": 0.02069954704493284,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03409386649727821,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0016141470987349749,
"signal/accgated_coverage_25/centered_abs_mean": 0.016882821545004846,
"signal/accgated_coverage_25/group_std_mean": 0.02147441916167736,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03555565439164639,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0016882821451872586,
"signal/accgated_coverage_5/centered_abs_mean": 0.03756747543811798,
"signal/accgated_coverage_5/group_std_mean": 0.04909084066748619,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07907658964395523,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0037567476276308296,
"signal/accuracy_reward/centered_abs_mean": 0.071954345703125,
"signal/accuracy_reward/group_std_mean": 0.09801256954669953,
"signal/accuracy_reward/group_zero_std_frac": 0.709375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7553323626518249,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0359771728515625,
"signal/advantage_abs_mean": 0.771164846420288,
"signal/advantage_pre_scale_abs_mean": 0.05432874709367752,
"signal/advantage_pre_scale_std": 0.08939148485660553,
"signal/advantage_std": 0.9824316620826721,
"signal/brier_reward/centered_abs_mean": 0.08345125317573547,
"signal/brier_reward/group_std_mean": 0.10945889204740525,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1743919849395752,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008345125615596772,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01370535809546709,
"signal/confidence_uniqueness_reward/group_std_mean": 0.017885472998023032,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.028743793815374376,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001370535814203322,
"signal/format_reward/centered_abs_mean": 0.000567626953125,
"signal/format_reward/group_std_mean": 0.0016572814900428056,
"signal/format_reward/group_zero_std_frac": 0.990625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.005596400983631611,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0002838134765625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2834340214729309,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3541470766067505,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5911893963813781,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028343402594327927,
"step": 260
},
{
"calibration/aurc": 0.2750655015117763,
"calibration/batch_distribution_entropy": 0.9562382509199573,
"calibration/batch_entropy_100bins": 0.9590383083474707,
"calibration/batch_entropy_10bins": 0.9562382509199573,
"calibration/batch_entropy_50bins": 0.9646852909283483,
"calibration/batch_uniqueness": 0.949761962890625,
"calibration/buffer_distribution_entropy": 0.9902782682484583,
"calibration/buffer_entropy_100bins": 0.9945874719131824,
"calibration/buffer_entropy_10bins": 0.9902782682484583,
"calibration/buffer_entropy_50bins": 0.9938715393780704,
"calibration/confidence_entropy": 0.4638141897117471,
"calibration/coverage@0%": 0.13984375,
"calibration/coverage@1%": 0.15546875,
"calibration/coverage@10%": 0.221484375,
"calibration/coverage@15%": 0.269140625,
"calibration/coverage@20%": 0.465234375,
"calibration/coverage@25%": 0.528515625,
"calibration/coverage@30%": 0.580078125,
"calibration/coverage@5%": 0.194921875,
"calibration/distribution_entropy_10": 0.9562382509199573,
"calibration/distribution_entropy_100": 0.9590383083474707,
"calibration/ece": 0.14915820264775856,
"calibration/mean_confidence": 0.5373101235402765,
"calibration/unique_confidence_per_question": 0.997265625,
"calibration/unique_confidences": 510.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 715.4,
"completions/max_terminated_length": 715.4,
"completions/mean_length": 260.64208984375,
"completions/mean_terminated_length": 260.69431762695314,
"completions/min_length": 67.8,
"completions/min_terminated_length": 110.6,
"epoch": 0.848,
"grad_norm": 0.00669163977727294,
"learning_rate": 1e-06,
"loss": 0.0013,
"num_tokens": 912990159.0,
"reward": 0.9441014170646668,
"reward_std": 0.06691018417477608,
"rewards/accgated_coverage_0": 0.03886624053120613,
"rewards/accgated_coverage_1": 0.03886624053120613,
"rewards/accgated_coverage_10": 0.03872780688107014,
"rewards/accgated_coverage_15": 0.028617511689662933,
"rewards/accgated_coverage_20": 0.01971310339868069,
"rewards/accgated_coverage_25": 0.021255484595894814,
"rewards/accgated_coverage_5": 0.038853126019239424,
"rewards/accuracy_reward": 0.53466796875,
"rewards/brier_reward": 0.8316412091255188,
"rewards/confidence_uniqueness_reward": 0.9501094579696655,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.23799911737442017,
"signal/accgated_coverage_0/centered_abs_mean": 0.03503080531954765,
"signal/accgated_coverage_0/group_std_mean": 0.045668549835681915,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07662608250975608,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.003503080550581217,
"signal/accgated_coverage_1/centered_abs_mean": 0.03503080531954765,
"signal/accgated_coverage_1/group_std_mean": 0.045668549835681915,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07662608250975608,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.003503080550581217,
"signal/accgated_coverage_10/centered_abs_mean": 0.03483315706253052,
"signal/accgated_coverage_10/group_std_mean": 0.045417507737874986,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07618726566433906,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003483315883204341,
"signal/accgated_coverage_15/centered_abs_mean": 0.024957020208239554,
"signal/accgated_coverage_15/group_std_mean": 0.03236778676509857,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05451673343777656,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0024957021232694387,
"signal/accgated_coverage_20/centered_abs_mean": 0.01495315246284008,
"signal/accgated_coverage_20/group_std_mean": 0.019112042710185052,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03276476785540581,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.001495315320789814,
"signal/accgated_coverage_25/centered_abs_mean": 0.015340249985456467,
"signal/accgated_coverage_25/group_std_mean": 0.019527608156204225,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03356584683060646,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0015340251149609685,
"signal/accgated_coverage_5/centered_abs_mean": 0.03502167388796806,
"signal/accgated_coverage_5/group_std_mean": 0.04565635621547699,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07660583704710007,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.003502167575061321,
"signal/accuracy_reward/centered_abs_mean": 0.065289306640625,
"signal/accuracy_reward/group_std_mean": 0.0903944805264473,
"signal/accuracy_reward/group_zero_std_frac": 0.728125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7075724482536316,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0326446533203125,
"signal/advantage_abs_mean": 0.7694258213043212,
"signal/advantage_pre_scale_abs_mean": 0.051307087391614915,
"signal/advantage_pre_scale_std": 0.08402346521615982,
"signal/advantage_std": 0.9823080420494079,
"signal/brier_reward/centered_abs_mean": 0.08195143938064575,
"signal/brier_reward/group_std_mean": 0.10823454260826111,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1809752881526947,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.00819514412432909,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.012950626388192176,
"signal/confidence_uniqueness_reward/group_std_mean": 0.016688339039683343,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02874571979045868,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012950626667588949,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.004131428152322769,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.27336313724517824,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3424546182155609,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.6065775632858277,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02733631506562233,
"step": 265
},
{
"calibration/aurc": 0.25395768420987397,
"calibration/batch_distribution_entropy": 0.9540329990648997,
"calibration/batch_entropy_100bins": 0.9559437727780715,
"calibration/batch_entropy_10bins": 0.9540329990648997,
"calibration/batch_entropy_50bins": 0.9622735843235143,
"calibration/batch_uniqueness": 0.9488739013671875,
"calibration/buffer_distribution_entropy": 0.9915750054109621,
"calibration/buffer_entropy_100bins": 0.9952719450497257,
"calibration/buffer_entropy_10bins": 0.9915750054109621,
"calibration/buffer_entropy_50bins": 0.994671015022415,
"calibration/confidence_entropy": 0.4723572430250617,
"calibration/coverage@0%": 0.0359375,
"calibration/coverage@1%": 0.0359375,
"calibration/coverage@10%": 0.178125,
"calibration/coverage@15%": 0.245703125,
"calibration/coverage@20%": 0.403515625,
"calibration/coverage@25%": 0.465234375,
"calibration/coverage@30%": 0.712109375,
"calibration/coverage@5%": 0.115625,
"calibration/distribution_entropy_10": 0.9540329990648997,
"calibration/distribution_entropy_100": 0.9559437727780715,
"calibration/ece": 0.14175661914947257,
"calibration/mean_confidence": 0.5905959991293376,
"calibration/unique_confidence_per_question": 0.996484375,
"calibration/unique_confidences": 510.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 713.8,
"completions/max_terminated_length": 713.8,
"completions/mean_length": 263.1501953125,
"completions/mean_terminated_length": 263.2016174316406,
"completions/min_length": 81.2,
"completions/min_terminated_length": 126.8,
"epoch": 0.864,
"grad_norm": 0.00721458625048399,
"learning_rate": 1e-06,
"loss": -0.0027,
"num_tokens": 930671633.0,
"reward": 0.966060483455658,
"reward_std": 0.06951428726315498,
"rewards/accgated_coverage_0": 0.028997783735394477,
"rewards/accgated_coverage_1": 0.028997783735394477,
"rewards/accgated_coverage_10": 0.028951923549175262,
"rewards/accgated_coverage_15": 0.024892372637987138,
"rewards/accgated_coverage_20": 0.020579223334789277,
"rewards/accgated_coverage_25": 0.026464609429240227,
"rewards/accgated_coverage_5": 0.028995355218648912,
"rewards/accuracy_reward": 0.59462890625,
"rewards/brier_reward": 0.81304532289505,
"rewards/confidence_uniqueness_reward": 0.9487526774406433,
"rewards/format_reward": 0.9998046875,
"rewards/frontier_entropy_batch_reward": -0.26124017834663393,
"signal/accgated_coverage_0/centered_abs_mean": 0.046113957464694974,
"signal/accgated_coverage_0/group_std_mean": 0.0601910762488842,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.09388052672147751,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004611395951360464,
"signal/accgated_coverage_1/centered_abs_mean": 0.046113957464694974,
"signal/accgated_coverage_1/group_std_mean": 0.0601910762488842,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.09388052672147751,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004611395951360464,
"signal/accgated_coverage_10/centered_abs_mean": 0.04555861875414848,
"signal/accgated_coverage_10/group_std_mean": 0.059482603520154956,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.09275497049093247,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004555861884728074,
"signal/accgated_coverage_15/centered_abs_mean": 0.030330605059862136,
"signal/accgated_coverage_15/group_std_mean": 0.039680345356464385,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06174131035804749,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003033060533925891,
"signal/accgated_coverage_20/centered_abs_mean": 0.01887478232383728,
"signal/accgated_coverage_20/group_std_mean": 0.024249549955129623,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.038462989032268524,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0018874783534556627,
"signal/accgated_coverage_25/centered_abs_mean": 0.020123864337801934,
"signal/accgated_coverage_25/group_std_mean": 0.02544081285595894,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.04106858782470226,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.002012386452406645,
"signal/accgated_coverage_5/centered_abs_mean": 0.046077194064855574,
"signal/accgated_coverage_5/group_std_mean": 0.060144300758838656,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.09380602091550827,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0046077192761003975,
"signal/accuracy_reward/centered_abs_mean": 0.072625732421875,
"signal/accuracy_reward/group_std_mean": 0.09984676092863083,
"signal/accuracy_reward/group_zero_std_frac": 0.703125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7361932516098022,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0363128662109375,
"signal/advantage_abs_mean": 0.7788555741310119,
"signal/advantage_pre_scale_abs_mean": 0.05345872864127159,
"signal/advantage_pre_scale_std": 0.08519237637519836,
"signal/advantage_std": 0.9824930191040039,
"signal/brier_reward/centered_abs_mean": 0.09002687335014344,
"signal/brier_reward/group_std_mean": 0.11738848984241486,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18250847160816192,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009002687036991119,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.013763246685266494,
"signal/confidence_uniqueness_reward/group_std_mean": 0.017733591236174108,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02789420336484909,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001376324682496488,
"signal/format_reward/centered_abs_mean": 0.00037841796875,
"signal/format_reward/group_std_mean": 0.0011048543266952038,
"signal/format_reward/group_zero_std_frac": 0.99375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.0036525866016745567,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000189208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28499422669410707,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3584535658359528,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.578595507144928,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02849942184984684,
"step": 270
},
{
"calibration/aurc": 0.35463664527273814,
"calibration/batch_distribution_entropy": 0.9756094231968394,
"calibration/batch_entropy_100bins": 0.9678146898865461,
"calibration/batch_entropy_10bins": 0.9756094231968394,
"calibration/batch_entropy_50bins": 0.9758044038621028,
"calibration/batch_uniqueness": 0.952691650390625,
"calibration/buffer_distribution_entropy": 0.9918414732318643,
"calibration/buffer_entropy_100bins": 0.9954415638877917,
"calibration/buffer_entropy_10bins": 0.9918414732318643,
"calibration/buffer_entropy_50bins": 0.9948503689431911,
"calibration/confidence_entropy": 0.4500348146102168,
"calibration/coverage@0%": 0.008984375,
"calibration/coverage@1%": 0.008984375,
"calibration/coverage@10%": 0.024609375,
"calibration/coverage@15%": 0.06796875,
"calibration/coverage@20%": 0.16484375,
"calibration/coverage@25%": 0.316015625,
"calibration/coverage@30%": 0.44140625,
"calibration/coverage@5%": 0.008984375,
"calibration/distribution_entropy_10": 0.9756094231968394,
"calibration/distribution_entropy_100": 0.9678146898865461,
"calibration/ece": 0.1535016748514101,
"calibration/mean_confidence": 0.5189274415080511,
"calibration/unique_confidence_per_question": 0.99921875,
"calibration/unique_confidences": 511.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00068359375,
"completions/max_length": 609.2,
"completions/max_terminated_length": 609.2,
"completions/mean_length": 258.9390625,
"completions/mean_terminated_length": 259.1177551269531,
"completions/min_length": 21.6,
"completions/min_terminated_length": 109.4,
"epoch": 0.88,
"grad_norm": 0.006898901890963316,
"learning_rate": 1e-06,
"loss": -0.0054,
"num_tokens": 948470241.0,
"reward": 0.9347312092781067,
"reward_std": 0.07466730326414109,
"rewards/accgated_coverage_0": 0.034718108177185056,
"rewards/accgated_coverage_1": 0.034718108177185056,
"rewards/accgated_coverage_10": 0.03438863828778267,
"rewards/accgated_coverage_15": 0.025535573810338975,
"rewards/accgated_coverage_20": 0.01815590523183346,
"rewards/accgated_coverage_25": 0.021791455894708635,
"rewards/accgated_coverage_5": 0.0346952348947525,
"rewards/accuracy_reward": 0.52548828125,
"rewards/brier_reward": 0.8171032190322876,
"rewards/confidence_uniqueness_reward": 0.9488176584243775,
"rewards/format_reward": 0.99931640625,
"rewards/frontier_entropy_batch_reward": -0.24663541913032533,
"signal/accgated_coverage_0/centered_abs_mean": 0.03735544756054878,
"signal/accgated_coverage_0/group_std_mean": 0.04825745970010757,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07862244322896003,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0037355448585003614,
"signal/accgated_coverage_1/centered_abs_mean": 0.03735544756054878,
"signal/accgated_coverage_1/group_std_mean": 0.04825745970010757,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07862244322896003,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0037355448585003614,
"signal/accgated_coverage_10/centered_abs_mean": 0.03692154288291931,
"signal/accgated_coverage_10/group_std_mean": 0.047697975486516955,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07770741656422615,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.00369215440005064,
"signal/accgated_coverage_15/centered_abs_mean": 0.025861211493611334,
"signal/accgated_coverage_15/group_std_mean": 0.03348269909620285,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05441740080714226,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0025861212983727456,
"signal/accgated_coverage_20/centered_abs_mean": 0.015574687719345092,
"signal/accgated_coverage_20/group_std_mean": 0.019897479191422464,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03283597081899643,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0015574687160551548,
"signal/accgated_coverage_25/centered_abs_mean": 0.016172933951020242,
"signal/accgated_coverage_25/group_std_mean": 0.020602439157664777,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03414270132780075,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0016172933857887984,
"signal/accgated_coverage_5/centered_abs_mean": 0.03732409104704857,
"signal/accgated_coverage_5/group_std_mean": 0.04821686372160912,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07855672612786294,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0037324091885238887,
"signal/accuracy_reward/centered_abs_mean": 0.083770751953125,
"signal/accuracy_reward/group_std_mean": 0.1117611512541771,
"signal/accuracy_reward/group_zero_std_frac": 0.675,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.8866649389266967,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0418853759765625,
"signal/advantage_abs_mean": 0.7670110106468201,
"signal/advantage_pre_scale_abs_mean": 0.05729171857237816,
"signal/advantage_pre_scale_std": 0.09516999870538712,
"signal/advantage_std": 0.9824028611183167,
"signal/brier_reward/centered_abs_mean": 0.09362064898014069,
"signal/brier_reward/group_std_mean": 0.12397909313440322,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19862198531627656,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009362064860761165,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.014180111698806286,
"signal/confidence_uniqueness_reward/group_std_mean": 0.019144942983984947,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.030146772041916847,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001418011193163693,
"signal/format_reward/centered_abs_mean": 0.001287841796875,
"signal/format_reward/group_std_mean": 0.0031351604498922824,
"signal/format_reward/group_zero_std_frac": 0.984375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.013877593167126178,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0006439208984375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2651355504989624,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3373463451862335,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5628416895866394,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026513554528355597,
"step": 275
},
{
"calibration/aurc": 0.3375147710713128,
"calibration/batch_distribution_entropy": 0.9799140151845895,
"calibration/batch_entropy_100bins": 0.968195441146262,
"calibration/batch_entropy_10bins": 0.9799140151845895,
"calibration/batch_entropy_50bins": 0.9764960021474703,
"calibration/batch_uniqueness": 0.9531788541447075,
"calibration/buffer_distribution_entropy": 0.9913242468707061,
"calibration/buffer_entropy_100bins": 0.995159034297236,
"calibration/buffer_entropy_10bins": 0.9913242468707061,
"calibration/buffer_entropy_50bins": 0.9945321552122197,
"calibration/confidence_entropy": 0.47799451054612563,
"calibration/coverage@0%": 0.008203889432485323,
"calibration/coverage@1%": 0.008203889432485323,
"calibration/coverage@10%": 0.06796951443248532,
"calibration/coverage@15%": 0.11562576443248532,
"calibration/coverage@20%": 0.2332107693248532,
"calibration/coverage@25%": 0.40625764432485323,
"calibration/coverage@30%": 0.5176117600293543,
"calibration/coverage@5%": 0.02265701443248532,
"calibration/distribution_entropy_10": 0.9799140151845895,
"calibration/distribution_entropy_100": 0.968195441146262,
"calibration/ece": 0.1378294590076745,
"calibration/mean_confidence": 0.5229784787870747,
"calibration/unique_confidence_per_question": 0.998046875,
"calibration/unique_confidences": 511.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0005859375,
"completions/max_length": 630.2,
"completions/max_terminated_length": 630.2,
"completions/mean_length": 257.19365234375,
"completions/mean_terminated_length": 257.3445220947266,
"completions/min_length": 26.6,
"completions/min_terminated_length": 125.0,
"epoch": 0.896,
"grad_norm": 0.007914412766695023,
"learning_rate": 1e-06,
"loss": -0.0003,
"num_tokens": 966214752.0,
"reward": 0.9528495907783509,
"reward_std": 0.06934682950377465,
"rewards/accgated_coverage_0": 0.023688822239637374,
"rewards/accgated_coverage_1": 0.023688822239637374,
"rewards/accgated_coverage_10": 0.02374329250305891,
"rewards/accgated_coverage_15": 0.019953744672238826,
"rewards/accgated_coverage_20": 0.017547025717794894,
"rewards/accgated_coverage_25": 0.025568272359669207,
"rewards/accgated_coverage_5": 0.023703657649457456,
"rewards/accuracy_reward": 0.571875,
"rewards/brier_reward": 0.8153186202049255,
"rewards/confidence_uniqueness_reward": 0.9498520612716674,
"rewards/format_reward": 0.9994140625,
"rewards/frontier_entropy_batch_reward": -0.25101393461227417,
"signal/accgated_coverage_0/centered_abs_mean": 0.041368260979652405,
"signal/accgated_coverage_0/group_std_mean": 0.05282168909907341,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08526693433523178,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004136826237663626,
"signal/accgated_coverage_1/centered_abs_mean": 0.041368260979652405,
"signal/accgated_coverage_1/group_std_mean": 0.05282168909907341,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08526693433523178,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004136826237663626,
"signal/accgated_coverage_10/centered_abs_mean": 0.0407241303473711,
"signal/accgated_coverage_10/group_std_mean": 0.05204328745603561,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08393167555332184,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004072412941604852,
"signal/accgated_coverage_15/centered_abs_mean": 0.02837865948677063,
"signal/accgated_coverage_15/group_std_mean": 0.03663924932479858,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05862174928188324,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.002837865985929966,
"signal/accgated_coverage_20/centered_abs_mean": 0.01716951485723257,
"signal/accgated_coverage_20/group_std_mean": 0.021865909919142723,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.035478471964597705,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0017169515136629342,
"signal/accgated_coverage_25/centered_abs_mean": 0.018951852433383466,
"signal/accgated_coverage_25/group_std_mean": 0.023946811631321906,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.039357070624828336,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0018951852805912494,
"signal/accgated_coverage_5/centered_abs_mean": 0.04132920615375042,
"signal/accgated_coverage_5/group_std_mean": 0.05277537032961845,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08518491759896278,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004132920596748591,
"signal/accuracy_reward/centered_abs_mean": 0.06832275390625,
"signal/accuracy_reward/group_std_mean": 0.09114883244037628,
"signal/accuracy_reward/group_zero_std_frac": 0.734375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6938984274864197,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.034161376953125,
"signal/advantage_abs_mean": 0.781197476387024,
"signal/advantage_pre_scale_abs_mean": 0.053500982373952864,
"signal/advantage_pre_scale_std": 0.08611135184764862,
"signal/advantage_std": 0.9824514150619507,
"signal/brier_reward/centered_abs_mean": 0.08797314614057541,
"signal/brier_reward/group_std_mean": 0.11378528028726578,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1813061386346817,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008797314763069154,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.01324790045619011,
"signal/confidence_uniqueness_reward/group_std_mean": 0.01836627870798111,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.027479372546076776,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013247900642454624,
"signal/format_reward/centered_abs_mean": 0.00113525390625,
"signal/format_reward/group_std_mean": 0.0033145629800856113,
"signal/format_reward/group_zero_std_frac": 0.98125,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.011674307473003864,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.000567626953125,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2860772132873535,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.35528329014778137,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5943777918815613,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028607721999287607,
"step": 280
},
{
"calibration/aurc": 0.32501498874480483,
"calibration/batch_distribution_entropy": 0.972210059582548,
"calibration/batch_entropy_100bins": 0.9654372376100943,
"calibration/batch_entropy_10bins": 0.972210059582548,
"calibration/batch_entropy_50bins": 0.9728120906048648,
"calibration/batch_uniqueness": 0.9520518050201329,
"calibration/buffer_distribution_entropy": 0.9915784673227502,
"calibration/buffer_entropy_100bins": 0.9952996720322081,
"calibration/buffer_entropy_10bins": 0.9915784673227502,
"calibration/buffer_entropy_50bins": 0.9947027550696191,
"calibration/confidence_entropy": 0.4685566777731044,
"calibration/coverage@0%": 0.02539291829745597,
"calibration/coverage@1%": 0.02539291829745597,
"calibration/coverage@10%": 0.08441322162426615,
"calibration/coverage@15%": 0.22323568982387476,
"calibration/coverage@20%": 0.3982777336105675,
"calibration/coverage@25%": 0.4607991377201565,
"calibration/coverage@30%": 0.5155080418297456,
"calibration/coverage@5%": 0.04806292808219178,
"calibration/distribution_entropy_10": 0.972210059582548,
"calibration/distribution_entropy_100": 0.9654372376100943,
"calibration/ece": 0.12314312933519764,
"calibration/mean_confidence": 0.4998538969843671,
"calibration/unique_confidence_per_question": 0.996484375,
"calibration/unique_confidences": 510.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 633.6,
"completions/max_terminated_length": 633.6,
"completions/mean_length": 254.82490234375,
"completions/mean_terminated_length": 254.87522583007814,
"completions/min_length": 76.6,
"completions/min_terminated_length": 127.4,
"epoch": 0.912,
"grad_norm": 0.007527172099798918,
"learning_rate": 1e-06,
"loss": -0.0036,
"num_tokens": 983875455.0,
"reward": 0.948394238948822,
"reward_std": 0.06839405074715614,
"rewards/accgated_coverage_0": 0.03512025102972984,
"rewards/accgated_coverage_1": 0.03512025102972984,
"rewards/accgated_coverage_10": 0.03483972474932671,
"rewards/accgated_coverage_15": 0.026702636480331422,
"rewards/accgated_coverage_20": 0.020785190910100938,
"rewards/accgated_coverage_25": 0.026924856565892698,
"rewards/accgated_coverage_5": 0.0351193081587553,
"rewards/accuracy_reward": 0.54521484375,
"rewards/brier_reward": 0.8287744760513306,
"rewards/confidence_uniqueness_reward": 0.9504788875579834,
"rewards/format_reward": 0.99970703125,
"rewards/frontier_entropy_batch_reward": -0.23453265726566314,
"signal/accgated_coverage_0/centered_abs_mean": 0.0383393406867981,
"signal/accgated_coverage_0/group_std_mean": 0.050081229954957965,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08058620393276214,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0038339341524988415,
"signal/accgated_coverage_1/centered_abs_mean": 0.0383393406867981,
"signal/accgated_coverage_1/group_std_mean": 0.050081229954957965,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08058620393276214,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0038339341524988415,
"signal/accgated_coverage_10/centered_abs_mean": 0.03798588439822197,
"signal/accgated_coverage_10/group_std_mean": 0.049623039364814756,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07982371896505355,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0037985884118825197,
"signal/accgated_coverage_15/centered_abs_mean": 0.027651983499526977,
"signal/accgated_coverage_15/group_std_mean": 0.03628884553909302,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05809515565633774,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0027651983313262463,
"signal/accgated_coverage_20/centered_abs_mean": 0.016608574241399766,
"signal/accgated_coverage_20/group_std_mean": 0.021373267844319345,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03512862734496593,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0016608575358986854,
"signal/accgated_coverage_25/centered_abs_mean": 0.017372503690421582,
"signal/accgated_coverage_25/group_std_mean": 0.021950911730527878,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.0368684321641922,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0017372504342347383,
"signal/accgated_coverage_5/centered_abs_mean": 0.0383368007838726,
"signal/accgated_coverage_5/group_std_mean": 0.0500779464840889,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08058080971240997,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0038336800411343575,
"signal/accuracy_reward/centered_abs_mean": 0.073284912109375,
"signal/accuracy_reward/group_std_mean": 0.09905661940574646,
"signal/accuracy_reward/group_zero_std_frac": 0.703125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7615991353988647,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0366424560546875,
"signal/advantage_abs_mean": 0.7667989134788513,
"signal/advantage_pre_scale_abs_mean": 0.05260523781180382,
"signal/advantage_pre_scale_std": 0.08477867543697357,
"signal/advantage_std": 0.9824009418487549,
"signal/brier_reward/centered_abs_mean": 0.08646931946277618,
"signal/brier_reward/group_std_mean": 0.11288875192403794,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18267132937908173,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008646932151168584,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.012981362268328666,
"signal/confidence_uniqueness_reward/group_std_mean": 0.017048532143235207,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.027543330937623976,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001298136287368834,
"signal/format_reward/centered_abs_mean": 0.000567626953125,
"signal/format_reward/group_std_mean": 0.0016572814900428056,
"signal/format_reward/group_zero_std_frac": 0.990625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.005691004358232022,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0002838134765625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2698351740837097,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.34424389004707334,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5730916380882263,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026983517780900003,
"step": 285
},
{
"calibration/aurc": 0.36007844980074855,
"calibration/batch_distribution_entropy": 0.9681150768944194,
"calibration/batch_entropy_100bins": 0.9639867694430668,
"calibration/batch_entropy_10bins": 0.9681150768944194,
"calibration/batch_entropy_50bins": 0.9704624502040309,
"calibration/batch_uniqueness": 0.9509854443177244,
"calibration/buffer_distribution_entropy": 0.9922717211693882,
"calibration/buffer_entropy_100bins": 0.9957055094687668,
"calibration/buffer_entropy_10bins": 0.9922717211693882,
"calibration/buffer_entropy_50bins": 0.9951627786858855,
"calibration/confidence_entropy": 0.4618129154760141,
"calibration/coverage@0%": 0.02109375,
"calibration/coverage@1%": 0.02109375,
"calibration/coverage@10%": 0.1390625,
"calibration/coverage@15%": 0.172265625,
"calibration/coverage@20%": 0.2105774522994129,
"calibration/coverage@25%": 0.2856324914383562,
"calibration/coverage@30%": 0.3470187133072407,
"calibration/coverage@5%": 0.059765625,
"calibration/distribution_entropy_10": 0.9681150768944194,
"calibration/distribution_entropy_100": 0.9639867694430668,
"calibration/ece": 0.12372173297931303,
"calibration/mean_confidence": 0.4747332510295301,
"calibration/unique_confidence_per_question": 0.997265625,
"calibration/unique_confidences": 510.6,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 543.0,
"completions/max_terminated_length": 543.0,
"completions/mean_length": 248.94404296875,
"completions/mean_terminated_length": 248.9938171386719,
"completions/min_length": 97.8,
"completions/min_terminated_length": 122.4,
"epoch": 0.928,
"grad_norm": 0.0065694162622094154,
"learning_rate": 1e-06,
"loss": -0.0021,
"num_tokens": 1001451458.0,
"reward": 0.9415223479270936,
"reward_std": 0.06293211877346039,
"rewards/accgated_coverage_0": 0.040309954434633255,
"rewards/accgated_coverage_1": 0.040309954434633255,
"rewards/accgated_coverage_10": 0.03992871195077896,
"rewards/accgated_coverage_15": 0.030955337360501288,
"rewards/accgated_coverage_20": 0.02219003066420555,
"rewards/accgated_coverage_25": 0.025744926184415817,
"rewards/accgated_coverage_5": 0.04030746668577194,
"rewards/accuracy_reward": 0.53896484375,
"rewards/brier_reward": 0.824522066116333,
"rewards/confidence_uniqueness_reward": 0.9471201777458191,
"rewards/format_reward": 0.99970703125,
"rewards/frontier_entropy_batch_reward": -0.2895245313644409,
"signal/accgated_coverage_0/centered_abs_mean": 0.03615171872079372,
"signal/accgated_coverage_0/group_std_mean": 0.047892171144485476,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08107624128460884,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0036151719745248557,
"signal/accgated_coverage_1/centered_abs_mean": 0.03615171872079372,
"signal/accgated_coverage_1/group_std_mean": 0.047892171144485476,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08107624128460884,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0036151719745248557,
"signal/accgated_coverage_10/centered_abs_mean": 0.0357041236013174,
"signal/accgated_coverage_10/group_std_mean": 0.04731676504015923,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08007377609610558,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.003570412378758192,
"signal/accgated_coverage_15/centered_abs_mean": 0.02639325223863125,
"signal/accgated_coverage_15/group_std_mean": 0.03501335047185421,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.05928588733077049,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0026393252424895763,
"signal/accgated_coverage_20/centered_abs_mean": 0.01628087479621172,
"signal/accgated_coverage_20/group_std_mean": 0.020937855169177056,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.0365210197865963,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0016280874609947205,
"signal/accgated_coverage_25/centered_abs_mean": 0.01691762190312147,
"signal/accgated_coverage_25/group_std_mean": 0.021358567103743554,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03800369873642921,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0016917621716856957,
"signal/accgated_coverage_5/centered_abs_mean": 0.03614833541214466,
"signal/accgated_coverage_5/group_std_mean": 0.047887755930423735,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08106868118047714,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.003614833578467369,
"signal/accuracy_reward/centered_abs_mean": 0.055328369140625,
"signal/accuracy_reward/group_std_mean": 0.08216724917292595,
"signal/accuracy_reward/group_zero_std_frac": 0.73125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6096242845058442,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0276641845703125,
"signal/advantage_abs_mean": 0.7644638061523438,
"signal/advantage_pre_scale_abs_mean": 0.046795842796564104,
"signal/advantage_pre_scale_std": 0.07591844350099564,
"signal/advantage_std": 0.982296884059906,
"signal/brier_reward/centered_abs_mean": 0.08460079729557038,
"signal/brier_reward/group_std_mean": 0.11188524514436722,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18806430995464324,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008460079878568649,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.014814193360507488,
"signal/confidence_uniqueness_reward/group_std_mean": 0.01929043047130108,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0329870592802763,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014814193593338132,
"signal/format_reward/centered_abs_mean": 0.000567626953125,
"signal/format_reward/group_std_mean": 0.0016572814900428056,
"signal/format_reward/group_zero_std_frac": 0.990625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.006176374293863774,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0002838134765625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2983026444911957,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.36996843814849856,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.665130889415741,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.029830264672636986,
"step": 290
},
{
"calibration/aurc": 0.25867875098025755,
"calibration/batch_distribution_entropy": 0.9718980786826137,
"calibration/batch_entropy_100bins": 0.9615980296544752,
"calibration/batch_entropy_10bins": 0.9718980786826137,
"calibration/batch_entropy_50bins": 0.9712295548644937,
"calibration/batch_uniqueness": 0.9509344162988951,
"calibration/buffer_distribution_entropy": 0.9923605608234588,
"calibration/buffer_entropy_100bins": 0.9957812903097041,
"calibration/buffer_entropy_10bins": 0.9923605608234588,
"calibration/buffer_entropy_50bins": 0.9952399716386944,
"calibration/confidence_entropy": 0.47118211635788254,
"calibration/coverage@0%": 0.021875764432485324,
"calibration/coverage@1%": 0.021875764432485324,
"calibration/coverage@10%": 0.2300788894324853,
"calibration/coverage@15%": 0.3234382644324853,
"calibration/coverage@20%": 0.39532778864970647,
"calibration/coverage@25%": 0.5254701259784735,
"calibration/coverage@30%": 0.6329248715753424,
"calibration/coverage@5%": 0.09882888943248533,
"calibration/distribution_entropy_10": 0.9718980786826137,
"calibration/distribution_entropy_100": 0.9615980296544752,
"calibration/ece": 0.11829233966323109,
"calibration/mean_confidence": 0.48343642093352895,
"calibration/unique_confidence_per_question": 0.996484375,
"calibration/unique_confidences": 510.2,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 720.2,
"completions/max_terminated_length": 720.2,
"completions/mean_length": 255.13916015625,
"completions/mean_terminated_length": 255.16352233886718,
"completions/min_length": 90.0,
"completions/min_terminated_length": 113.0,
"epoch": 0.944,
"grad_norm": 0.017357630655169487,
"learning_rate": 1e-06,
"loss": -0.0001,
"num_tokens": 1019039507.0,
"reward": 0.9455022096633912,
"reward_std": 0.0739950880408287,
"rewards/accgated_coverage_0": 0.035088099911808966,
"rewards/accgated_coverage_1": 0.035088099911808966,
"rewards/accgated_coverage_10": 0.03473405130207539,
"rewards/accgated_coverage_15": 0.02834833338856697,
"rewards/accgated_coverage_20": 0.020123963244259357,
"rewards/accgated_coverage_25": 0.02252167705446482,
"rewards/accgated_coverage_5": 0.03508702889084816,
"rewards/accuracy_reward": 0.5501953125,
"rewards/brier_reward": 0.8146409511566162,
"rewards/confidence_uniqueness_reward": 0.9485449671745301,
"rewards/format_reward": 0.99990234375,
"rewards/frontier_entropy_batch_reward": -0.2696434497833252,
"signal/accgated_coverage_0/centered_abs_mean": 0.04485566839575768,
"signal/accgated_coverage_0/group_std_mean": 0.057407965511083604,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.09075943529605865,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004485567007213831,
"signal/accgated_coverage_1/centered_abs_mean": 0.04485566839575768,
"signal/accgated_coverage_1/group_std_mean": 0.057407965511083604,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.09075943529605865,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004485567007213831,
"signal/accgated_coverage_10/centered_abs_mean": 0.04434861540794373,
"signal/accgated_coverage_10/group_std_mean": 0.056770000606775284,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08973048031330108,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004434861522167921,
"signal/accgated_coverage_15/centered_abs_mean": 0.03330285735428333,
"signal/accgated_coverage_15/group_std_mean": 0.04269464835524559,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06738676428794861,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0033302858471870424,
"signal/accgated_coverage_20/centered_abs_mean": 0.01831945925951004,
"signal/accgated_coverage_20/group_std_mean": 0.023091159760951996,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.037078937888145445,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.001831945963203907,
"signal/accgated_coverage_25/centered_abs_mean": 0.017375039309263228,
"signal/accgated_coverage_25/group_std_mean": 0.021864892542362215,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03524128533899784,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0017375039868056774,
"signal/accgated_coverage_5/centered_abs_mean": 0.044851568341255185,
"signal/accgated_coverage_5/group_std_mean": 0.057402705401182176,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.090751151740551,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004485156945884227,
"signal/accuracy_reward/centered_abs_mean": 0.091015625,
"signal/accuracy_reward/group_std_mean": 0.11987628191709518,
"signal/accuracy_reward/group_zero_std_frac": 0.659375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.9179443717002869,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0455078125,
"signal/advantage_abs_mean": 0.7828087091445923,
"signal/advantage_pre_scale_abs_mean": 0.05813053473830223,
"signal/advantage_pre_scale_std": 0.09080860167741775,
"signal/advantage_std": 0.9825004458427429,
"signal/brier_reward/centered_abs_mean": 0.0965838834643364,
"signal/brier_reward/group_std_mean": 0.1238422930240631,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.19514789879322053,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009658388514071702,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.013734615594148635,
"signal/confidence_uniqueness_reward/group_std_mean": 0.017413123138248922,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.027795661240816116,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013734615640714765,
"signal/format_reward/centered_abs_mean": 0.000189208984375,
"signal/format_reward/group_std_mean": 0.0005524271633476019,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.001997852511703968,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 9.46044921875e-05,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2891539216041565,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3595525801181793,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5847546458244324,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028915392979979516,
"step": 295
},
{
"calibration/aurc": 0.31199443796174353,
"calibration/batch_distribution_entropy": 0.9466913738300166,
"calibration/batch_entropy_100bins": 0.9517573639197658,
"calibration/batch_entropy_10bins": 0.9466913738300166,
"calibration/batch_entropy_50bins": 0.9573104912141733,
"calibration/batch_uniqueness": 0.9474395751953125,
"calibration/buffer_distribution_entropy": 0.9920504630636374,
"calibration/buffer_entropy_100bins": 0.9956035848261736,
"calibration/buffer_entropy_10bins": 0.9920504630636374,
"calibration/buffer_entropy_50bins": 0.9950359333856049,
"calibration/confidence_entropy": 0.4392061302464286,
"calibration/coverage@0%": 0.0125,
"calibration/coverage@1%": 0.0125,
"calibration/coverage@10%": 0.018359375,
"calibration/coverage@15%": 0.23828125,
"calibration/coverage@20%": 0.288671875,
"calibration/coverage@25%": 0.34375,
"calibration/coverage@30%": 0.503125,
"calibration/coverage@5%": 0.01328125,
"calibration/distribution_entropy_10": 0.9466913738300166,
"calibration/distribution_entropy_100": 0.9517573639197658,
"calibration/ece": 0.1278921471672227,
"calibration/mean_confidence": 0.564862040921503,
"calibration/unique_confidence_per_question": 0.995703125,
"calibration/unique_confidences": 509.8,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 641.8,
"completions/max_terminated_length": 641.8,
"completions/mean_length": 258.07294921875,
"completions/mean_terminated_length": 258.12323608398435,
"completions/min_length": 76.4,
"completions/min_terminated_length": 125.0,
"epoch": 0.96,
"grad_norm": 0.007013469934463501,
"learning_rate": 1e-06,
"loss": -0.0022,
"num_tokens": 1036622494.0,
"reward": 0.9375743627548218,
"reward_std": 0.06640594154596328,
"rewards/accgated_coverage_0": 0.03816794380545616,
"rewards/accgated_coverage_1": 0.03816794380545616,
"rewards/accgated_coverage_10": 0.037914327904582025,
"rewards/accgated_coverage_15": 0.031234052032232285,
"rewards/accgated_coverage_20": 0.021295220032334326,
"rewards/accgated_coverage_25": 0.026653024181723595,
"rewards/accgated_coverage_5": 0.03816728405654431,
"rewards/accuracy_reward": 0.5361328125,
"rewards/brier_reward": 0.8256236791610718,
"rewards/confidence_uniqueness_reward": 0.9464357614517211,
"rewards/format_reward": 0.99970703125,
"rewards/frontier_entropy_batch_reward": -0.3071149528026581,
"signal/accgated_coverage_0/centered_abs_mean": 0.03682501018047333,
"signal/accgated_coverage_0/group_std_mean": 0.047518537193536756,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07903042733669281,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0036825011018663647,
"signal/accgated_coverage_1/centered_abs_mean": 0.03682501018047333,
"signal/accgated_coverage_1/group_std_mean": 0.047518537193536756,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07903042733669281,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0036825011018663647,
"signal/accgated_coverage_10/centered_abs_mean": 0.03660396635532379,
"signal/accgated_coverage_10/group_std_mean": 0.04722764790058136,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07855970412492752,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0036603968124836682,
"signal/accgated_coverage_15/centered_abs_mean": 0.028405727818608284,
"signal/accgated_coverage_15/group_std_mean": 0.03654852136969566,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06116994246840477,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0028405728749930858,
"signal/accgated_coverage_20/centered_abs_mean": 0.01682550981640816,
"signal/accgated_coverage_20/group_std_mean": 0.021106991916894913,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03618932627141476,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0016825510188937186,
"signal/accgated_coverage_25/centered_abs_mean": 0.018247338384389876,
"signal/accgated_coverage_25/group_std_mean": 0.022746633738279343,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.039190822094678876,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.001824733824469149,
"signal/accgated_coverage_5/centered_abs_mean": 0.036820971965789796,
"signal/accgated_coverage_5/group_std_mean": 0.04751335978507996,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07902174293994904,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.00368209732696414,
"signal/accuracy_reward/centered_abs_mean": 0.06614990234375,
"signal/accuracy_reward/group_std_mean": 0.08971318304538727,
"signal/accuracy_reward/group_zero_std_frac": 0.734375,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6973819017410279,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.033074951171875,
"signal/advantage_abs_mean": 0.7846113562583923,
"signal/advantage_pre_scale_abs_mean": 0.05130809471011162,
"signal/advantage_pre_scale_std": 0.08110593557357788,
"signal/advantage_std": 0.9823913455009461,
"signal/brier_reward/centered_abs_mean": 0.08842150419950486,
"signal/brier_reward/group_std_mean": 0.11597429811954499,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18856617212295532,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.008842150680720806,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.014768127351999283,
"signal/confidence_uniqueness_reward/group_std_mean": 0.019342204183340074,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.03170429207384586,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014768127584829926,
"signal/format_reward/centered_abs_mean": 0.000567626953125,
"signal/format_reward/group_std_mean": 0.0016572814900428056,
"signal/format_reward/group_zero_std_frac": 0.990625,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.006288815476000309,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0002838134765625,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.30554712414741514,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3763702929019928,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.6513253450393677,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.030554712191224097,
"step": 300
},
{
"epoch": 0.96,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 486.25,
"eval_completions/max_terminated_length": 486.25,
"eval_completions/mean_length": 258.4506301879883,
"eval_completions/mean_terminated_length": 258.4506301879883,
"eval_completions/min_length": 142.5,
"eval_completions/min_terminated_length": 142.5,
"eval_loss": 0.0,
"eval_num_tokens": 1036622494.0,
"eval_reward": 0.8093127459287643,
"eval_reward_std": 0.22334988415241241,
"eval_rewards/accgated_coverage_0": 0.038257550448179245,
"eval_rewards/accgated_coverage_1": 0.038257550448179245,
"eval_rewards/accgated_coverage_10": 0.03812482440844178,
"eval_rewards/accgated_coverage_15": 0.030789055861532688,
"eval_rewards/accgated_coverage_20": 0.01574985240586102,
"eval_rewards/accgated_coverage_25": 0.012001962284557521,
"eval_rewards/accgated_coverage_5": 0.0382560845464468,
"eval_rewards/accuracy_reward": 0.435546875,
"eval_rewards/brier_reward": 0.8099131137132645,
"eval_rewards/confidence_uniqueness_reward": 0.89404296875,
"eval_rewards/format_reward": 1.0,
"eval_rewards/frontier_entropy_batch_reward": -1.0,
"eval_runtime": 23.5653,
"eval_samples_per_second": 21.218,
"eval_signal/accgated_coverage_0/centered_abs_mean": 0.07457183301448822,
"eval_signal/accgated_coverage_0/group_std_mean": 0.08923457935452461,
"eval_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.033520148135721684,
"eval_signal/accgated_coverage_0/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.007457182859070599,
"eval_signal/accgated_coverage_1/centered_abs_mean": 0.07457183301448822,
"eval_signal/accgated_coverage_1/group_std_mean": 0.08923457935452461,
"eval_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.033520148135721684,
"eval_signal/accgated_coverage_1/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.007457182859070599,
"eval_signal/accgated_coverage_10/centered_abs_mean": 0.07421804405748844,
"eval_signal/accgated_coverage_10/group_std_mean": 0.08883162401616573,
"eval_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.03336044121533632,
"eval_signal/accgated_coverage_10/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.007421804475598037,
"eval_signal/accgated_coverage_15/centered_abs_mean": 0.05673111416399479,
"eval_signal/accgated_coverage_15/group_std_mean": 0.06877893023192883,
"eval_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.02549748169258237,
"eval_signal/accgated_coverage_15/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.005673111765645444,
"eval_signal/accgated_coverage_20/centered_abs_mean": 0.022827666252851486,
"eval_signal/accgated_coverage_20/group_std_mean": 0.028775827959179878,
"eval_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.010260861599817872,
"eval_signal/accgated_coverage_20/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0022827666252851486,
"eval_signal/accgated_coverage_25/centered_abs_mean": 0.018368238117545843,
"eval_signal/accgated_coverage_25/group_std_mean": 0.023854423314332962,
"eval_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.008252283791080117,
"eval_signal/accgated_coverage_25/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0018368238816037774,
"eval_signal/accgated_coverage_5/centered_abs_mean": 0.07456586603075266,
"eval_signal/accgated_coverage_5/group_std_mean": 0.0892275907099247,
"eval_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"eval_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.03351746080443263,
"eval_signal/accgated_coverage_5/weight": 0.10000000149011612,
"eval_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.007456586696207523,
"eval_signal/accuracy_reward/centered_abs_mean": 0.4771728515625,
"eval_signal/accuracy_reward/group_std_mean": 0.4959787428379059,
"eval_signal/accuracy_reward/group_zero_std_frac": 0.0,
"eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.071576565504074,
"eval_signal/accuracy_reward/weight": 0.5,
"eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.23858642578125,
"eval_signal/advantage_abs_mean": 0.931053102016449,
"eval_signal/advantage_pre_scale_abs_mean": 0.20829248800873756,
"eval_signal/advantage_pre_scale_std": 0.22105665132403374,
"eval_signal/advantage_std": 0.9876744598150253,
"eval_signal/brier_reward/centered_abs_mean": 0.1904592178761959,
"eval_signal/brier_reward/group_std_mean": 0.24390720203518867,
"eval_signal/brier_reward/group_zero_std_frac": 0.0,
"eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.0856118556112051,
"eval_signal/brier_reward/weight": 0.10000000149011612,
"eval_signal/brier_reward/weighted_centered_abs_mean": 0.01904592150822282,
"eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.041290283203125,
"eval_signal/confidence_uniqueness_reward/group_std_mean": 0.049044907093048096,
"eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.018552962224930525,
"eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004129028180614114,
"eval_signal/format_reward/centered_abs_mean": 0.0,
"eval_signal/format_reward/group_std_mean": 0.0,
"eval_signal/format_reward/group_zero_std_frac": 1.0,
"eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/format_reward/weight": 0.5,
"eval_signal/format_reward/weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"eval_steps_per_second": 0.17,
"step": 300
},
{
"epoch": 0.96,
"step": 300,
"train_probe_completions/clipped_ratio": 0.0,
"train_probe_completions/max_length": 544.75,
"train_probe_completions/max_terminated_length": 544.75,
"train_probe_completions/mean_length": 256.78542709350586,
"train_probe_completions/mean_terminated_length": 256.78542709350586,
"train_probe_completions/min_length": 135.0,
"train_probe_completions/min_terminated_length": 135.0,
"train_probe_loss": 0.0,
"train_probe_num_tokens": 1036622494.0,
"train_probe_reward": 0.9038369059562683,
"train_probe_reward_std": 0.1930069997906685,
"train_probe_rewards/accgated_coverage_0": 0.03255564160645008,
"train_probe_rewards/accgated_coverage_1": 0.03255564160645008,
"train_probe_rewards/accgated_coverage_10": 0.03251849673688412,
"train_probe_rewards/accgated_coverage_15": 0.03028361964970827,
"train_probe_rewards/accgated_coverage_20": 0.02410488505847752,
"train_probe_rewards/accgated_coverage_25": 0.03020387515425682,
"train_probe_rewards/accgated_coverage_5": 0.03255576081573963,
"train_probe_rewards/accuracy_reward": 0.62109375,
"train_probe_rewards/brier_reward": 0.8250560164451599,
"train_probe_rewards/confidence_uniqueness_reward": 0.89306640625,
"train_probe_rewards/format_reward": 1.0,
"train_probe_rewards/frontier_entropy_batch_reward": -1.0,
"train_probe_runtime": 25.4798,
"train_probe_samples_per_second": 19.623,
"train_probe_signal/accgated_coverage_0/centered_abs_mean": 0.1296680662781,
"train_probe_signal/accgated_coverage_0/group_std_mean": 0.16905909031629562,
"train_probe_signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.0696471193805337,
"train_probe_signal/accgated_coverage_0/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_0/weighted_centered_abs_mean": 0.012966807233169675,
"train_probe_signal/accgated_coverage_1/centered_abs_mean": 0.1296680662781,
"train_probe_signal/accgated_coverage_1/group_std_mean": 0.16905909031629562,
"train_probe_signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.0696471193805337,
"train_probe_signal/accgated_coverage_1/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_1/weighted_centered_abs_mean": 0.012966807233169675,
"train_probe_signal/accgated_coverage_10/centered_abs_mean": 0.12882263399660587,
"train_probe_signal/accgated_coverage_10/group_std_mean": 0.16800688207149506,
"train_probe_signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.06918841321021318,
"train_probe_signal/accgated_coverage_10/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_10/weighted_centered_abs_mean": 0.012882263865321875,
"train_probe_signal/accgated_coverage_15/centered_abs_mean": 0.09863024391233921,
"train_probe_signal/accgated_coverage_15/group_std_mean": 0.12959836795926094,
"train_probe_signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.052916424348950386,
"train_probe_signal/accgated_coverage_15/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_15/weighted_centered_abs_mean": 0.009863024577498436,
"train_probe_signal/accgated_coverage_20/centered_abs_mean": 0.040732065215706825,
"train_probe_signal/accgated_coverage_20/group_std_mean": 0.053746563382446766,
"train_probe_signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.02184022730216384,
"train_probe_signal/accgated_coverage_20/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_20/weighted_centered_abs_mean": 0.004073206626344472,
"train_probe_signal/accgated_coverage_25/centered_abs_mean": 0.037193214520812035,
"train_probe_signal/accgated_coverage_25/group_std_mean": 0.04660756699740887,
"train_probe_signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.019912811694666743,
"train_probe_signal/accgated_coverage_25/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0037193215684965253,
"train_probe_signal/accgated_coverage_5/centered_abs_mean": 0.12965733744204044,
"train_probe_signal/accgated_coverage_5/group_std_mean": 0.16904626414179802,
"train_probe_signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"train_probe_signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.06964142806828022,
"train_probe_signal/accgated_coverage_5/weight": 0.10000000149011612,
"train_probe_signal/accgated_coverage_5/weighted_centered_abs_mean": 0.012965733651071787,
"train_probe_signal/accuracy_reward/centered_abs_mean": 0.455078125,
"train_probe_signal/accuracy_reward/group_std_mean": 0.4842090755701065,
"train_probe_signal/accuracy_reward/group_zero_std_frac": 0.0,
"train_probe_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 1.1914450526237488,
"train_probe_signal/accuracy_reward/weight": 0.5,
"train_probe_signal/accuracy_reward/weighted_centered_abs_mean": 0.2275390625,
"train_probe_signal/advantage_abs_mean": 0.8866461962461472,
"train_probe_signal/advantage_pre_scale_abs_mean": 0.17213350534439087,
"train_probe_signal/advantage_pre_scale_std": 0.19129860401153564,
"train_probe_signal/advantage_std": 0.98759725689888,
"train_probe_signal/brier_reward/centered_abs_mean": 0.17349136993288994,
"train_probe_signal/brier_reward/group_std_mean": 0.22335120663046837,
"train_probe_signal/brier_reward/group_zero_std_frac": 0.0,
"train_probe_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.09027872607111931,
"train_probe_signal/brier_reward/weight": 0.10000000149011612,
"train_probe_signal/brier_reward/weighted_centered_abs_mean": 0.017349137226119637,
"train_probe_signal/confidence_uniqueness_reward/centered_abs_mean": 0.04248046875,
"train_probe_signal/confidence_uniqueness_reward/group_std_mean": 0.05080406926572323,
"train_probe_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"train_probe_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.022486352361738682,
"train_probe_signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"train_probe_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004248047014698386,
"train_probe_signal/format_reward/centered_abs_mean": 0.0,
"train_probe_signal/format_reward/group_std_mean": 0.0,
"train_probe_signal/format_reward/group_zero_std_frac": 1.0,
"train_probe_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/format_reward/weight": 0.5,
"train_probe_signal/format_reward/weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_std_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/group_zero_std_frac": 1.0,
"train_probe_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0,
"train_probe_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"train_probe_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0,
"train_probe_steps_per_second": 0.157
},
{
"calibration/aurc": 0.21932777041170565,
"calibration/batch_distribution_entropy": 0.9781350322086324,
"calibration/batch_entropy_100bins": 0.9653315439810859,
"calibration/batch_entropy_10bins": 0.9781350322086324,
"calibration/batch_entropy_50bins": 0.9754277787813974,
"calibration/batch_uniqueness": 0.9525615741063964,
"calibration/buffer_distribution_entropy": 0.9906859739988491,
"calibration/buffer_entropy_100bins": 0.9949244929119552,
"calibration/buffer_entropy_10bins": 0.9906859739988491,
"calibration/buffer_entropy_50bins": 0.9942236126601367,
"calibration/confidence_entropy": 0.47164863915663,
"calibration/coverage@0%": 0.06174703400195695,
"calibration/coverage@1%": 0.06252828400195695,
"calibration/coverage@10%": 0.3598841120352251,
"calibration/coverage@15%": 0.4357005259295499,
"calibration/coverage@20%": 0.5150195694716243,
"calibration/coverage@25%": 0.5724651418786693,
"calibration/coverage@30%": 0.6400730797455969,
"calibration/coverage@5%": 0.19500290484344424,
"calibration/distribution_entropy_10": 0.9781350322086324,
"calibration/distribution_entropy_100": 0.9653315439810859,
"calibration/ece": 0.13754592922330594,
"calibration/mean_confidence": 0.5109261548232384,
"calibration/unique_confidence_per_question": 0.994921875,
"calibration/unique_confidences": 509.4,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.000390625,
"completions/max_length": 747.0,
"completions/max_terminated_length": 747.0,
"completions/mean_length": 258.12626953125,
"completions/mean_terminated_length": 258.2263977050781,
"completions/min_length": 49.0,
"completions/min_terminated_length": 121.4,
"epoch": 0.976,
"grad_norm": 0.00685409689322114,
"learning_rate": 1e-06,
"loss": -0.0018,
"num_tokens": 1054126827.0,
"reward": 0.9543739318847656,
"reward_std": 0.06868521645665168,
"rewards/accgated_coverage_0": 0.03272476131096482,
"rewards/accgated_coverage_1": 0.03272476131096482,
"rewards/accgated_coverage_10": 0.032653958443552254,
"rewards/accgated_coverage_15": 0.02865794561803341,
"rewards/accgated_coverage_20": 0.019704227894544603,
"rewards/accgated_coverage_25": 0.022663561254739763,
"rewards/accgated_coverage_5": 0.03272439446300268,
"rewards/accuracy_reward": 0.56171875,
"rewards/brier_reward": 0.8187178611755371,
"rewards/confidence_uniqueness_reward": 0.9504208087921142,
"rewards/format_reward": 0.99951171875,
"rewards/frontier_entropy_batch_reward": -0.23340550661087037,
"signal/accgated_coverage_0/centered_abs_mean": 0.04211543202400207,
"signal/accgated_coverage_0/group_std_mean": 0.05437769591808319,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08571670949459076,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.0042115431278944016,
"signal/accgated_coverage_1/centered_abs_mean": 0.04211543202400207,
"signal/accgated_coverage_1/group_std_mean": 0.05437769591808319,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08571670949459076,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.0042115431278944016,
"signal/accgated_coverage_10/centered_abs_mean": 0.041787787526845935,
"signal/accgated_coverage_10/group_std_mean": 0.053954492509365085,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08507721722126008,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.004178778827190399,
"signal/accgated_coverage_15/centered_abs_mean": 0.03185315728187561,
"signal/accgated_coverage_15/group_std_mean": 0.041186673939228056,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06508256793022156,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0031853157095611097,
"signal/accgated_coverage_20/centered_abs_mean": 0.017366071790456773,
"signal/accgated_coverage_20/group_std_mean": 0.02201416380703449,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03550922945141792,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0017366073094308375,
"signal/accgated_coverage_25/centered_abs_mean": 0.018135173246264458,
"signal/accgated_coverage_25/group_std_mean": 0.022788660228252412,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03720953315496445,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0018135173711925745,
"signal/accgated_coverage_5/centered_abs_mean": 0.04211136251688004,
"signal/accgated_coverage_5/group_std_mean": 0.05437261387705803,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08570821881294251,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004211136559024453,
"signal/accuracy_reward/centered_abs_mean": 0.0708984375,
"signal/accuracy_reward/group_std_mean": 0.10080507844686508,
"signal/accuracy_reward/group_zero_std_frac": 0.68125,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.7143431305885315,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.03544921875,
"signal/advantage_abs_mean": 0.7610075950622559,
"signal/advantage_pre_scale_abs_mean": 0.051202143728733066,
"signal/advantage_pre_scale_std": 0.08305490612983704,
"signal/advantage_std": 0.9824317812919616,
"signal/brier_reward/centered_abs_mean": 0.08758036494255066,
"signal/brier_reward/group_std_mean": 0.11371626257896424,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.1802319645881653,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.00875803716480732,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.013032703660428524,
"signal/confidence_uniqueness_reward/group_std_mean": 0.017780544608831404,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.027440373227000235,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001303270412608981,
"signal/format_reward/centered_abs_mean": 0.000946044921875,
"signal/format_reward/group_std_mean": 0.0027621358167380095,
"signal/format_reward/group_zero_std_frac": 0.984375,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.009943905472755431,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.0004730224609375,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26888933777809143,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3424463629722595,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5625767886638642,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026888934895396233,
"step": 305
},
{
"calibration/aurc": 0.34134532141488927,
"calibration/batch_distribution_entropy": 0.9716329597394855,
"calibration/batch_entropy_100bins": 0.9666794622881343,
"calibration/batch_entropy_10bins": 0.9716329597394855,
"calibration/batch_entropy_50bins": 0.9719893099832102,
"calibration/batch_uniqueness": 0.952001953125,
"calibration/buffer_distribution_entropy": 0.9905505884576691,
"calibration/buffer_entropy_100bins": 0.9948464286455015,
"calibration/buffer_entropy_10bins": 0.9905505884576691,
"calibration/buffer_entropy_50bins": 0.9941327329868417,
"calibration/confidence_entropy": 0.4617977948845026,
"calibration/coverage@0%": 0.01171875,
"calibration/coverage@1%": 0.01171875,
"calibration/coverage@10%": 0.05625,
"calibration/coverage@15%": 0.077734375,
"calibration/coverage@20%": 0.317578125,
"calibration/coverage@25%": 0.39921875,
"calibration/coverage@30%": 0.47265625,
"calibration/coverage@5%": 0.01171875,
"calibration/distribution_entropy_10": 0.9716329597394855,
"calibration/distribution_entropy_100": 0.9666794622881343,
"calibration/ece": 0.14105595503308838,
"calibration/mean_confidence": 0.4634805730401171,
"calibration/unique_confidence_per_question": 0.99609375,
"calibration/unique_confidences": 510.0,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 9.765625e-05,
"completions/max_length": 637.2,
"completions/max_terminated_length": 637.2,
"completions/mean_length": 248.23583984375,
"completions/mean_terminated_length": 248.26107177734374,
"completions/min_length": 84.0,
"completions/min_terminated_length": 107.2,
"epoch": 0.992,
"grad_norm": 0.007217899896204472,
"learning_rate": 1e-06,
"loss": 0.0012,
"num_tokens": 1071797242.0,
"reward": 0.9381205320358277,
"reward_std": 0.06416866332292556,
"rewards/accgated_coverage_0": 0.02983600050210953,
"rewards/accgated_coverage_1": 0.02983600050210953,
"rewards/accgated_coverage_10": 0.029756875708699226,
"rewards/accgated_coverage_15": 0.025021048448979854,
"rewards/accgated_coverage_20": 0.017900421842932702,
"rewards/accgated_coverage_25": 0.023241546005010605,
"rewards/accgated_coverage_5": 0.02983627952635288,
"rewards/accuracy_reward": 0.53349609375,
"rewards/brier_reward": 0.8161760807037354,
"rewards/confidence_uniqueness_reward": 0.9509012460708618,
"rewards/format_reward": 0.99990234375,
"rewards/frontier_entropy_batch_reward": -0.23829238414764403,
"signal/accgated_coverage_0/centered_abs_mean": 0.035384940728545186,
"signal/accgated_coverage_0/group_std_mean": 0.04552566036581993,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.07650707140564919,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.003538494324311614,
"signal/accgated_coverage_1/centered_abs_mean": 0.035384940728545186,
"signal/accgated_coverage_1/group_std_mean": 0.04552566036581993,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.07650707140564919,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.003538494324311614,
"signal/accgated_coverage_10/centered_abs_mean": 0.03522704914212227,
"signal/accgated_coverage_10/group_std_mean": 0.045329224318265915,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.07616176977753639,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.0035227050073444845,
"signal/accgated_coverage_15/centered_abs_mean": 0.02708212360739708,
"signal/accgated_coverage_15/group_std_mean": 0.03494739718735218,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.0586032934486866,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.0027082124259322883,
"signal/accgated_coverage_20/centered_abs_mean": 0.015357860177755357,
"signal/accgated_coverage_20/group_std_mean": 0.0195449635386467,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03336197212338447,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.001535786036401987,
"signal/accgated_coverage_25/centered_abs_mean": 0.016744259931147098,
"signal/accgated_coverage_25/group_std_mean": 0.021145326271653175,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.036504270508885384,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0016744260676205159,
"signal/accgated_coverage_5/centered_abs_mean": 0.035381896421313286,
"signal/accgated_coverage_5/group_std_mean": 0.045521778613328935,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.07650035098195077,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.0035381895955651998,
"signal/accuracy_reward/centered_abs_mean": 0.058001708984375,
"signal/accuracy_reward/group_std_mean": 0.08016391545534134,
"signal/accuracy_reward/group_zero_std_frac": 0.75625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.6156765937805175,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.0290008544921875,
"signal/advantage_abs_mean": 0.7771116256713867,
"signal/advantage_pre_scale_abs_mean": 0.04954228028655052,
"signal/advantage_pre_scale_std": 0.07848731800913811,
"signal/advantage_std": 0.9823615074157714,
"signal/brier_reward/centered_abs_mean": 0.08547381907701493,
"signal/brier_reward/group_std_mean": 0.1098471462726593,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18402215838432312,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.00854738187044859,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.012716401740908623,
"signal/confidence_uniqueness_reward/group_std_mean": 0.01609993875026703,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.027407549694180487,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012716402066871523,
"signal/format_reward/centered_abs_mean": 0.000189208984375,
"signal/format_reward/group_std_mean": 0.0005524271633476019,
"signal/format_reward/group_zero_std_frac": 0.996875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.001923634484410286,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 9.46044921875e-05,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2780640244483948,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.3484475135803223,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.6001133680343628,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027806402742862703,
"step": 310
},
{
"calibration/aurc": 0.284015807048594,
"calibration/batch_distribution_entropy": 0.9722640913241893,
"calibration/batch_entropy_100bins": 0.9685421670055321,
"calibration/batch_entropy_10bins": 0.9722640913241893,
"calibration/batch_entropy_50bins": 0.9752045050612412,
"calibration/batch_uniqueness": 0.9531097412109375,
"calibration/buffer_distribution_entropy": 0.9908205294158174,
"calibration/buffer_entropy_100bins": 0.9949635077656247,
"calibration/buffer_entropy_10bins": 0.9908205294158174,
"calibration/buffer_entropy_50bins": 0.9942837222617167,
"calibration/confidence_entropy": 0.4589883674588277,
"calibration/coverage@0%": 0.0205078125,
"calibration/coverage@1%": 0.0205078125,
"calibration/coverage@10%": 0.0341796875,
"calibration/coverage@15%": 0.0556640625,
"calibration/coverage@20%": 0.177734375,
"calibration/coverage@25%": 0.4765625,
"calibration/coverage@30%": 0.6240234375,
"calibration/coverage@5%": 0.0205078125,
"calibration/distribution_entropy_10": 0.9722640913241893,
"calibration/distribution_entropy_100": 0.9685421670055321,
"calibration/ece": 0.11670367578399148,
"calibration/mean_confidence": 0.5715617792935408,
"calibration/unique_confidence_per_question": 0.9970703125,
"calibration/unique_confidences": 510.5,
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 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": 538.5,
"completions/max_terminated_length": 538.5,
"completions/mean_length": 248.3640594482422,
"completions/mean_terminated_length": 248.3640594482422,
"completions/min_length": 102.0,
"completions/min_terminated_length": 102.0,
"epoch": 0.9984,
"num_tokens": 1078816210.0,
"reward": 0.9534339010715485,
"reward_std": 0.07603390514850616,
"rewards/accgated_coverage_0": 0.017301072366535664,
"rewards/accgated_coverage_1": 0.017301072366535664,
"rewards/accgated_coverage_10": 0.017331603914499283,
"rewards/accgated_coverage_15": 0.016173084266483784,
"rewards/accgated_coverage_20": 0.014687915332615376,
"rewards/accgated_coverage_25": 0.02235917467623949,
"rewards/accgated_coverage_5": 0.01730012008920312,
"rewards/accuracy_reward": 0.571044921875,
"rewards/brier_reward": 0.7989811301231384,
"rewards/confidence_uniqueness_reward": 0.9529584348201752,
"rewards/format_reward": 0.999755859375,
"rewards/frontier_entropy_batch_reward": -0.19405893981456757,
"signal/accgated_coverage_0/centered_abs_mean": 0.046160491183400154,
"signal/accgated_coverage_0/group_std_mean": 0.0586283914744854,
"signal/accgated_coverage_0/group_zero_std_frac": 0.0,
"signal/accgated_coverage_0/scaled_weighted_centered_abs_mean": 0.08909685164690018,
"signal/accgated_coverage_0/weight": 0.10000000149011612,
"signal/accgated_coverage_0/weighted_centered_abs_mean": 0.004616048885509372,
"signal/accgated_coverage_1/centered_abs_mean": 0.046160491183400154,
"signal/accgated_coverage_1/group_std_mean": 0.0586283914744854,
"signal/accgated_coverage_1/group_zero_std_frac": 0.0,
"signal/accgated_coverage_1/scaled_weighted_centered_abs_mean": 0.08909685164690018,
"signal/accgated_coverage_1/weight": 0.10000000149011612,
"signal/accgated_coverage_1/weighted_centered_abs_mean": 0.004616048885509372,
"signal/accgated_coverage_10/centered_abs_mean": 0.046025536954402924,
"signal/accgated_coverage_10/group_std_mean": 0.058451514691114426,
"signal/accgated_coverage_10/group_zero_std_frac": 0.0,
"signal/accgated_coverage_10/scaled_weighted_centered_abs_mean": 0.08883628994226456,
"signal/accgated_coverage_10/weight": 0.10000000149011612,
"signal/accgated_coverage_10/weighted_centered_abs_mean": 0.00460255378857255,
"signal/accgated_coverage_15/centered_abs_mean": 0.035478485748171806,
"signal/accgated_coverage_15/group_std_mean": 0.0448724627494812,
"signal/accgated_coverage_15/group_zero_std_frac": 0.0,
"signal/accgated_coverage_15/scaled_weighted_centered_abs_mean": 0.06851176545023918,
"signal/accgated_coverage_15/weight": 0.10000000149011612,
"signal/accgated_coverage_15/weighted_centered_abs_mean": 0.003547848667949438,
"signal/accgated_coverage_20/centered_abs_mean": 0.017994429916143417,
"signal/accgated_coverage_20/group_std_mean": 0.022838743403553963,
"signal/accgated_coverage_20/group_zero_std_frac": 0.0,
"signal/accgated_coverage_20/scaled_weighted_centered_abs_mean": 0.03469894826412201,
"signal/accgated_coverage_20/weight": 0.10000000149011612,
"signal/accgated_coverage_20/weighted_centered_abs_mean": 0.0017994431545957923,
"signal/accgated_coverage_25/centered_abs_mean": 0.018403842113912106,
"signal/accgated_coverage_25/group_std_mean": 0.023735282942652702,
"signal/accgated_coverage_25/group_zero_std_frac": 0.0,
"signal/accgated_coverage_25/scaled_weighted_centered_abs_mean": 0.03545524738729,
"signal/accgated_coverage_25/weight": 0.10000000149011612,
"signal/accgated_coverage_25/weighted_centered_abs_mean": 0.0018403843278065324,
"signal/accgated_coverage_5/centered_abs_mean": 0.04615779221057892,
"signal/accgated_coverage_5/group_std_mean": 0.05862509831786156,
"signal/accgated_coverage_5/group_zero_std_frac": 0.0,
"signal/accgated_coverage_5/scaled_weighted_centered_abs_mean": 0.08909168839454651,
"signal/accgated_coverage_5/weight": 0.10000000149011612,
"signal/accgated_coverage_5/weighted_centered_abs_mean": 0.004615779500454664,
"signal/accuracy_reward/centered_abs_mean": 0.0850372314453125,
"signal/accuracy_reward/group_std_mean": 0.1174321100115776,
"signal/accuracy_reward/group_zero_std_frac": 0.640625,
"signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.8238133192062378,
"signal/accuracy_reward/weight": 0.5,
"signal/accuracy_reward/weighted_centered_abs_mean": 0.04251861572265625,
"signal/advantage_abs_mean": 0.7755907475948334,
"signal/advantage_pre_scale_abs_mean": 0.05832046829164028,
"signal/advantage_pre_scale_std": 0.09210299327969551,
"signal/advantage_std": 0.9825921952724457,
"signal/brier_reward/centered_abs_mean": 0.09697600081562996,
"signal/brier_reward/group_std_mean": 0.12307576835155487,
"signal/brier_reward/group_zero_std_frac": 0.0,
"signal/brier_reward/scaled_weighted_centered_abs_mean": 0.18729493767023087,
"signal/brier_reward/weight": 0.10000000149011612,
"signal/brier_reward/weighted_centered_abs_mean": 0.009697599802166224,
"signal/confidence_uniqueness_reward/centered_abs_mean": 0.011839714366942644,
"signal/confidence_uniqueness_reward/group_std_mean": 0.015590203925967216,
"signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0,
"signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.022851058281958103,
"signal/confidence_uniqueness_reward/weight": 0.10000000149011612,
"signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001183971471618861,
"signal/format_reward/centered_abs_mean": 0.0004730224609375,
"signal/format_reward/group_std_mean": 0.0013810679083690047,
"signal/format_reward/group_zero_std_frac": 0.9921875,
"signal/format_reward/scaled_weighted_centered_abs_mean": 0.00465081911534071,
"signal/format_reward/weight": 0.5,
"signal/format_reward/weighted_centered_abs_mean": 0.00023651123046875,
"signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26447156071662903,
"signal/frontier_entropy_batch_reward/group_std_mean": 0.34236086905002594,
"signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0,
"signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.5100432634353638,
"signal/frontier_entropy_batch_reward/weight": 0.10000000149011612,
"signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026447158306837082,
"step": 312,
"total_flos": 0.0,
"train_loss": -0.010174515421684377,
"train_runtime": 65977.4028,
"train_samples_per_second": 0.303,
"train_steps_per_second": 0.005
}
],
"logging_steps": 5,
"max_steps": 312,
"num_input_tokens_seen": 1078816210,
"num_train_epochs": 1,
"save_steps": 60,
"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
}