{ "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": 0.6528365431216747, "calibration/batch_distribution_entropy": 0.6504403816717517, "calibration/confidence_entropy": 0.3485475713141097, "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/ece": 0.5135036252491615, "calibration/mean_confidence": 0.7932241738665146, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.034375, "completions/max_length": 1495.0, "completions/max_terminated_length": 1495.0, "completions/mean_length": 214.21494140625, "completions/mean_terminated_length": 221.82879943847655, "completions/min_length": 0.0, "completions/min_terminated_length": 2.0, "epoch": 0.016, "grad_norm": 0.029577190056443214, "learning_rate": 3.1249999999999997e-07, "loss": 0.0071, "num_tokens": 17037593.0, "reward": 0.5223751187324523, "reward_std": 0.395353239774704, "rewards/accuracy_reward": 0.21650390625, "rewards/batch_coverage_0": 0.054699184745550154, "rewards/batch_coverage_1": 0.054699184745550154, "rewards/batch_coverage_5": 0.06712760254740716, "rewards/brier_reward": 0.37456473112106325, "rewards/confidence_uniqueness_reward": 0.4886078596115112, "rewards/format_reward": 0.68427734375, "rewards/frontier_aurc_reward": 0.2997752159833908, "rewards/frontier_coverage_10": 0.2997752159833908, "rewards/frontier_coverage_15": 0.2997752159833908, "rewards/frontier_coverage_20": 0.2997752159833908, "rewards/frontier_coverage_25": 0.2997752159833908, "rewards/frontier_ece_reward": 0.2997752159833908, "rewards/frontier_entropy_batch_reward": -0.6525280594825744, "signal/accuracy_reward/centered_abs_mean": 0.235845947265625, "signal/accuracy_reward/group_std_mean": 0.2778962790966034, "signal/accuracy_reward/group_zero_std_frac": 0.325, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.1179229736328125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.1179229736328125, "signal/advantage_abs_mean": 0.33545403480529784, "signal/advantage_pre_scale_abs_mean": 0.33545403480529784, "signal/advantage_pre_scale_std": 0.40500656962394715, "signal/advantage_std": 0.40500656962394715, "signal/batch_coverage_0/centered_abs_mean": 0.07738140523433686, "signal/batch_coverage_0/group_std_mean": 0.12824641913175583, "signal/batch_coverage_0/group_zero_std_frac": 0.003125, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0009672675398178399, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0009672675398178399, "signal/batch_coverage_1/centered_abs_mean": 0.07738140523433686, "signal/batch_coverage_1/group_std_mean": 0.12824641913175583, "signal/batch_coverage_1/group_zero_std_frac": 0.003125, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0009672675398178399, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0009672675398178399, "signal/batch_coverage_5/centered_abs_mean": 0.08599689900875092, "signal/batch_coverage_5/group_std_mean": 0.13797477185726165, "signal/batch_coverage_5/group_zero_std_frac": 0.003125, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0010749612585641444, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0010749612585641444, "signal/brier_reward/centered_abs_mean": 0.3168258905410767, "signal/brier_reward/group_std_mean": 0.36247108578681947, "signal/brier_reward/group_zero_std_frac": 0.003125, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.03168258890509605, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.03168258890509605, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.29753745198249815, "signal/confidence_uniqueness_reward/group_std_mean": 0.3484658181667328, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.02975374534726143, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.02975374534726143, "signal/format_reward/centered_abs_mean": 0.401458740234375, "signal/format_reward/group_std_mean": 0.45175864100456237, "signal/format_reward/group_zero_std_frac": 0.003125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.2007293701171875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.2007293701171875, "signal/frontier_aurc_reward/centered_abs_mean": 0.29053200483322145, "signal/frontier_aurc_reward/group_std_mean": 0.3419479191303253, "signal/frontier_aurc_reward/group_zero_std_frac": 0.003125, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 0.0036316501442342996, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 0.0036316501442342996, "signal/frontier_coverage_10/centered_abs_mean": 0.29053200483322145, "signal/frontier_coverage_10/group_std_mean": 0.3419479191303253, "signal/frontier_coverage_10/group_zero_std_frac": 0.003125, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0036316501442342996, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0036316501442342996, "signal/frontier_coverage_15/centered_abs_mean": 0.29053200483322145, "signal/frontier_coverage_15/group_std_mean": 0.3419479191303253, "signal/frontier_coverage_15/group_zero_std_frac": 0.003125, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0036316501442342996, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0036316501442342996, "signal/frontier_coverage_20/centered_abs_mean": 0.29053200483322145, "signal/frontier_coverage_20/group_std_mean": 0.3419479191303253, "signal/frontier_coverage_20/group_zero_std_frac": 0.003125, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0036316501442342996, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0036316501442342996, "signal/frontier_coverage_25/centered_abs_mean": 0.29053200483322145, "signal/frontier_coverage_25/group_std_mean": 0.3419479191303253, "signal/frontier_coverage_25/group_zero_std_frac": 0.003125, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0036316501442342996, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0036316501442342996, "signal/frontier_ece_reward/centered_abs_mean": 0.29053200483322145, "signal/frontier_ece_reward/group_std_mean": 0.3419479191303253, "signal/frontier_ece_reward/group_zero_std_frac": 0.003125, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.029053201153874397, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.029053201153874397, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.4259944140911102, "signal/frontier_entropy_batch_reward/group_std_mean": 0.47283905148506167, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0425994411110878, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0425994411110878, "step": 5 }, { "calibration/aurc": 0.6652684193150369, "calibration/batch_distribution_entropy": 0.6638764889598275, "calibration/confidence_entropy": 0.3446802269995745, "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/ece": 0.522101985321513, "calibration/mean_confidence": 0.7831486012756634, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02998046875, "completions/max_length": 1501.2, "completions/max_terminated_length": 1501.2, "completions/mean_length": 208.8384765625, "completions/mean_terminated_length": 215.32691345214843, "completions/min_length": 0.0, "completions/min_terminated_length": 2.0, "epoch": 0.032, "grad_norm": 0.025925613939762115, "learning_rate": 6.249999999999999e-07, "loss": 0.0082, "num_tokens": 34276451.0, "reward": 0.5391811728477478, "reward_std": 0.37500656843185426, "rewards/accuracy_reward": 0.21044921875, "rewards/batch_coverage_0": 0.0569451816380024, "rewards/batch_coverage_1": 0.0569451816380024, "rewards/batch_coverage_5": 0.06479373648762703, "rewards/brier_reward": 0.38155901432037354, "rewards/confidence_uniqueness_reward": 0.5201361238956451, "rewards/format_reward": 0.7228515625, "rewards/frontier_aurc_reward": 0.3006913423538208, "rewards/frontier_coverage_10": 0.3006913423538208, "rewards/frontier_coverage_15": 0.3006913423538208, "rewards/frontier_coverage_20": 0.3006913423538208, "rewards/frontier_coverage_25": 0.3006913423538208, "rewards/frontier_ece_reward": 0.3006913423538208, "rewards/frontier_entropy_batch_reward": -0.6873462796211243, "signal/accuracy_reward/centered_abs_mean": 0.220037841796875, "signal/accuracy_reward/group_std_mean": 0.2649418592453003, "signal/accuracy_reward/group_zero_std_frac": 0.33125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.1100189208984375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.1100189208984375, "signal/advantage_abs_mean": 0.3092048764228821, "signal/advantage_pre_scale_abs_mean": 0.3092048764228821, "signal/advantage_pre_scale_std": 0.3854846656322479, "signal/advantage_std": 0.3854846656322479, "signal/batch_coverage_0/centered_abs_mean": 0.07865111678838729, "signal/batch_coverage_0/group_std_mean": 0.12817984372377395, "signal/batch_coverage_0/group_zero_std_frac": 0.003125, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.000983138952869922, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.000983138952869922, "signal/batch_coverage_1/centered_abs_mean": 0.07865111678838729, "signal/batch_coverage_1/group_std_mean": 0.12817984372377395, "signal/batch_coverage_1/group_zero_std_frac": 0.003125, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.000983138952869922, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.000983138952869922, "signal/batch_coverage_5/centered_abs_mean": 0.08280284851789474, "signal/batch_coverage_5/group_std_mean": 0.13257062435150146, "signal/batch_coverage_5/group_zero_std_frac": 0.003125, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.001035035611130297, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.001035035611130297, "signal/brier_reward/centered_abs_mean": 0.30248112678527833, "signal/brier_reward/group_std_mean": 0.3510852336883545, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.030248113721609116, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.030248113721609116, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.2778193086385727, "signal/confidence_uniqueness_reward/group_std_mean": 0.33605759143829345, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.027781930938363075, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.027781930938363075, "signal/format_reward/centered_abs_mean": 0.36656494140625, "signal/format_reward/group_std_mean": 0.4304173469543457, "signal/format_reward/group_zero_std_frac": 0.0, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.183282470703125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.183282470703125, "signal/frontier_aurc_reward/centered_abs_mean": 0.2775486588478088, "signal/frontier_aurc_reward/group_std_mean": 0.33209188580513, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 0.003469358244910836, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 0.003469358244910836, "signal/frontier_coverage_10/centered_abs_mean": 0.2775486588478088, "signal/frontier_coverage_10/group_std_mean": 0.33209188580513, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.003469358244910836, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.003469358244910836, "signal/frontier_coverage_15/centered_abs_mean": 0.2775486588478088, "signal/frontier_coverage_15/group_std_mean": 0.33209188580513, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.003469358244910836, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.003469358244910836, "signal/frontier_coverage_20/centered_abs_mean": 0.2775486588478088, "signal/frontier_coverage_20/group_std_mean": 0.33209188580513, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.003469358244910836, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.003469358244910836, "signal/frontier_coverage_25/centered_abs_mean": 0.2775486588478088, "signal/frontier_coverage_25/group_std_mean": 0.33209188580513, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.003469358244910836, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.003469358244910836, "signal/frontier_ece_reward/centered_abs_mean": 0.2775486588478088, "signal/frontier_ece_reward/group_std_mean": 0.33209188580513, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.02775486595928669, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.02775486595928669, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.3982545256614685, "signal/frontier_entropy_batch_reward/group_std_mean": 0.45577409863471985, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.03982545360922814, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.03982545360922814, "step": 10 }, { "calibration/aurc": 0.6049888831291755, "calibration/batch_distribution_entropy": 0.6394673167151603, "calibration/buffer_distribution_entropy": 0.668499663871192, "calibration/confidence_entropy": 0.33684570655177054, "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/ece": 0.4769498508426146, "calibration/mean_confidence": 0.8059454365796774, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.01943359375, "completions/max_length": 1481.0, "completions/max_terminated_length": 1481.0, "completions/mean_length": 172.1814453125, "completions/mean_terminated_length": 175.6769805908203, "completions/min_length": 0.0, "completions/min_terminated_length": 2.2, "epoch": 0.048, "grad_norm": 0.01964188925921917, "learning_rate": 9.374999999999999e-07, "loss": 0.0028, "num_tokens": 51088325.0, "reward": 0.6567001819610596, "reward_std": 0.29598453640937805, "rewards/accuracy_reward": 0.27041015625, "rewards/batch_coverage_0": 0.08630179464817048, "rewards/batch_coverage_1": 0.08630179464817048, "rewards/batch_coverage_5": 0.10200787782669067, "rewards/brier_reward": 0.47995066046714785, "rewards/confidence_uniqueness_reward": 0.6418472051620483, "rewards/format_reward": 0.88330078125, "rewards/frontier_aurc_reward": 0.296812772937119, "rewards/frontier_coverage_10": 0.31256224811077116, "rewards/frontier_coverage_15": 0.31256224811077116, "rewards/frontier_coverage_20": 0.31256224811077116, "rewards/frontier_coverage_25": 0.31256224811077116, "rewards/frontier_ece_reward": 0.28651840090751646, "rewards/frontier_entropy_batch_reward": -0.8375783324241638, "signal/accuracy_reward/centered_abs_mean": 0.198297119140625, "signal/accuracy_reward/group_std_mean": 0.24569556713104249, "signal/accuracy_reward/group_zero_std_frac": 0.3625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0991485595703125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0991485595703125, "signal/advantage_abs_mean": 0.22403306663036346, "signal/advantage_pre_scale_abs_mean": 0.22403306663036346, "signal/advantage_pre_scale_std": 0.307987841963768, "signal/advantage_std": 0.307987841963768, "signal/batch_coverage_0/centered_abs_mean": 0.10079470723867416, "signal/batch_coverage_0/group_std_mean": 0.15550697445869446, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.00125993387773633, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.00125993387773633, "signal/batch_coverage_1/centered_abs_mean": 0.10079470723867416, "signal/batch_coverage_1/group_std_mean": 0.15550697445869446, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.00125993387773633, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.00125993387773633, "signal/batch_coverage_5/centered_abs_mean": 0.10616336464881897, "signal/batch_coverage_5/group_std_mean": 0.161780846118927, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0013270420487970114, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0013270420487970114, "signal/brier_reward/centered_abs_mean": 0.2741372585296631, "signal/brier_reward/group_std_mean": 0.3282070577144623, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.02741372548043728, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.02741372548043728, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.1985466718673706, "signal/confidence_uniqueness_reward/group_std_mean": 0.26411537528038026, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.019854667782783508, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.019854667782783508, "signal/format_reward/centered_abs_mean": 0.193170166015625, "signal/format_reward/group_std_mean": 0.294920951128006, "signal/format_reward/group_zero_std_frac": 0.05625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0965850830078125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0965850830078125, "signal/frontier_aurc_reward/centered_abs_mean": 0.21754129705950617, "signal/frontier_aurc_reward/group_std_mean": 0.262114205211401, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 0.00271926637069555, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 0.00271926637069555, "signal/frontier_coverage_10/centered_abs_mean": 0.23712547421455382, "signal/frontier_coverage_10/group_std_mean": 0.2920158565044403, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.002964068599976599, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.002964068599976599, "signal/frontier_coverage_15/centered_abs_mean": 0.23712547421455382, "signal/frontier_coverage_15/group_std_mean": 0.2920158565044403, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002964068599976599, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002964068599976599, "signal/frontier_coverage_20/centered_abs_mean": 0.23712547421455382, "signal/frontier_coverage_20/group_std_mean": 0.2920158565044403, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002964068599976599, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002964068599976599, "signal/frontier_coverage_25/centered_abs_mean": 0.23712547421455382, "signal/frontier_coverage_25/group_std_mean": 0.2920158565044403, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.002964068599976599, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.002964068599976599, "signal/frontier_ece_reward/centered_abs_mean": 0.2458635687828064, "signal/frontier_ece_reward/group_std_mean": 0.29540451765060427, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.024586357921361924, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.024586357921361924, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.25697159171104433, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3632834553718567, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0125, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.025697159022092818, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025697159022092818, "step": 15 }, { "calibration/aurc": 0.5352051226031345, "calibration/batch_distribution_entropy": 0.7138786257256148, "calibration/buffer_distribution_entropy": 0.666008141017535, "calibration/confidence_entropy": 0.3719886738659953, "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/ece": 0.36500567161410663, "calibration/mean_confidence": 0.7703579009131649, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0056640625, "completions/max_length": 1295.6, "completions/max_terminated_length": 1295.6, "completions/mean_length": 133.05732421875, "completions/mean_terminated_length": 133.8147399902344, "completions/min_length": 0.0, "completions/min_terminated_length": 22.4, "epoch": 0.064, "grad_norm": 0.0575578399002552, "learning_rate": 1e-06, "loss": -0.0017, "num_tokens": 67369232.0, "reward": 0.6984567999839782, "reward_std": 0.20601321756839752, "rewards/accuracy_reward": 0.3392578125, "rewards/batch_coverage_0": 0.08189218938350677, "rewards/batch_coverage_1": 0.08189218938350677, "rewards/batch_coverage_5": 0.10462159365415573, "rewards/brier_reward": 0.567774772644043, "rewards/confidence_uniqueness_reward": 0.73758544921875, "rewards/format_reward": 0.97353515625, "rewards/frontier_aurc_reward": -0.006997331976890564, "rewards/frontier_coverage_10": 0.06370379328727722, "rewards/frontier_coverage_15": 0.06370379328727722, "rewards/frontier_coverage_20": 0.06370379328727722, "rewards/frontier_coverage_25": 0.06370379328727722, "rewards/frontier_ece_reward": -0.04985757041722536, "rewards/frontier_entropy_batch_reward": -0.8994276046752929, "signal/accuracy_reward/centered_abs_mean": 0.2031982421875, "signal/accuracy_reward/group_std_mean": 0.2531350880861282, "signal/accuracy_reward/group_zero_std_frac": 0.34375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.10159912109375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.10159912109375, "signal/advantage_abs_mean": 0.15646747648715972, "signal/advantage_pre_scale_abs_mean": 0.15646747648715972, "signal/advantage_pre_scale_std": 0.22189362347126007, "signal/advantage_std": 0.22189362347126007, "signal/batch_coverage_0/centered_abs_mean": 0.11014256626367569, "signal/batch_coverage_0/group_std_mean": 0.16949896216392518, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0013767820782959461, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0013767820782959461, "signal/batch_coverage_1/centered_abs_mean": 0.11014256626367569, "signal/batch_coverage_1/group_std_mean": 0.16949896216392518, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0013767820782959461, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0013767820782959461, "signal/batch_coverage_5/centered_abs_mean": 0.11534451395273208, "signal/batch_coverage_5/group_std_mean": 0.1762994647026062, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0014418064383789897, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0014418064383789897, "signal/brier_reward/centered_abs_mean": 0.24833780527114868, "signal/brier_reward/group_std_mean": 0.3036909818649292, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.02483378052711487, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.02483378052711487, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.13121245801448822, "signal/confidence_uniqueness_reward/group_std_mean": 0.17133590281009675, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.013121245615184306, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.013121245615184306, "signal/format_reward/centered_abs_mean": 0.048956298828125, "signal/format_reward/group_std_mean": 0.11224002465605736, "signal/format_reward/group_zero_std_frac": 0.471875, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0244781494140625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0244781494140625, "signal/frontier_aurc_reward/centered_abs_mean": 0.0051182197406888005, "signal/frontier_aurc_reward/group_std_mean": 0.007049326226115227, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 6.397774486686104e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 6.397774486686104e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.10937852710485459, "signal/frontier_coverage_10/group_std_mean": 0.1701591819524765, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0013672316214069723, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0013672316214069723, "signal/frontier_coverage_15/centered_abs_mean": 0.10937852710485459, "signal/frontier_coverage_15/group_std_mean": 0.1701591819524765, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0013672316214069723, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0013672316214069723, "signal/frontier_coverage_20/centered_abs_mean": 0.10937852710485459, "signal/frontier_coverage_20/group_std_mean": 0.1701591819524765, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0013672316214069723, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0013672316214069723, "signal/frontier_coverage_25/centered_abs_mean": 0.10937852710485459, "signal/frontier_coverage_25/group_std_mean": 0.1701591819524765, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0013672316214069723, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0013672316214069723, "signal/frontier_ece_reward/centered_abs_mean": 0.14623606503009795, "signal/frontier_ece_reward/group_std_mean": 0.17439992129802703, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.014623606950044632, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.014623606950044632, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.17345191836357116, "signal/frontier_entropy_batch_reward/group_std_mean": 0.2986772358417511, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.096875, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.01734519228339195, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.01734519228339195, "step": 20 }, { "calibration/aurc": 0.6172633570224754, "calibration/batch_distribution_entropy": 0.8339899041337885, "calibration/buffer_distribution_entropy": 0.6997543931890366, "calibration/confidence_entropy": 0.45167123075067, "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/ece": 0.3919382461290124, "calibration/mean_confidence": 0.7034418278152188, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 806.6, "completions/max_terminated_length": 806.6, "completions/mean_length": 116.523046875, "completions/mean_terminated_length": 116.74108428955078, "completions/min_length": 0.0, "completions/min_terminated_length": 30.6, "epoch": 0.08, "grad_norm": 0.014022819697856903, "learning_rate": 1e-06, "loss": -0.0028, "num_tokens": 83495580.0, "reward": 0.7385941505432129, "reward_std": 0.17612878680229188, "rewards/accuracy_reward": 0.355078125, "rewards/batch_coverage_0": 0.10197570025920868, "rewards/batch_coverage_1": 0.10197570025920868, "rewards/batch_coverage_5": 0.1384865313768387, "rewards/brier_reward": 0.6277936935424805, "rewards/confidence_uniqueness_reward": 0.8350537300109864, "rewards/format_reward": 0.991796875, "rewards/frontier_aurc_reward": -0.005960373487323522, "rewards/frontier_coverage_10": 0.0744130864739418, "rewards/frontier_coverage_15": 0.0744130864739418, "rewards/frontier_coverage_20": 0.0744130864739418, "rewards/frontier_coverage_25": 0.0744130864739418, "rewards/frontier_ece_reward": -0.040788016840815546, "rewards/frontier_entropy_batch_reward": -0.8497591137886047, "signal/accuracy_reward/centered_abs_mean": 0.18973388671875, "signal/accuracy_reward/group_std_mean": 0.2429521083831787, "signal/accuracy_reward/group_zero_std_frac": 0.3375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.094866943359375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.094866943359375, "signal/advantage_abs_mean": 0.13535310328006744, "signal/advantage_pre_scale_abs_mean": 0.13535310328006744, "signal/advantage_pre_scale_std": 0.19212235808372496, "signal/advantage_std": 0.19212235808372496, "signal/batch_coverage_0/centered_abs_mean": 0.14148729294538498, "signal/batch_coverage_0/group_std_mean": 0.20446310043334961, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.001768591208383441, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.001768591208383441, "signal/batch_coverage_1/centered_abs_mean": 0.14148729294538498, "signal/batch_coverage_1/group_std_mean": 0.20446310043334961, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.001768591208383441, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.001768591208383441, "signal/batch_coverage_5/centered_abs_mean": 0.15119329690933228, "signal/batch_coverage_5/group_std_mean": 0.21579791009426116, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.001889916229993105, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.001889916229993105, "signal/brier_reward/centered_abs_mean": 0.22526630461215974, "signal/brier_reward/group_std_mean": 0.27675821185112, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.022526631504297255, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.022526631504297255, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.07187793254852295, "signal/confidence_uniqueness_reward/group_std_mean": 0.1005837857723236, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.007187793496996164, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.007187793496996164, "signal/format_reward/centered_abs_mean": 0.01546630859375, "signal/format_reward/group_std_mean": 0.039999290555715564, "signal/format_reward/group_zero_std_frac": 0.790625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.007733154296875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.007733154296875, "signal/frontier_aurc_reward/centered_abs_mean": 0.0038858432322740554, "signal/frontier_aurc_reward/group_std_mean": 0.005908959917724133, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 4.857304156757891e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 4.857304156757891e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.14614013433456421, "signal/frontier_coverage_10/group_std_mean": 0.21353738009929657, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0018267517210915686, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0018267517210915686, "signal/frontier_coverage_15/centered_abs_mean": 0.14614013433456421, "signal/frontier_coverage_15/group_std_mean": 0.21353738009929657, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0018267517210915686, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0018267517210915686, "signal/frontier_coverage_20/centered_abs_mean": 0.14614013433456421, "signal/frontier_coverage_20/group_std_mean": 0.21353738009929657, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0018267517210915686, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0018267517210915686, "signal/frontier_coverage_25/centered_abs_mean": 0.14614013433456421, "signal/frontier_coverage_25/group_std_mean": 0.21353738009929657, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0018267517210915686, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0018267517210915686, "signal/frontier_ece_reward/centered_abs_mean": 0.1286403000354767, "signal/frontier_ece_reward/group_std_mean": 0.15866474211215972, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.012864029593765735, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.012864029593765735, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.24675768315792085, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3810174763202667, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.03125, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02467576749622822, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02467576749622822, "step": 25 }, { "calibration/aurc": 0.633727663732538, "calibration/batch_distribution_entropy": 0.9509955867294941, "calibration/buffer_distribution_entropy": 0.7582111631112705, "calibration/confidence_entropy": 0.5070069103263735, "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/ece": 0.2808798025077909, "calibration/mean_confidence": 0.5523479158013329, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00126953125, "completions/max_length": 830.8, "completions/max_terminated_length": 830.8, "completions/mean_length": 105.15400390625, "completions/mean_terminated_length": 105.28469848632812, "completions/min_length": 0.0, "completions/min_terminated_length": 30.4, "epoch": 0.096, "grad_norm": 0.0034083914943039417, "learning_rate": 1e-06, "loss": -0.002, "num_tokens": 99616965.0, "reward": 0.7852896571159362, "reward_std": 0.1621668130159378, "rewards/accuracy_reward": 0.3599609375, "rewards/batch_coverage_0": 0.15466825067996978, "rewards/batch_coverage_1": 0.15466825067996978, "rewards/batch_coverage_5": 0.18448047637939452, "rewards/brier_reward": 0.682706093788147, "rewards/confidence_uniqueness_reward": 0.9147889256477356, "rewards/format_reward": 0.995703125, "rewards/frontier_aurc_reward": -0.005372170079499483, "rewards/frontier_coverage_10": 0.10586128979921341, "rewards/frontier_coverage_15": 0.10586128979921341, "rewards/frontier_coverage_20": 0.10586128979921341, "rewards/frontier_coverage_25": 0.10586128979921341, "rewards/frontier_ece_reward": -0.029005835950374603, "rewards/frontier_entropy_batch_reward": -0.6078990817070007, "signal/accuracy_reward/centered_abs_mean": 0.18262939453125, "signal/accuracy_reward/group_std_mean": 0.22975040972232819, "signal/accuracy_reward/group_zero_std_frac": 0.390625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.091314697265625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.091314697265625, "signal/advantage_abs_mean": 0.12659611254930497, "signal/advantage_pre_scale_abs_mean": 0.12659611254930497, "signal/advantage_pre_scale_std": 0.17533020675182343, "signal/advantage_std": 0.17533020675182343, "signal/batch_coverage_0/centered_abs_mean": 0.21107754111289978, "signal/batch_coverage_0/group_std_mean": 0.2733610153198242, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0026384693570435046, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0026384693570435046, "signal/batch_coverage_1/centered_abs_mean": 0.21107754111289978, "signal/batch_coverage_1/group_std_mean": 0.2733610153198242, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0026384693570435046, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0026384693570435046, "signal/batch_coverage_5/centered_abs_mean": 0.2218673586845398, "signal/batch_coverage_5/group_std_mean": 0.285660994052887, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0027733420953154563, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0027733420953154563, "signal/brier_reward/centered_abs_mean": 0.22906748056411744, "signal/brier_reward/group_std_mean": 0.27832997441291807, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.022906748205423356, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.022906748205423356, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.04866270124912262, "signal/confidence_uniqueness_reward/group_std_mean": 0.07076105996966361, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.004866270301863551, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004866270301863551, "signal/format_reward/centered_abs_mean": 0.00828857421875, "signal/format_reward/group_std_mean": 0.02329786978662014, "signal/format_reward/group_zero_std_frac": 0.871875, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.004144287109375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.004144287109375, "signal/frontier_aurc_reward/centered_abs_mean": 0.0026717749424278736, "signal/frontier_aurc_reward/group_std_mean": 0.004080273536965251, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.339718750794418e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.339718750794418e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.2277726113796234, "signal/frontier_coverage_10/group_std_mean": 0.29606238305568694, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.002847157744690776, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.002847157744690776, "signal/frontier_coverage_15/centered_abs_mean": 0.2277726113796234, "signal/frontier_coverage_15/group_std_mean": 0.29606238305568694, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002847157744690776, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002847157744690776, "signal/frontier_coverage_20/centered_abs_mean": 0.2277726113796234, "signal/frontier_coverage_20/group_std_mean": 0.29606238305568694, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002847157744690776, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002847157744690776, "signal/frontier_coverage_25/centered_abs_mean": 0.2277726113796234, "signal/frontier_coverage_25/group_std_mean": 0.29606238305568694, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.002847157744690776, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.002847157744690776, "signal/frontier_ece_reward/centered_abs_mean": 0.10935526341199875, "signal/frontier_ece_reward/group_std_mean": 0.14532690048217772, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.010935526341199875, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.010935526341199875, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.4438786208629608, "signal/frontier_entropy_batch_reward/group_std_mean": 0.5260465621948243, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0443878635764122, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0443878635764122, "step": 30 }, { "calibration/aurc": 0.5202571097103064, "calibration/batch_distribution_entropy": 0.9430957297684482, "calibration/buffer_distribution_entropy": 0.8389954532230423, "calibration/confidence_entropy": 0.4974867099038153, "calibration/coverage@0%": 0.0, "calibration/coverage@1%": 0.0, "calibration/coverage@10%": 0.0, "calibration/coverage@15%": 0.0, "calibration/coverage@20%": 0.00392156862745098, "calibration/coverage@25%": 0.00392156862745098, "calibration/coverage@30%": 0.033725490196078435, "calibration/coverage@5%": 0.0, "calibration/ece": 0.19978526084483866, "calibration/mean_confidence": 0.3805440661690666, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00244140625, "completions/max_length": 591.6, "completions/max_terminated_length": 591.6, "completions/mean_length": 101.0359375, "completions/mean_terminated_length": 101.28242797851563, "completions/min_length": 0.0, "completions/min_terminated_length": 39.0, "epoch": 0.112, "grad_norm": 0.0025749315973371267, "learning_rate": 1e-06, "loss": -0.0016, "num_tokens": 115761045.0, "reward": 0.8336851239204407, "reward_std": 0.13831743001937866, "rewards/accuracy_reward": 0.38740234375, "rewards/batch_coverage_0": 0.2132992446422577, "rewards/batch_coverage_1": 0.2132992446422577, "rewards/batch_coverage_5": 0.2325060784816742, "rewards/brier_reward": 0.7129961490631104, "rewards/confidence_uniqueness_reward": 0.9415440559387207, "rewards/format_reward": 0.9958984375, "rewards/frontier_aurc_reward": -0.004701448418200016, "rewards/frontier_coverage_10": 0.1297334685921669, "rewards/frontier_coverage_15": 0.1297334685921669, "rewards/frontier_coverage_20": 0.1297334685921669, "rewards/frontier_coverage_25": 0.1297334685921669, "rewards/frontier_ece_reward": -0.00449736500158906, "rewards/frontier_entropy_batch_reward": -0.37636255025863646, "signal/accuracy_reward/centered_abs_mean": 0.186004638671875, "signal/accuracy_reward/group_std_mean": 0.23665676414966583, "signal/accuracy_reward/group_zero_std_frac": 0.35625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0930023193359375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0930023193359375, "signal/advantage_abs_mean": 0.10636455714702606, "signal/advantage_pre_scale_abs_mean": 0.10636455714702606, "signal/advantage_pre_scale_std": 0.15220401287078858, "signal/advantage_std": 0.15220401287078858, "signal/batch_coverage_0/centered_abs_mean": 0.2513628154993057, "signal/batch_coverage_0/group_std_mean": 0.31081662774086, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.00314203524030745, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.00314203524030745, "signal/batch_coverage_1/centered_abs_mean": 0.2513628154993057, "signal/batch_coverage_1/group_std_mean": 0.31081662774086, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.00314203524030745, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.00314203524030745, "signal/batch_coverage_5/centered_abs_mean": 0.25727218091487886, "signal/batch_coverage_5/group_std_mean": 0.3185530245304108, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.003215902438387275, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.003215902438387275, "signal/brier_reward/centered_abs_mean": 0.21385617852210997, "signal/brier_reward/group_std_mean": 0.2658445775508881, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.021385619416832924, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.021385619416832924, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.02815374471247196, "signal/confidence_uniqueness_reward/group_std_mean": 0.04658707231283188, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.002815374592319131, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.002815374592319131, "signal/format_reward/centered_abs_mean": 0.00791015625, "signal/format_reward/group_std_mean": 0.02219301536679268, "signal/format_reward/group_zero_std_frac": 0.878125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.003955078125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.003955078125, "signal/frontier_aurc_reward/centered_abs_mean": 0.0015520456014201044, "signal/frontier_aurc_reward/group_std_mean": 0.0025679299142211676, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 1.940057009051088e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 1.940057009051088e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.31559001803398135, "signal/frontier_coverage_10/group_std_mean": 0.38809067010879517, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.003944875253364444, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.003944875253364444, "signal/frontier_coverage_15/centered_abs_mean": 0.31559001803398135, "signal/frontier_coverage_15/group_std_mean": 0.38809067010879517, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.003944875253364444, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.003944875253364444, "signal/frontier_coverage_20/centered_abs_mean": 0.31559001803398135, "signal/frontier_coverage_20/group_std_mean": 0.38809067010879517, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.003944875253364444, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.003944875253364444, "signal/frontier_coverage_25/centered_abs_mean": 0.31559001803398135, "signal/frontier_coverage_25/group_std_mean": 0.38809067010879517, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.003944875253364444, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.003944875253364444, "signal/frontier_ece_reward/centered_abs_mean": 0.06571868434548378, "signal/frontier_ece_reward/group_std_mean": 0.09877323508262634, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.006571868434548378, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.006571868434548378, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.4296597421169281, "signal/frontier_entropy_batch_reward/group_std_mean": 0.4898779034614563, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.042965974658727646, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.042965974658727646, "step": 35 }, { "calibration/aurc": 0.5639305656958816, "calibration/batch_distribution_entropy": 0.9186417246944301, "calibration/buffer_distribution_entropy": 0.8989803262156284, "calibration/confidence_entropy": 0.4828925259188782, "calibration/coverage@0%": 0.0023498714554314876, "calibration/coverage@1%": 0.0023498714554314876, "calibration/coverage@10%": 0.0023498714554314876, "calibration/coverage@15%": 0.004310655769156978, "calibration/coverage@20%": 0.0047020452016423, "calibration/coverage@25%": 0.00665975979432869, "calibration/coverage@30%": 0.01135338424753463, "calibration/coverage@5%": 0.0023498714554314876, "calibration/ece": 0.20512244938093138, "calibration/mean_confidence": 0.3380585148631722, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0017578125, "completions/max_length": 808.4, "completions/max_terminated_length": 808.4, "completions/mean_length": 101.72802734375, "completions/mean_terminated_length": 101.90912780761718, "completions/min_length": 0.0, "completions/min_terminated_length": 39.6, "epoch": 0.128, "grad_norm": 0.0022302691359072924, "learning_rate": 1e-06, "loss": -0.0014, "num_tokens": 131719412.0, "reward": 0.8352240324020386, "reward_std": 0.12470821291208267, "rewards/accuracy_reward": 0.3888671875, "rewards/batch_coverage_0": 0.25125581324100493, "rewards/batch_coverage_1": 0.25125581324100493, "rewards/batch_coverage_5": 0.2699826031923294, "rewards/brier_reward": 0.717239499092102, "rewards/confidence_uniqueness_reward": 0.9374270796775818, "rewards/format_reward": 0.9962890625, "rewards/frontier_aurc_reward": -0.004471705388277769, "rewards/frontier_coverage_10": 0.1387957066297531, "rewards/frontier_coverage_15": 0.1387957066297531, "rewards/frontier_coverage_20": 0.1387957066297531, "rewards/frontier_coverage_25": 0.1387957066297531, "rewards/frontier_ece_reward": 0.0029714343952946367, "rewards/frontier_entropy_batch_reward": -0.3965793430805206, "signal/accuracy_reward/centered_abs_mean": 0.16063232421875, "signal/accuracy_reward/group_std_mean": 0.21138851046562196, "signal/accuracy_reward/group_zero_std_frac": 0.396875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.080316162109375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.080316162109375, "signal/advantage_abs_mean": 0.09423303902149201, "signal/advantage_pre_scale_abs_mean": 0.09423303902149201, "signal/advantage_pre_scale_std": 0.13935055136680602, "signal/advantage_std": 0.13935055136680602, "signal/batch_coverage_0/centered_abs_mean": 0.2517063319683075, "signal/batch_coverage_0/group_std_mean": 0.3126410245895386, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0031463291961699724, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0031463291961699724, "signal/batch_coverage_1/centered_abs_mean": 0.2517063319683075, "signal/batch_coverage_1/group_std_mean": 0.3126410245895386, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0031463291961699724, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0031463291961699724, "signal/batch_coverage_5/centered_abs_mean": 0.25431095361709594, "signal/batch_coverage_5/group_std_mean": 0.3168126463890076, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.003178886929526925, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.003178886929526925, "signal/brier_reward/centered_abs_mean": 0.2044551283121109, "signal/brier_reward/group_std_mean": 0.25647562444210054, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.020445513352751733, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.020445513352751733, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.026767711713910104, "signal/confidence_uniqueness_reward/group_std_mean": 0.04347552061080932, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.002676771255210042, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.002676771255210042, "signal/format_reward/centered_abs_mean": 0.00716552734375, "signal/format_reward/group_std_mean": 0.02031961511820555, "signal/format_reward/group_zero_std_frac": 0.8875, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.003582763671875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.003582763671875, "signal/frontier_aurc_reward/centered_abs_mean": 0.0014250480569899082, "signal/frontier_aurc_reward/group_std_mean": 0.002257279586046934, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 1.7813101294450462e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 1.7813101294450462e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.3143028438091278, "signal/frontier_coverage_10/group_std_mean": 0.3871733367443085, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.003928785584867001, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.003928785584867001, "signal/frontier_coverage_15/centered_abs_mean": 0.3143028438091278, "signal/frontier_coverage_15/group_std_mean": 0.3871733367443085, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.003928785584867001, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.003928785584867001, "signal/frontier_coverage_20/centered_abs_mean": 0.3143028438091278, "signal/frontier_coverage_20/group_std_mean": 0.3871733367443085, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.003928785584867001, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.003928785584867001, "signal/frontier_coverage_25/centered_abs_mean": 0.3143028438091278, "signal/frontier_coverage_25/group_std_mean": 0.3871733367443085, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.003928785584867001, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.003928785584867001, "signal/frontier_ece_reward/centered_abs_mean": 0.05491202026605606, "signal/frontier_ece_reward/group_std_mean": 0.08285191357135772, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.005491202138364315, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.005491202138364315, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.4243355572223663, "signal/frontier_entropy_batch_reward/group_std_mean": 0.4852792859077454, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.04243355467915535, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.04243355467915535, "step": 40 }, { "calibration/aurc": 0.4154914142086052, "calibration/batch_distribution_entropy": 0.9833036512104236, "calibration/buffer_distribution_entropy": 0.9302459812552428, "calibration/confidence_entropy": 0.523829010994173, "calibration/coverage@0%": 0.0035294358885286396, "calibration/coverage@1%": 0.0035294358885286396, "calibration/coverage@10%": 0.01134193588852864, "calibration/coverage@15%": 0.038299656476763935, "calibration/coverage@20%": 0.07423715647676393, "calibration/coverage@25%": 0.09689493833950903, "calibration/coverage@30%": 0.22477999073645205, "calibration/coverage@5%": 0.01095131088852864, "calibration/ece": 0.2092989795907501, "calibration/mean_confidence": 0.4798063499165311, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00166015625, "completions/max_length": 657.0, "completions/max_terminated_length": 657.0, "completions/mean_length": 107.8806640625, "completions/mean_terminated_length": 108.06053771972657, "completions/min_length": 0.0, "completions/min_terminated_length": 46.4, "epoch": 0.144, "grad_norm": 0.002824458759278059, "learning_rate": 1e-06, "loss": -0.0017, "num_tokens": 147774542.0, "reward": 0.8979710221290589, "reward_std": 0.1393474370241165, "rewards/accuracy_reward": 0.49228515625, "rewards/batch_coverage_0": 0.17017147541046143, "rewards/batch_coverage_1": 0.17017147541046143, "rewards/batch_coverage_5": 0.1970554769039154, "rewards/brier_reward": 0.6984734654426574, "rewards/confidence_uniqueness_reward": 0.950539493560791, "rewards/format_reward": 0.99677734375, "rewards/frontier_aurc_reward": -0.004099717363715172, "rewards/frontier_coverage_10": 0.028359479270875454, "rewards/frontier_coverage_15": 0.028359479270875454, "rewards/frontier_coverage_20": 0.028359479270875454, "rewards/frontier_coverage_25": 0.028359479270875454, "rewards/frontier_ece_reward": 0.003376691718585789, "rewards/frontier_entropy_batch_reward": -0.1988338440656662, "signal/accuracy_reward/centered_abs_mean": 0.172650146484375, "signal/accuracy_reward/group_std_mean": 0.22143488228321076, "signal/accuracy_reward/group_zero_std_frac": 0.390625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0863250732421875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0863250732421875, "signal/advantage_abs_mean": 0.10827811062335968, "signal/advantage_pre_scale_abs_mean": 0.10827811062335968, "signal/advantage_pre_scale_std": 0.152878013253212, "signal/advantage_std": 0.152878013253212, "signal/batch_coverage_0/centered_abs_mean": 0.23307644426822663, "signal/batch_coverage_0/group_std_mean": 0.2904390752315521, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002913455618545413, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002913455618545413, "signal/batch_coverage_1/centered_abs_mean": 0.23307644426822663, "signal/batch_coverage_1/group_std_mean": 0.2904390752315521, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002913455618545413, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002913455618545413, "signal/batch_coverage_5/centered_abs_mean": 0.23927061259746552, "signal/batch_coverage_5/group_std_mean": 0.29758256673812866, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.00299088261090219, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.00299088261090219, "signal/brier_reward/centered_abs_mean": 0.21872566044330596, "signal/brier_reward/group_std_mean": 0.26644076108932496, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.021872566640377046, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.021872566640377046, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.017737414687871933, "signal/confidence_uniqueness_reward/group_std_mean": 0.031206076219677924, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0017737414687871932, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0017737414687871932, "signal/format_reward/centered_abs_mean": 0.006207275390625, "signal/format_reward/group_std_mean": 0.017221171595156192, "signal/format_reward/group_zero_std_frac": 0.90625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0031036376953125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0031036376953125, "signal/frontier_aurc_reward/centered_abs_mean": 0.0021680801641196013, "signal/frontier_aurc_reward/group_std_mean": 0.0031417491380125286, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.7101004525320606e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.7101004525320606e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.26910161077976225, "signal/frontier_coverage_10/group_std_mean": 0.33466690182685854, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0033637700602412225, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0033637700602412225, "signal/frontier_coverage_15/centered_abs_mean": 0.26910161077976225, "signal/frontier_coverage_15/group_std_mean": 0.33466690182685854, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0033637700602412225, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0033637700602412225, "signal/frontier_coverage_20/centered_abs_mean": 0.26910161077976225, "signal/frontier_coverage_20/group_std_mean": 0.33466690182685854, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0033637700602412225, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0033637700602412225, "signal/frontier_coverage_25/centered_abs_mean": 0.26910161077976225, "signal/frontier_coverage_25/group_std_mean": 0.33466690182685854, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0033637700602412225, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0033637700602412225, "signal/frontier_ece_reward/centered_abs_mean": 0.06440802365541458, "signal/frontier_ece_reward/group_std_mean": 0.09016914367675781, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.00644080201163888, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.00644080201163888, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28723893463611605, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3630897760391235, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02872389554977417, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02872389554977417, "step": 45 }, { "calibration/aurc": 0.49485550097123693, "calibration/batch_distribution_entropy": 0.9809334020823716, "calibration/buffer_distribution_entropy": 0.9450654151293921, "calibration/confidence_entropy": 0.5272734969224558, "calibration/coverage@0%": 0.00078431674093326, "calibration/coverage@1%": 0.00078431674093326, "calibration/coverage@10%": 0.00078431674093326, "calibration/coverage@15%": 0.00078431674093326, "calibration/coverage@20%": 0.0023498744708745516, "calibration/coverage@25%": 0.0023498744708745516, "calibration/coverage@30%": 0.007457929480697734, "calibration/coverage@5%": 0.00078431674093326, "calibration/ece": 0.18138254550070013, "calibration/mean_confidence": 0.5474443198503514, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 800.8, "completions/max_terminated_length": 800.8, "completions/mean_length": 121.887109375, "completions/mean_terminated_length": 122.02821350097656, "completions/min_length": 0.0, "completions/min_terminated_length": 50.0, "epoch": 0.16, "grad_norm": 0.0017115556402131915, "learning_rate": 1e-06, "loss": -0.002, "num_tokens": 164043594.0, "reward": 0.8776283621788025, "reward_std": 0.14146997332572936, "rewards/accuracy_reward": 0.44033203125, "rewards/batch_coverage_0": 0.1819753795862198, "rewards/batch_coverage_1": 0.1819753795862198, "rewards/batch_coverage_5": 0.2082147181034088, "rewards/brier_reward": 0.7022818803787232, "rewards/confidence_uniqueness_reward": 0.9528778910636901, "rewards/format_reward": 0.9982421875, "rewards/frontier_aurc_reward": -0.004562536813318729, "rewards/frontier_coverage_10": 0.06561319380998612, "rewards/frontier_coverage_15": 0.06561319380998612, "rewards/frontier_coverage_20": 0.06561319380998612, "rewards/frontier_coverage_25": 0.06561319380998612, "rewards/frontier_ece_reward": 0.0010979614977259188, "rewards/frontier_entropy_batch_reward": -0.17660210132598878, "signal/accuracy_reward/centered_abs_mean": 0.163360595703125, "signal/accuracy_reward/group_std_mean": 0.21059593558311462, "signal/accuracy_reward/group_zero_std_frac": 0.4125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0816802978515625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0816802978515625, "signal/advantage_abs_mean": 0.11046318858861923, "signal/advantage_pre_scale_abs_mean": 0.11046318858861923, "signal/advantage_pre_scale_std": 0.1564834088087082, "signal/advantage_std": 0.1564834088087082, "signal/batch_coverage_0/centered_abs_mean": 0.21572335362434386, "signal/batch_coverage_0/group_std_mean": 0.27485826015472414, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002696541976183653, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002696541976183653, "signal/batch_coverage_1/centered_abs_mean": 0.21572335362434386, "signal/batch_coverage_1/group_std_mean": 0.27485826015472414, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002696541976183653, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002696541976183653, "signal/batch_coverage_5/centered_abs_mean": 0.22292961478233336, "signal/batch_coverage_5/group_std_mean": 0.283195298910141, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0027866202872246504, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0027866202872246504, "signal/brier_reward/centered_abs_mean": 0.21594384610652922, "signal/brier_reward/group_std_mean": 0.2642689645290375, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.02159438580274582, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.02159438580274582, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.01417789962142706, "signal/confidence_uniqueness_reward/group_std_mean": 0.023011171817779542, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0014177899807691573, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014177899807691573, "signal/format_reward/centered_abs_mean": 0.0033935546875, "signal/format_reward/group_std_mean": 0.009607380582019686, "signal/format_reward/group_zero_std_frac": 0.946875, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00169677734375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00169677734375, "signal/frontier_aurc_reward/centered_abs_mean": 0.0028273236472159623, "signal/frontier_aurc_reward/group_std_mean": 0.004012005869299173, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.534154748194851e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.534154748194851e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.221804678440094, "signal/frontier_coverage_10/group_std_mean": 0.28661670088768004, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0027725585270673035, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0027725585270673035, "signal/frontier_coverage_15/centered_abs_mean": 0.221804678440094, "signal/frontier_coverage_15/group_std_mean": 0.28661670088768004, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0027725585270673035, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0027725585270673035, "signal/frontier_coverage_20/centered_abs_mean": 0.221804678440094, "signal/frontier_coverage_20/group_std_mean": 0.28661670088768004, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0027725585270673035, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0027725585270673035, "signal/frontier_coverage_25/centered_abs_mean": 0.221804678440094, "signal/frontier_coverage_25/group_std_mean": 0.28661670088768004, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0027725585270673035, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0027725585270673035, "signal/frontier_ece_reward/centered_abs_mean": 0.0714937224984169, "signal/frontier_ece_reward/group_std_mean": 0.09538796991109848, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.007149372342973947, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.007149372342973947, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26375205516815187, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34235548973083496, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026375206559896468, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026375206559896468, "step": 50 }, { "epoch": 0.16, "eval_calibration/aurc": 0.6977046381221063, "eval_calibration/batch_distribution_entropy": 0.9297253001835979, "eval_calibration/buffer_distribution_entropy": 0.9512632358171704, "eval_calibration/confidence_entropy": 0.4963632176432223, "eval_calibration/coverage@0%": 0.0234375, "eval_calibration/coverage@1%": 0.0234375, "eval_calibration/coverage@10%": 0.0234375, "eval_calibration/coverage@15%": 0.0234375, "eval_calibration/coverage@20%": 0.0234375, "eval_calibration/coverage@25%": 0.0234375, "eval_calibration/coverage@30%": 0.0234375, "eval_calibration/coverage@5%": 0.0234375, "eval_calibration/ece": 0.3679003958436029, "eval_calibration/mean_confidence": 0.5406141961402707, "eval_completions/clipped_ratio": 0.0, "eval_completions/max_length": 303.75, "eval_completions/max_terminated_length": 303.75, "eval_completions/mean_length": 138.17194366455078, "eval_completions/mean_terminated_length": 138.17194366455078, "eval_completions/min_length": 65.75, "eval_completions/min_terminated_length": 65.75, "eval_loss": 0.0, "eval_num_tokens": 164043594.0, "eval_reward": 0.7527717649936676, "eval_reward_std": 0.23287427425384521, "eval_rewards/accuracy_reward": 0.369140625, "eval_rewards/batch_coverage_0": 0.10159964673221111, "eval_rewards/batch_coverage_1": 0.10159964673221111, "eval_rewards/batch_coverage_5": 0.10159964673221111, "eval_rewards/brier_reward": 0.6937532722949982, "eval_rewards/confidence_uniqueness_reward": 0.898193359375, "eval_rewards/format_reward": 1.0, "eval_rewards/frontier_aurc_reward": -0.004968127002939582, "eval_rewards/frontier_coverage_10": 0.11007783003151417, "eval_rewards/frontier_coverage_15": 0.11007783003151417, "eval_rewards/frontier_coverage_20": 0.11007783003151417, "eval_rewards/frontier_coverage_25": 0.11007783003151417, "eval_rewards/frontier_ece_reward": -0.002450033149216324, "eval_rewards/frontier_entropy_batch_reward": -1.0, "eval_runtime": 16.4084, "eval_samples_per_second": 30.472, "eval_signal/accuracy_reward/centered_abs_mean": 0.4503173828125, "eval_signal/accuracy_reward/group_std_mean": 0.48090726137161255, "eval_signal/accuracy_reward/group_zero_std_frac": 0.0, "eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.22515869140625, "eval_signal/accuracy_reward/weight": 0.5, "eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.22515869140625, "eval_signal/advantage_abs_mean": 0.21093403920531273, "eval_signal/advantage_pre_scale_abs_mean": 0.21093403920531273, "eval_signal/advantage_pre_scale_std": 0.23057331144809723, "eval_signal/advantage_std": 0.23057331144809723, "eval_signal/batch_coverage_0/centered_abs_mean": 0.2845957465469837, "eval_signal/batch_coverage_0/group_std_mean": 0.38306722790002823, "eval_signal/batch_coverage_0/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.003557447053026408, "eval_signal/batch_coverage_0/weight": 0.012500000186264515, "eval_signal/batch_coverage_0/weighted_centered_abs_mean": 0.003557447053026408, "eval_signal/batch_coverage_1/centered_abs_mean": 0.2845957465469837, "eval_signal/batch_coverage_1/group_std_mean": 0.38306722790002823, "eval_signal/batch_coverage_1/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.003557447053026408, "eval_signal/batch_coverage_1/weight": 0.012500000186264515, "eval_signal/batch_coverage_1/weighted_centered_abs_mean": 0.003557447053026408, "eval_signal/batch_coverage_5/centered_abs_mean": 0.2845957465469837, "eval_signal/batch_coverage_5/group_std_mean": 0.38306722790002823, "eval_signal/batch_coverage_5/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.003557447053026408, "eval_signal/batch_coverage_5/weight": 0.012500000186264515, "eval_signal/batch_coverage_5/weighted_centered_abs_mean": 0.003557447053026408, "eval_signal/brier_reward/centered_abs_mean": 0.23709464073181152, "eval_signal/brier_reward/group_std_mean": 0.28304120153188705, "eval_signal/brier_reward/group_zero_std_frac": 0.0, "eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.023709464352577925, "eval_signal/brier_reward/weight": 0.10000000149011612, "eval_signal/brier_reward/weighted_centered_abs_mean": 0.023709464352577925, "eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.041259765625, "eval_signal/confidence_uniqueness_reward/group_std_mean": 0.04886958468705416, "eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.004125976585783064, "eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004125976585783064, "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_aurc_reward/centered_abs_mean": 0.003675252024549991, "eval_signal/frontier_aurc_reward/group_std_mean": 0.005108524113893509, "eval_signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 4.594065103447065e-05, "eval_signal/frontier_aurc_reward/weight": 0.012500000186264515, "eval_signal/frontier_aurc_reward/weighted_centered_abs_mean": 4.594065103447065e-05, "eval_signal/frontier_coverage_10/centered_abs_mean": 0.2993464916944504, "eval_signal/frontier_coverage_10/group_std_mean": 0.4017925262451172, "eval_signal/frontier_coverage_10/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.003741831344086677, "eval_signal/frontier_coverage_10/weight": 0.012500000186264515, "eval_signal/frontier_coverage_10/weighted_centered_abs_mean": 0.003741831344086677, "eval_signal/frontier_coverage_15/centered_abs_mean": 0.2993464916944504, "eval_signal/frontier_coverage_15/group_std_mean": 0.4017925262451172, "eval_signal/frontier_coverage_15/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.003741831344086677, "eval_signal/frontier_coverage_15/weight": 0.012500000186264515, "eval_signal/frontier_coverage_15/weighted_centered_abs_mean": 0.003741831344086677, "eval_signal/frontier_coverage_20/centered_abs_mean": 0.2993464916944504, "eval_signal/frontier_coverage_20/group_std_mean": 0.4017925262451172, "eval_signal/frontier_coverage_20/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.003741831344086677, "eval_signal/frontier_coverage_20/weight": 0.012500000186264515, "eval_signal/frontier_coverage_20/weighted_centered_abs_mean": 0.003741831344086677, "eval_signal/frontier_coverage_25/centered_abs_mean": 0.2993464916944504, "eval_signal/frontier_coverage_25/group_std_mean": 0.4017925262451172, "eval_signal/frontier_coverage_25/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.003741831344086677, "eval_signal/frontier_coverage_25/weight": 0.012500000186264515, "eval_signal/frontier_coverage_25/weighted_centered_abs_mean": 0.003741831344086677, "eval_signal/frontier_ece_reward/centered_abs_mean": 0.07061541080474854, "eval_signal/frontier_ece_reward/group_std_mean": 0.10019382834434509, "eval_signal/frontier_ece_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0070615410804748535, "eval_signal/frontier_ece_reward/weight": 0.10000000149011612, "eval_signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0070615410804748535, "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": 0.244, "step": 50 }, { "calibration/aurc": 0.43317665499190533, "calibration/batch_distribution_entropy": 0.9925566160711995, "calibration/buffer_distribution_entropy": 0.9551033987084541, "calibration/confidence_entropy": 0.5108712319956926, "calibration/coverage@0%": 0.0027366682974559685, "calibration/coverage@1%": 0.0027366682974559685, "calibration/coverage@10%": 0.0027366682974559685, "calibration/coverage@15%": 0.005085004892367906, "calibration/coverage@20%": 0.005085004892367906, "calibration/coverage@25%": 0.012128485812133072, "calibration/coverage@30%": 0.13617294520547946, "calibration/coverage@5%": 0.0027366682974559685, "calibration/ece": 0.18712385392418693, "calibration/mean_confidence": 0.5095072176956078, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 908.6, "completions/max_terminated_length": 908.6, "completions/mean_length": 144.1236328125, "completions/mean_terminated_length": 144.2341522216797, "completions/min_length": 14.2, "completions/min_terminated_length": 62.6, "epoch": 0.176, "grad_norm": 0.001296878675930202, "learning_rate": 1e-06, "loss": -0.0007, "num_tokens": 180756540.0, "reward": 0.8917116761207581, "reward_std": 0.12562428116798402, "rewards/accuracy_reward": 0.452734375, "rewards/batch_coverage_0": 0.21493849158287048, "rewards/batch_coverage_1": 0.21493849158287048, "rewards/batch_coverage_5": 0.24146150350570678, "rewards/brier_reward": 0.7259267091751098, "rewards/confidence_uniqueness_reward": 0.9546158909797668, "rewards/format_reward": 0.99892578125, "rewards/frontier_aurc_reward": -0.00400357972830534, "rewards/frontier_coverage_10": 0.0874374970793724, "rewards/frontier_coverage_15": 0.0874374970793724, "rewards/frontier_coverage_20": 0.0874374970793724, "rewards/frontier_coverage_25": 0.0874374970793724, "rewards/frontier_ece_reward": 0.009037808421999216, "rewards/frontier_entropy_batch_reward": -0.157900133728981, "signal/accuracy_reward/centered_abs_mean": 0.15015869140625, "signal/accuracy_reward/group_std_mean": 0.1922037661075592, "signal/accuracy_reward/group_zero_std_frac": 0.471875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.075079345703125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.075079345703125, "signal/advantage_abs_mean": 0.09891923367977143, "signal/advantage_pre_scale_abs_mean": 0.09891923367977143, "signal/advantage_pre_scale_std": 0.14161589741706848, "signal/advantage_std": 0.14161589741706848, "signal/batch_coverage_0/centered_abs_mean": 0.22720018923282623, "signal/batch_coverage_0/group_std_mean": 0.2847258746623993, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0028400023467838762, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0028400023467838762, "signal/batch_coverage_1/centered_abs_mean": 0.22720018923282623, "signal/batch_coverage_1/group_std_mean": 0.2847258746623993, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0028400023467838762, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0028400023467838762, "signal/batch_coverage_5/centered_abs_mean": 0.2349332571029663, "signal/batch_coverage_5/group_std_mean": 0.2940861344337463, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0029366659000515938, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0029366659000515938, "signal/brier_reward/centered_abs_mean": 0.20331737995147706, "signal/brier_reward/group_std_mean": 0.2511123865842819, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.020331737771630286, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.020331737771630286, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.012506001628935337, "signal/confidence_uniqueness_reward/group_std_mean": 0.018862531706690788, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0012506002094596624, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012506002094596624, "signal/format_reward/centered_abs_mean": 0.002081298828125, "signal/format_reward/group_std_mean": 0.006076698750257492, "signal/format_reward/group_zero_std_frac": 0.965625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0010406494140625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0010406494140625, "signal/frontier_aurc_reward/centered_abs_mean": 0.0023525694850832224, "signal/frontier_aurc_reward/group_std_mean": 0.00342489960603416, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.940711856354028e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.940711856354028e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.2465378850698471, "signal/frontier_coverage_10/group_std_mean": 0.31073635816574097, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0030817235819995403, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0030817235819995403, "signal/frontier_coverage_15/centered_abs_mean": 0.2465378850698471, "signal/frontier_coverage_15/group_std_mean": 0.31073635816574097, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0030817235819995403, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0030817235819995403, "signal/frontier_coverage_20/centered_abs_mean": 0.2465378850698471, "signal/frontier_coverage_20/group_std_mean": 0.31073635816574097, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0030817235819995403, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0030817235819995403, "signal/frontier_coverage_25/centered_abs_mean": 0.2465378850698471, "signal/frontier_coverage_25/group_std_mean": 0.31073635816574097, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0030817235819995403, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0030817235819995403, "signal/frontier_ece_reward/centered_abs_mean": 0.060089949518442154, "signal/frontier_ece_reward/group_std_mean": 0.08282739967107773, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.006008995324373245, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.006008995324373245, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.24448440372943878, "signal/frontier_entropy_batch_reward/group_std_mean": 0.32362828850746156, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.024448440968990327, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.024448440968990327, "step": 55 }, { "calibration/aurc": 0.34776572662075855, "calibration/batch_distribution_entropy": 0.9890232962960198, "calibration/buffer_distribution_entropy": 0.9632647561791915, "calibration/confidence_entropy": 0.49148455616670833, "calibration/coverage@0%": 0.007820144324853229, "calibration/coverage@1%": 0.007820144324853229, "calibration/coverage@10%": 0.04414903375733855, "calibration/coverage@15%": 0.05586778375733855, "calibration/coverage@20%": 0.09378134173189824, "calibration/coverage@25%": 0.20015212206457927, "calibration/coverage@30%": 0.36321321550880625, "calibration/coverage@5%": 0.03242951932485323, "calibration/ece": 0.1301401746722361, "calibration/mean_confidence": 0.4719012042330407, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 627.6, "completions/max_terminated_length": 627.6, "completions/mean_length": 160.3814453125, "completions/mean_terminated_length": 160.5065490722656, "completions/min_length": 0.0, "completions/min_terminated_length": 69.8, "epoch": 0.192, "grad_norm": 0.001202436862513423, "learning_rate": 1e-06, "loss": -0.0004, "num_tokens": 197213662.0, "reward": 0.910670530796051, "reward_std": 0.1196125790476799, "rewards/accuracy_reward": 0.48564453125, "rewards/batch_coverage_0": 0.2758574724197388, "rewards/batch_coverage_1": 0.2758574724197388, "rewards/batch_coverage_5": 0.2934402644634247, "rewards/brier_reward": 0.7440332412719727, "rewards/confidence_uniqueness_reward": 0.9530950307846069, "rewards/format_reward": 0.99873046875, "rewards/frontier_aurc_reward": -0.0034424461890012028, "rewards/frontier_coverage_10": 0.09780684933066368, "rewards/frontier_coverage_15": 0.09780684933066368, "rewards/frontier_coverage_20": 0.09780684933066368, "rewards/frontier_coverage_25": 0.09780684933066368, "rewards/frontier_ece_reward": 0.01746120397001505, "rewards/frontier_entropy_batch_reward": -0.18387694358825685, "signal/accuracy_reward/centered_abs_mean": 0.141888427734375, "signal/accuracy_reward/group_std_mean": 0.18309329748153685, "signal/accuracy_reward/group_zero_std_frac": 0.49375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0709442138671875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0709442138671875, "signal/advantage_abs_mean": 0.09309611171483993, "signal/advantage_pre_scale_abs_mean": 0.09309611171483993, "signal/advantage_pre_scale_std": 0.13633327782154084, "signal/advantage_std": 0.13633327782154084, "signal/batch_coverage_0/centered_abs_mean": 0.23735251724720002, "signal/batch_coverage_0/group_std_mean": 0.2954968333244324, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0029669065494090318, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0029669065494090318, "signal/batch_coverage_1/centered_abs_mean": 0.23735251724720002, "signal/batch_coverage_1/group_std_mean": 0.2954968333244324, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0029669065494090318, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0029669065494090318, "signal/batch_coverage_5/centered_abs_mean": 0.24333244562149048, "signal/batch_coverage_5/group_std_mean": 0.3029507100582123, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.00304165561683476, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.00304165561683476, "signal/brier_reward/centered_abs_mean": 0.19848443567752838, "signal/brier_reward/group_std_mean": 0.2473463922739029, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.019848443567752838, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.019848443567752838, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.014001734368503094, "signal/confidence_uniqueness_reward/group_std_mean": 0.021521516144275665, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0014001735020428897, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014001735020428897, "signal/format_reward/centered_abs_mean": 0.002459716796875, "signal/format_reward/group_std_mean": 0.007181553030386567, "signal/format_reward/group_zero_std_frac": 0.959375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0012298583984375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0012298583984375, "signal/frontier_aurc_reward/centered_abs_mean": 0.002208374394103885, "signal/frontier_aurc_reward/group_std_mean": 0.0032601014245301485, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.7604680508375167e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.7604680508375167e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.25742127895355227, "signal/frontier_coverage_10/group_std_mean": 0.32326231002807615, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0032177661079913376, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0032177661079913376, "signal/frontier_coverage_15/centered_abs_mean": 0.25742127895355227, "signal/frontier_coverage_15/group_std_mean": 0.32326231002807615, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0032177661079913376, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0032177661079913376, "signal/frontier_coverage_20/centered_abs_mean": 0.25742127895355227, "signal/frontier_coverage_20/group_std_mean": 0.32326231002807615, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0032177661079913376, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0032177661079913376, "signal/frontier_coverage_25/centered_abs_mean": 0.25742127895355227, "signal/frontier_coverage_25/group_std_mean": 0.32326231002807615, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0032177661079913376, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0032177661079913376, "signal/frontier_ece_reward/centered_abs_mean": 0.054076675325632095, "signal/frontier_ece_reward/group_std_mean": 0.07428217381238937, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0054076677188277245, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0054076677188277245, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2703059077262878, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34906224608421327, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.027030591294169425, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027030591294169425, "step": 60 }, { "calibration/aurc": 0.2967466269490937, "calibration/batch_distribution_entropy": 0.9885498616950793, "calibration/buffer_distribution_entropy": 0.9688090339090151, "calibration/confidence_entropy": 0.4733263843348757, "calibration/coverage@0%": 0.005859375, "calibration/coverage@1%": 0.005859375, "calibration/coverage@10%": 0.0609375, "calibration/coverage@15%": 0.1984375, "calibration/coverage@20%": 0.373828125, "calibration/coverage@25%": 0.482421875, "calibration/coverage@30%": 0.577734375, "calibration/coverage@5%": 0.0203125, "calibration/ece": 0.1817669776137724, "calibration/mean_confidence": 0.507299410735321, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 649.6, "completions/max_terminated_length": 649.6, "completions/mean_length": 173.56572265625, "completions/mean_terminated_length": 173.6514404296875, "completions/min_length": 30.6, "completions/min_terminated_length": 76.8, "epoch": 0.208, "grad_norm": 0.0011576537508517504, "learning_rate": 1e-06, "loss": -0.0007, "num_tokens": 214023199.0, "reward": 0.9320768237113952, "reward_std": 0.11440627723932266, "rewards/accuracy_reward": 0.52939453125, "rewards/batch_coverage_0": 0.2709454894065857, "rewards/batch_coverage_1": 0.2709454894065857, "rewards/batch_coverage_5": 0.2978154242038727, "rewards/brier_reward": 0.741813600063324, "rewards/confidence_uniqueness_reward": 0.9541807055473328, "rewards/format_reward": 0.99931640625, "rewards/frontier_aurc_reward": -0.003160778433084488, "rewards/frontier_coverage_10": 0.06550543606281281, "rewards/frontier_coverage_15": 0.06550543606281281, "rewards/frontier_coverage_20": 0.06550543606281281, "rewards/frontier_coverage_25": 0.06550543606281281, "rewards/frontier_ece_reward": 0.019096315279603003, "rewards/frontier_entropy_batch_reward": -0.17519809007644654, "signal/accuracy_reward/centered_abs_mean": 0.128802490234375, "signal/accuracy_reward/group_std_mean": 0.1724784791469574, "signal/accuracy_reward/group_zero_std_frac": 0.503125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0644012451171875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0644012451171875, "signal/advantage_abs_mean": 0.08731248527765274, "signal/advantage_pre_scale_abs_mean": 0.08731248527765274, "signal/advantage_pre_scale_std": 0.13033602982759476, "signal/advantage_std": 0.13033602982759476, "signal/batch_coverage_0/centered_abs_mean": 0.2275046318769455, "signal/batch_coverage_0/group_std_mean": 0.2831751525402069, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.00284380791708827, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.00284380791708827, "signal/batch_coverage_1/centered_abs_mean": 0.2275046318769455, "signal/batch_coverage_1/group_std_mean": 0.2831751525402069, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.00284380791708827, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.00284380791708827, "signal/batch_coverage_5/centered_abs_mean": 0.23631368577480316, "signal/batch_coverage_5/group_std_mean": 0.29307244420051576, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002953921118751168, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002953921118751168, "signal/brier_reward/centered_abs_mean": 0.1987267553806305, "signal/brier_reward/group_std_mean": 0.24653929471969604, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.019872675463557244, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.019872675463557244, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.012689129076898098, "signal/confidence_uniqueness_reward/group_std_mean": 0.01802603043615818, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0012689128983765841, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012689128983765841, "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": 0.0006622314453125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0006622314453125, "signal/frontier_aurc_reward/centered_abs_mean": 0.0023700138088315727, "signal/frontier_aurc_reward/group_std_mean": 0.0034461815375834703, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.9625173920067027e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.9625173920067027e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.24167124629020692, "signal/frontier_coverage_10/group_std_mean": 0.3068415105342865, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.003020890709012747, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.003020890709012747, "signal/frontier_coverage_15/centered_abs_mean": 0.24167124629020692, "signal/frontier_coverage_15/group_std_mean": 0.3068415105342865, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.003020890709012747, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.003020890709012747, "signal/frontier_coverage_20/centered_abs_mean": 0.24167124629020692, "signal/frontier_coverage_20/group_std_mean": 0.3068415105342865, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.003020890709012747, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.003020890709012747, "signal/frontier_coverage_25/centered_abs_mean": 0.24167124629020692, "signal/frontier_coverage_25/group_std_mean": 0.3068415105342865, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.003020890709012747, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.003020890709012747, "signal/frontier_ece_reward/centered_abs_mean": 0.052906306087970735, "signal/frontier_ece_reward/group_std_mean": 0.07203125953674316, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.005290630832314491, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.005290630832314491, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2662981808185577, "signal/frontier_entropy_batch_reward/group_std_mean": 0.348252934217453, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026629818603396416, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026629818603396416, "step": 65 }, { "calibration/aurc": 0.33070615070415854, "calibration/batch_distribution_entropy": 0.9869468524486052, "calibration/buffer_distribution_entropy": 0.9730908563547878, "calibration/confidence_entropy": 0.46874678229755284, "calibration/coverage@0%": 0.011354296635266735, "calibration/coverage@1%": 0.011354296635266735, "calibration/coverage@10%": 0.08287933733262698, "calibration/coverage@15%": 0.13485768860464262, "calibration/coverage@20%": 0.250264020204418, "calibration/coverage@25%": 0.4106883719053601, "calibration/coverage@30%": 0.5164756173478348, "calibration/coverage@5%": 0.021119921635266735, "calibration/ece": 0.1684663184505916, "calibration/mean_confidence": 0.4663901185503979, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00146484375, "completions/max_length": 841.6, "completions/max_terminated_length": 841.6, "completions/mean_length": 183.25546875, "completions/mean_terminated_length": 183.52782592773437, "completions/min_length": 15.8, "completions/min_terminated_length": 85.4, "epoch": 0.224, "grad_norm": 0.0009952709078788757, "learning_rate": 1e-06, "loss": -0.0017, "num_tokens": 231052919.0, "reward": 0.9125337243080139, "reward_std": 0.11022349447011948, "rewards/accuracy_reward": 0.47978515625, "rewards/batch_coverage_0": 0.3065375924110413, "rewards/batch_coverage_1": 0.3065375924110413, "rewards/batch_coverage_5": 0.3402839720249176, "rewards/brier_reward": 0.7596318006515503, "rewards/confidence_uniqueness_reward": 0.9531329512596131, "rewards/format_reward": 0.9982421875, "rewards/frontier_aurc_reward": -0.003217305149883032, "rewards/frontier_coverage_10": 0.12076858133077621, "rewards/frontier_coverage_15": 0.12076858133077621, "rewards/frontier_coverage_20": 0.12076858133077621, "rewards/frontier_coverage_25": 0.12076858133077621, "rewards/frontier_ece_reward": 0.020412345044314863, "rewards/frontier_entropy_batch_reward": -0.17712857127189635, "signal/accuracy_reward/centered_abs_mean": 0.122882080078125, "signal/accuracy_reward/group_std_mean": 0.1640130877494812, "signal/accuracy_reward/group_zero_std_frac": 0.525, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0614410400390625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0614410400390625, "signal/advantage_abs_mean": 0.08355161696672439, "signal/advantage_pre_scale_abs_mean": 0.08355161696672439, "signal/advantage_pre_scale_std": 0.1283587023615837, "signal/advantage_std": 0.1283587023615837, "signal/batch_coverage_0/centered_abs_mean": 0.22491827607154846, "signal/batch_coverage_0/group_std_mean": 0.27926393747329714, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0028114784974604843, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0028114784974604843, "signal/batch_coverage_1/centered_abs_mean": 0.22491827607154846, "signal/batch_coverage_1/group_std_mean": 0.27926393747329714, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0028114784974604843, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0028114784974604843, "signal/batch_coverage_5/centered_abs_mean": 0.2368843674659729, "signal/batch_coverage_5/group_std_mean": 0.29381625056266786, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0029610546305775644, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0029610546305775644, "signal/brier_reward/centered_abs_mean": 0.1879715472459793, "signal/brier_reward/group_std_mean": 0.23527895510196686, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.018797155097126962, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.018797155097126962, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.014573495090007781, "signal/confidence_uniqueness_reward/group_std_mean": 0.022509437799453736, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0014573494670912623, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014573494670912623, "signal/format_reward/centered_abs_mean": 0.0033447265625, "signal/format_reward/group_std_mean": 0.008539242390543222, "signal/format_reward/group_zero_std_frac": 0.95625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00167236328125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00167236328125, "signal/frontier_aurc_reward/centered_abs_mean": 0.002444747556000948, "signal/frontier_aurc_reward/group_std_mean": 0.0036018223501741884, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.0559344304492696e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.0559344304492696e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.23408545255661012, "signal/frontier_coverage_10/group_std_mean": 0.2968937695026398, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.002926068194210529, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.002926068194210529, "signal/frontier_coverage_15/centered_abs_mean": 0.23408545255661012, "signal/frontier_coverage_15/group_std_mean": 0.2968937695026398, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002926068194210529, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002926068194210529, "signal/frontier_coverage_20/centered_abs_mean": 0.23408545255661012, "signal/frontier_coverage_20/group_std_mean": 0.2968937695026398, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002926068194210529, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002926068194210529, "signal/frontier_coverage_25/centered_abs_mean": 0.23408545255661012, "signal/frontier_coverage_25/group_std_mean": 0.2968937695026398, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.002926068194210529, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.002926068194210529, "signal/frontier_ece_reward/centered_abs_mean": 0.04961889386177063, "signal/frontier_ece_reward/group_std_mean": 0.06744030714035035, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.004961889423429966, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.004961889423429966, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2602704674005508, "signal/frontier_entropy_batch_reward/group_std_mean": 0.33613797426223757, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026027046889066697, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026027046889066697, "step": 70 }, { "calibration/aurc": 0.36998191958410465, "calibration/batch_distribution_entropy": 0.978596165964422, "calibration/buffer_distribution_entropy": 0.9767574627191198, "calibration/confidence_entropy": 0.48639702887033653, "calibration/coverage@0%": 0.009378060727715744, "calibration/coverage@1%": 0.009378060727715744, "calibration/coverage@10%": 0.0511902243774222, "calibration/coverage@15%": 0.13675162359464335, "calibration/coverage@20%": 0.20005199639787424, "calibration/coverage@25%": 0.24930896802214036, "calibration/coverage@30%": 0.2962741770509573, "calibration/coverage@5%": 0.011721810727715744, "calibration/ece": 0.16677656247922917, "calibration/mean_confidence": 0.49793785648164846, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0017578125, "completions/max_length": 987.2, "completions/max_terminated_length": 987.2, "completions/mean_length": 194.5134765625, "completions/mean_terminated_length": 194.8571563720703, "completions/min_length": 0.0, "completions/min_terminated_length": 86.8, "epoch": 0.24, "grad_norm": 0.0012529160594567657, "learning_rate": 1e-06, "loss": -0.0008, "num_tokens": 248296417.0, "reward": 0.938344395160675, "reward_std": 0.11969478279352189, "rewards/accuracy_reward": 0.5421875, "rewards/batch_coverage_0": 0.2755492031574249, "rewards/batch_coverage_1": 0.2755492031574249, "rewards/batch_coverage_5": 0.30430068969726565, "rewards/brier_reward": 0.7514105319976807, "rewards/confidence_uniqueness_reward": 0.952166223526001, "rewards/format_reward": 0.99794921875, "rewards/frontier_aurc_reward": -0.0029062157031148673, "rewards/frontier_coverage_10": 0.06537975631654262, "rewards/frontier_coverage_15": 0.06537975631654262, "rewards/frontier_coverage_20": 0.06537975631654262, "rewards/frontier_coverage_25": 0.06537975631654262, "rewards/frontier_ece_reward": 0.02016715370118618, "rewards/frontier_entropy_batch_reward": -0.18023498356342316, "signal/accuracy_reward/centered_abs_mean": 0.140185546875, "signal/accuracy_reward/group_std_mean": 0.18424464464187623, "signal/accuracy_reward/group_zero_std_frac": 0.478125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0700927734375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0700927734375, "signal/advantage_abs_mean": 0.09151992201805115, "signal/advantage_pre_scale_abs_mean": 0.09151992201805115, "signal/advantage_pre_scale_std": 0.13715528696775436, "signal/advantage_std": 0.13715528696775436, "signal/batch_coverage_0/centered_abs_mean": 0.21631483733654022, "signal/batch_coverage_0/group_std_mean": 0.27190067768096926, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0027039355598390104, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0027039355598390104, "signal/batch_coverage_1/centered_abs_mean": 0.21631483733654022, "signal/batch_coverage_1/group_std_mean": 0.27190067768096926, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0027039355598390104, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0027039355598390104, "signal/batch_coverage_5/centered_abs_mean": 0.22524747848510743, "signal/batch_coverage_5/group_std_mean": 0.28245970606803894, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002815593546256423, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002815593546256423, "signal/brier_reward/centered_abs_mean": 0.1892361134290695, "signal/brier_reward/group_std_mean": 0.23870718479156494, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.0189236119389534, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.0189236119389534, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.015436468087136746, "signal/confidence_uniqueness_reward/group_std_mean": 0.024331727251410484, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.001543646864593029, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001543646864593029, "signal/format_reward/centered_abs_mean": 0.003924560546875, "signal/format_reward/group_std_mean": 0.01025573662482202, "signal/format_reward/group_zero_std_frac": 0.946875, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0019622802734375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0019622802734375, "signal/frontier_aurc_reward/centered_abs_mean": 0.0024291576817631723, "signal/frontier_aurc_reward/group_std_mean": 0.003613337501883507, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.0364471604116262e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.0364471604116262e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.2288861870765686, "signal/frontier_coverage_10/group_std_mean": 0.2941165566444397, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0028610773850232364, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0028610773850232364, "signal/frontier_coverage_15/centered_abs_mean": 0.2288861870765686, "signal/frontier_coverage_15/group_std_mean": 0.2941165566444397, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0028610773850232364, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0028610773850232364, "signal/frontier_coverage_20/centered_abs_mean": 0.2288861870765686, "signal/frontier_coverage_20/group_std_mean": 0.2941165566444397, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0028610773850232364, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0028610773850232364, "signal/frontier_coverage_25/centered_abs_mean": 0.2288861870765686, "signal/frontier_coverage_25/group_std_mean": 0.2941165566444397, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0028610773850232364, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0028610773850232364, "signal/frontier_ece_reward/centered_abs_mean": 0.046427696198225024, "signal/frontier_ece_reward/group_std_mean": 0.06465726047754287, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.004642769601196051, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.004642769601196051, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2656397998332977, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34345542192459105, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026563981920480727, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026563981920480727, "step": 75 }, { "calibration/aurc": 0.29513670963214134, "calibration/batch_distribution_entropy": 0.982679243117417, "calibration/buffer_distribution_entropy": 0.9791440314685043, "calibration/confidence_entropy": 0.4703822915200556, "calibration/coverage@0%": 0.012962860860286302, "calibration/coverage@1%": 0.012962860860286302, "calibration/coverage@10%": 0.14173070042721542, "calibration/coverage@15%": 0.3022282243882614, "calibration/coverage@20%": 0.38253130320931633, "calibration/coverage@25%": 0.4636432057722237, "calibration/coverage@30%": 0.5654249654741359, "calibration/coverage@5%": 0.04595990810438079, "calibration/ece": 0.15268193158767535, "calibration/mean_confidence": 0.4963923044685613, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00234375, "completions/max_length": 848.6, "completions/max_terminated_length": 848.6, "completions/mean_length": 193.4853515625, "completions/mean_terminated_length": 193.942724609375, "completions/min_length": 0.0, "completions/min_terminated_length": 82.6, "epoch": 0.256, "grad_norm": 0.0010381774045526981, "learning_rate": 1e-06, "loss": -0.0014, "num_tokens": 265332523.0, "reward": 0.9300483703613281, "reward_std": 0.11342969089746475, "rewards/accuracy_reward": 0.5197265625, "rewards/batch_coverage_0": 0.30136329531669614, "rewards/batch_coverage_1": 0.30136329531669614, "rewards/batch_coverage_5": 0.3297120988368988, "rewards/brier_reward": 0.7636389136314392, "rewards/confidence_uniqueness_reward": 0.9517635226249694, "rewards/format_reward": 0.99755859375, "rewards/frontier_aurc_reward": -0.002878309693187475, "rewards/frontier_coverage_10": 0.09460367858409882, "rewards/frontier_coverage_15": 0.09460367858409882, "rewards/frontier_coverage_20": 0.09460367858409882, "rewards/frontier_coverage_25": 0.09460367858409882, "rewards/frontier_ece_reward": 0.020416869968175887, "rewards/frontier_entropy_batch_reward": -0.1852582335472107, "signal/accuracy_reward/centered_abs_mean": 0.1331298828125, "signal/accuracy_reward/group_std_mean": 0.16960339844226838, "signal/accuracy_reward/group_zero_std_frac": 0.53125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.06656494140625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.06656494140625, "signal/advantage_abs_mean": 0.08764777332544327, "signal/advantage_pre_scale_abs_mean": 0.08764777332544327, "signal/advantage_pre_scale_std": 0.13268282115459443, "signal/advantage_std": 0.13268282115459443, "signal/batch_coverage_0/centered_abs_mean": 0.21929797828197478, "signal/batch_coverage_0/group_std_mean": 0.2731813728809357, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0027412247378379106, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0027412247378379106, "signal/batch_coverage_1/centered_abs_mean": 0.21929797828197478, "signal/batch_coverage_1/group_std_mean": 0.2731813728809357, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0027412247378379106, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0027412247378379106, "signal/batch_coverage_5/centered_abs_mean": 0.22841758728027345, "signal/batch_coverage_5/group_std_mean": 0.28432170748710633, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002855219878256321, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002855219878256321, "signal/brier_reward/centered_abs_mean": 0.17741186022758484, "signal/brier_reward/group_std_mean": 0.22425712645053864, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.017741186544299126, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.017741186544299126, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.01610844414681196, "signal/confidence_uniqueness_reward/group_std_mean": 0.025824428349733353, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0016108444426208735, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0016108444426208735, "signal/format_reward/centered_abs_mean": 0.004632568359375, "signal/format_reward/group_std_mean": 0.011674402747303247, "signal/format_reward/group_zero_std_frac": 0.940625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0023162841796875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0023162841796875, "signal/frontier_aurc_reward/centered_abs_mean": 0.002340973261743784, "signal/frontier_aurc_reward/group_std_mean": 0.003478876780718565, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.9262166208354756e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.9262166208354756e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.22374011278152467, "signal/frontier_coverage_10/group_std_mean": 0.2843699276447296, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0027967514004558327, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0027967514004558327, "signal/frontier_coverage_15/centered_abs_mean": 0.22374011278152467, "signal/frontier_coverage_15/group_std_mean": 0.2843699276447296, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0027967514004558327, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0027967514004558327, "signal/frontier_coverage_20/centered_abs_mean": 0.22374011278152467, "signal/frontier_coverage_20/group_std_mean": 0.2843699276447296, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0027967514004558327, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0027967514004558327, "signal/frontier_coverage_25/centered_abs_mean": 0.22374011278152467, "signal/frontier_coverage_25/group_std_mean": 0.2843699276447296, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0027967514004558327, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0027967514004558327, "signal/frontier_ece_reward/centered_abs_mean": 0.041655081510543826, "signal/frontier_ece_reward/group_std_mean": 0.058604808151721956, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.004165508132427931, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.004165508132427931, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.27044638991355896, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34862378239631653, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.027044639736413956, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027044639736413956, "step": 80 }, { "calibration/aurc": 0.3896870872581615, "calibration/batch_distribution_entropy": 0.9905793044214641, "calibration/buffer_distribution_entropy": 0.9813633476097401, "calibration/confidence_entropy": 0.48943567152788453, "calibration/coverage@0%": 0.0062507644324853225, "calibration/coverage@1%": 0.0062507644324853225, "calibration/coverage@10%": 0.0480660217854265, "calibration/coverage@15%": 0.12033699781282377, "calibration/coverage@20%": 0.1715479308161621, "calibration/coverage@25%": 0.21648282515060818, "calibration/coverage@30%": 0.33884955309178466, "calibration/coverage@5%": 0.0062507644324853225, "calibration/ece": 0.14889071128473919, "calibration/mean_confidence": 0.4888013927339571, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 913.2, "completions/max_terminated_length": 913.2, "completions/mean_length": 203.5021484375, "completions/mean_terminated_length": 203.8803955078125, "completions/min_length": 0.0, "completions/min_terminated_length": 91.4, "epoch": 0.272, "grad_norm": 0.0009740217356011271, "learning_rate": 1e-06, "loss": -0.0011, "num_tokens": 282382081.0, "reward": 0.921879255771637, "reward_std": 0.11053736507892609, "rewards/accuracy_reward": 0.5021484375, "rewards/batch_coverage_0": 0.29376897811889646, "rewards/batch_coverage_1": 0.29376897811889646, "rewards/batch_coverage_5": 0.31758773922920225, "rewards/brier_reward": 0.7569353103637695, "rewards/confidence_uniqueness_reward": 0.9531200647354126, "rewards/format_reward": 0.99814453125, "rewards/frontier_aurc_reward": -0.002986749028787017, "rewards/frontier_coverage_10": 0.09633223749697209, "rewards/frontier_coverage_15": 0.09633223749697209, "rewards/frontier_coverage_20": 0.09633223749697209, "rewards/frontier_coverage_25": 0.09633223749697209, "rewards/frontier_ece_reward": 0.01580774337053299, "rewards/frontier_entropy_batch_reward": -0.16946856379508973, "signal/accuracy_reward/centered_abs_mean": 0.12703857421875, "signal/accuracy_reward/group_std_mean": 0.16171487867832185, "signal/accuracy_reward/group_zero_std_frac": 0.55625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.063519287109375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.063519287109375, "signal/advantage_abs_mean": 0.08598263412714005, "signal/advantage_pre_scale_abs_mean": 0.08598263412714005, "signal/advantage_pre_scale_std": 0.12972784489393235, "signal/advantage_std": 0.12972784489393235, "signal/batch_coverage_0/centered_abs_mean": 0.20856670439243316, "signal/batch_coverage_0/group_std_mean": 0.2598303914070129, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002607083786278963, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002607083786278963, "signal/batch_coverage_1/centered_abs_mean": 0.20856670439243316, "signal/batch_coverage_1/group_std_mean": 0.2598303914070129, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002607083786278963, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002607083786278963, "signal/batch_coverage_5/centered_abs_mean": 0.21692942678928376, "signal/batch_coverage_5/group_std_mean": 0.26979697942733766, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0027116178534924984, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0027116178534924984, "signal/brier_reward/centered_abs_mean": 0.1775350570678711, "signal/brier_reward/group_std_mean": 0.22339357137680055, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01775350533425808, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01775350533425808, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013990981131792068, "signal/confidence_uniqueness_reward/group_std_mean": 0.023021527752280235, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.001399098103865981, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001399098103865981, "signal/format_reward/centered_abs_mean": 0.003582763671875, "signal/format_reward/group_std_mean": 0.010159807652235031, "signal/format_reward/group_zero_std_frac": 0.94375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0017913818359375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0017913818359375, "signal/frontier_aurc_reward/centered_abs_mean": 0.0022375767584890127, "signal/frontier_aurc_reward/group_std_mean": 0.003281994676217437, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.796970984491054e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.796970984491054e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.22695637941360475, "signal/frontier_coverage_10/group_std_mean": 0.28653684854507444, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0028369548264890907, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0028369548264890907, "signal/frontier_coverage_15/centered_abs_mean": 0.22695637941360475, "signal/frontier_coverage_15/group_std_mean": 0.28653684854507444, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0028369548264890907, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0028369548264890907, "signal/frontier_coverage_20/centered_abs_mean": 0.22695637941360475, "signal/frontier_coverage_20/group_std_mean": 0.28653684854507444, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0028369548264890907, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0028369548264890907, "signal/frontier_coverage_25/centered_abs_mean": 0.22695637941360475, "signal/frontier_coverage_25/group_std_mean": 0.28653684854507444, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0028369548264890907, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0028369548264890907, "signal/frontier_ece_reward/centered_abs_mean": 0.03842918649315834, "signal/frontier_ece_reward/group_std_mean": 0.054614823311567307, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0038429186679422855, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0038429186679422855, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.25364412665367125, "signal/frontier_entropy_batch_reward/group_std_mean": 0.33080282211303713, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0253644123673439, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0253644123673439, "step": 85 }, { "calibration/aurc": 0.34329576101250975, "calibration/batch_distribution_entropy": 0.9900253191364354, "calibration/buffer_distribution_entropy": 0.9835057909534346, "calibration/confidence_entropy": 0.499811333671287, "calibration/coverage@0%": 0.004689028864970646, "calibration/coverage@1%": 0.004689028864970646, "calibration/coverage@10%": 0.030079653864970645, "calibration/coverage@15%": 0.09414826932485323, "calibration/coverage@20%": 0.12422715875733856, "calibration/coverage@25%": 0.21614481409001957, "calibration/coverage@30%": 0.32567728718199607, "calibration/coverage@5%": 0.017970278864970646, "calibration/ece": 0.12948712843764051, "calibration/mean_confidence": 0.5311546999422978, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 917.4, "completions/max_terminated_length": 917.4, "completions/mean_length": 198.8470703125, "completions/mean_terminated_length": 199.21812744140624, "completions/min_length": 0.0, "completions/min_terminated_length": 89.4, "epoch": 0.288, "grad_norm": 0.0014363753143697977, "learning_rate": 1e-06, "loss": -0.0011, "num_tokens": 299376451.0, "reward": 0.9277427077293396, "reward_std": 0.11109800636768341, "rewards/accuracy_reward": 0.512890625, "rewards/batch_coverage_0": 0.2883468419313431, "rewards/batch_coverage_1": 0.2883468419313431, "rewards/batch_coverage_5": 0.30783610939979555, "rewards/brier_reward": 0.7603950262069702, "rewards/confidence_uniqueness_reward": 0.9534291386604309, "rewards/format_reward": 0.99814453125, "rewards/frontier_aurc_reward": -0.0029745508916676046, "rewards/frontier_coverage_10": 0.09208662584424018, "rewards/frontier_coverage_15": 0.09208662584424018, "rewards/frontier_coverage_20": 0.09208662584424018, "rewards/frontier_coverage_25": 0.09208662584424018, "rewards/frontier_ece_reward": 0.01688596522435546, "rewards/frontier_entropy_batch_reward": -0.16469671428203583, "signal/accuracy_reward/centered_abs_mean": 0.12598876953125, "signal/accuracy_reward/group_std_mean": 0.16362096667289733, "signal/accuracy_reward/group_zero_std_frac": 0.546875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.062994384765625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.062994384765625, "signal/advantage_abs_mean": 0.08507915884256363, "signal/advantage_pre_scale_abs_mean": 0.08507915884256363, "signal/advantage_pre_scale_std": 0.12944555580615996, "signal/advantage_std": 0.12944555580615996, "signal/batch_coverage_0/centered_abs_mean": 0.19805439412593842, "signal/batch_coverage_0/group_std_mean": 0.2480872690677643, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0024756799452006818, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0024756799452006818, "signal/batch_coverage_1/centered_abs_mean": 0.19805439412593842, "signal/batch_coverage_1/group_std_mean": 0.2480872690677643, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0024756799452006818, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0024756799452006818, "signal/batch_coverage_5/centered_abs_mean": 0.20353636145591736, "signal/batch_coverage_5/group_std_mean": 0.25471295416355133, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0025442046578973533, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0025442046578973533, "signal/brier_reward/centered_abs_mean": 0.17084162533283234, "signal/brier_reward/group_std_mean": 0.21616483330726624, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.017084162682294846, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.017084162682294846, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013714919425547123, "signal/confidence_uniqueness_reward/group_std_mean": 0.021899756416678427, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013714919565245508, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013714919565245508, "signal/format_reward/centered_abs_mean": 0.003546142578125, "signal/format_reward/group_std_mean": 0.009427977912127971, "signal/format_reward/group_zero_std_frac": 0.95, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0017730712890625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0017730712890625, "signal/frontier_aurc_reward/centered_abs_mean": 0.002301326720044017, "signal/frontier_aurc_reward/group_std_mean": 0.0034188056364655496, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.876658436434809e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.876658436434809e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.21349818408489227, "signal/frontier_coverage_10/group_std_mean": 0.2712583661079407, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0026687273755669596, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0026687273755669596, "signal/frontier_coverage_15/centered_abs_mean": 0.21349818408489227, "signal/frontier_coverage_15/group_std_mean": 0.2712583661079407, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0026687273755669596, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0026687273755669596, "signal/frontier_coverage_20/centered_abs_mean": 0.21349818408489227, "signal/frontier_coverage_20/group_std_mean": 0.2712583661079407, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0026687273755669596, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0026687273755669596, "signal/frontier_coverage_25/centered_abs_mean": 0.21349818408489227, "signal/frontier_coverage_25/group_std_mean": 0.2712583661079407, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0026687273755669596, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0026687273755669596, "signal/frontier_ece_reward/centered_abs_mean": 0.03777876943349838, "signal/frontier_ece_reward/group_std_mean": 0.0529966726899147, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0037778770085424183, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0037778770085424183, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.25277996361255645, "signal/frontier_entropy_batch_reward/group_std_mean": 0.33164491653442385, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02527799643576145, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02527799643576145, "step": 90 }, { "calibration/aurc": 0.30309922932488725, "calibration/batch_distribution_entropy": 0.9856072076660519, "calibration/buffer_distribution_entropy": 0.984871346889653, "calibration/confidence_entropy": 0.4915576447192608, "calibration/coverage@0%": 0.014462298189823874, "calibration/coverage@1%": 0.014462298189823874, "calibration/coverage@10%": 0.11412671232876712, "calibration/coverage@15%": 0.15985965019569473, "calibration/coverage@20%": 0.2419344728473581, "calibration/coverage@25%": 0.36660270914872795, "calibration/coverage@30%": 0.44947717014888144, "calibration/coverage@5%": 0.050016817514677105, "calibration/ece": 0.114070829368294, "calibration/mean_confidence": 0.5385665254438745, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0013671875, "completions/max_length": 1131.0, "completions/max_terminated_length": 1131.0, "completions/mean_length": 202.87255859375, "completions/mean_terminated_length": 203.15162963867186, "completions/min_length": 0.0, "completions/min_terminated_length": 94.4, "epoch": 0.304, "grad_norm": 0.0014599341666325927, "learning_rate": 1e-06, "loss": -0.0009, "num_tokens": 316383818.0, "reward": 0.9255910992622376, "reward_std": 0.10754784196615219, "rewards/accuracy_reward": 0.5119140625, "rewards/batch_coverage_0": 0.2885267436504364, "rewards/batch_coverage_1": 0.2885267436504364, "rewards/batch_coverage_5": 0.31999329328536985, "rewards/brier_reward": 0.7544575691223144, "rewards/confidence_uniqueness_reward": 0.9530716061592102, "rewards/format_reward": 0.9984375, "rewards/frontier_aurc_reward": -0.0030529949348419906, "rewards/frontier_coverage_10": 0.08898131102323532, "rewards/frontier_coverage_15": 0.08898131102323532, "rewards/frontier_coverage_20": 0.08898131102323532, "rewards/frontier_coverage_25": 0.08898131102323532, "rewards/frontier_ece_reward": 0.014308569300919772, "rewards/frontier_entropy_batch_reward": -0.17392431497573851, "signal/accuracy_reward/centered_abs_mean": 0.11571044921875, "signal/accuracy_reward/group_std_mean": 0.15815874934196472, "signal/accuracy_reward/group_zero_std_frac": 0.525, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.057855224609375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.057855224609375, "signal/advantage_abs_mean": 0.08032247871160507, "signal/advantage_pre_scale_abs_mean": 0.08032247871160507, "signal/advantage_pre_scale_std": 0.12265940755605698, "signal/advantage_std": 0.12265940755605698, "signal/batch_coverage_0/centered_abs_mean": 0.19989262223243714, "signal/batch_coverage_0/group_std_mean": 0.24998952746391295, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002498657861724496, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002498657861724496, "signal/batch_coverage_1/centered_abs_mean": 0.19989262223243714, "signal/batch_coverage_1/group_std_mean": 0.24998952746391295, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002498657861724496, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002498657861724496, "signal/batch_coverage_5/centered_abs_mean": 0.20843929350376128, "signal/batch_coverage_5/group_std_mean": 0.2604684591293335, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0026054912246763705, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0026054912246763705, "signal/brier_reward/centered_abs_mean": 0.17239186763763428, "signal/brier_reward/group_std_mean": 0.218502077460289, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01723918616771698, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01723918616771698, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013918697461485862, "signal/confidence_uniqueness_reward/group_std_mean": 0.022194354236125945, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013918697368353605, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013918697368353605, "signal/format_reward/centered_abs_mean": 0.00302734375, "signal/format_reward/group_std_mean": 0.008838834520429373, "signal/format_reward/group_zero_std_frac": 0.95, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.001513671875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.001513671875, "signal/frontier_aurc_reward/centered_abs_mean": 0.0022680160123854875, "signal/frontier_aurc_reward/group_std_mean": 0.003406479489058256, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.8350201318971813e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.8350201318971813e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.2137775629758835, "signal/frontier_coverage_10/group_std_mean": 0.27304571568965913, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.002672219602391124, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.002672219602391124, "signal/frontier_coverage_15/centered_abs_mean": 0.2137775629758835, "signal/frontier_coverage_15/group_std_mean": 0.27304571568965913, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002672219602391124, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002672219602391124, "signal/frontier_coverage_20/centered_abs_mean": 0.2137775629758835, "signal/frontier_coverage_20/group_std_mean": 0.27304571568965913, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002672219602391124, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002672219602391124, "signal/frontier_coverage_25/centered_abs_mean": 0.2137775629758835, "signal/frontier_coverage_25/group_std_mean": 0.27304571568965913, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.002672219602391124, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.002672219602391124, "signal/frontier_ece_reward/centered_abs_mean": 0.03606079444289208, "signal/frontier_ece_reward/group_std_mean": 0.05103454366326332, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.003606079472228885, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.003606079472228885, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.25910104513168336, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3394786775112152, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.025910104811191558, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025910104811191558, "step": 95 }, { "calibration/aurc": 0.25488904799607315, "calibration/batch_distribution_entropy": 0.9853988122568185, "calibration/buffer_distribution_entropy": 0.9857836181215465, "calibration/confidence_entropy": 0.4843606522713391, "calibration/coverage@0%": 0.008985903864970645, "calibration/coverage@1%": 0.008985903864970645, "calibration/coverage@10%": 0.23848611790606652, "calibration/coverage@15%": 0.35181858488258316, "calibration/coverage@20%": 0.433882705479452, "calibration/coverage@25%": 0.5085471196183953, "calibration/coverage@30%": 0.6290170927103718, "calibration/coverage@5%": 0.07076504403131115, "calibration/ece": 0.1314380640390656, "calibration/mean_confidence": 0.5502947413033883, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0021484375, "completions/max_length": 806.2, "completions/max_terminated_length": 806.2, "completions/mean_length": 201.2990234375, "completions/mean_terminated_length": 201.73126220703125, "completions/min_length": 0.0, "completions/min_terminated_length": 97.2, "epoch": 0.32, "grad_norm": 0.0010468187974765897, "learning_rate": 1e-06, "loss": -0.0015, "num_tokens": 333533824.0, "reward": 0.9378777503967285, "reward_std": 0.09861989617347718, "rewards/accuracy_reward": 0.53017578125, "rewards/batch_coverage_0": 0.32871403694152834, "rewards/batch_coverage_1": 0.32871403694152834, "rewards/batch_coverage_5": 0.3625796020030975, "rewards/brier_reward": 0.7723654985427857, "rewards/confidence_uniqueness_reward": 0.9525111436843872, "rewards/format_reward": 0.9978515625, "rewards/frontier_aurc_reward": -0.0027396206744015215, "rewards/frontier_coverage_10": 0.09240553937852383, "rewards/frontier_coverage_15": 0.09240553937852383, "rewards/frontier_coverage_20": 0.09240553937852383, "rewards/frontier_coverage_25": 0.09240553937852383, "rewards/frontier_ece_reward": 0.01853880248963833, "rewards/frontier_entropy_batch_reward": -0.17813571989536287, "signal/accuracy_reward/centered_abs_mean": 0.094427490234375, "signal/accuracy_reward/group_std_mean": 0.13309391140937804, "signal/accuracy_reward/group_zero_std_frac": 0.584375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0472137451171875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0472137451171875, "signal/advantage_abs_mean": 0.07220838218927383, "signal/advantage_pre_scale_abs_mean": 0.07220838218927383, "signal/advantage_pre_scale_std": 0.1159633681178093, "signal/advantage_std": 0.1159633681178093, "signal/batch_coverage_0/centered_abs_mean": 0.1942468464374542, "signal/batch_coverage_0/group_std_mean": 0.24366948008537292, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0024280855897814034, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0024280855897814034, "signal/batch_coverage_1/centered_abs_mean": 0.1942468464374542, "signal/batch_coverage_1/group_std_mean": 0.24366948008537292, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0024280855897814034, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0024280855897814034, "signal/batch_coverage_5/centered_abs_mean": 0.20423491299152374, "signal/batch_coverage_5/group_std_mean": 0.25555876195430755, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0025529364589601754, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0025529364589601754, "signal/brier_reward/centered_abs_mean": 0.15765537917613984, "signal/brier_reward/group_std_mean": 0.20305769443511962, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01576553825289011, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01576553825289011, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.014990145899355411, "signal/confidence_uniqueness_reward/group_std_mean": 0.024523165822029114, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.001499014627188444, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001499014627188444, "signal/format_reward/centered_abs_mean": 0.0041259765625, "signal/format_reward/group_std_mean": 0.011144472099840642, "signal/format_reward/group_zero_std_frac": 0.940625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00206298828125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00206298828125, "signal/frontier_aurc_reward/centered_abs_mean": 0.0021417980547994374, "signal/frontier_aurc_reward/group_std_mean": 0.003238040255382657, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.6772475757752544e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.6772475757752544e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.19105646312236785, "signal/frontier_coverage_10/group_std_mean": 0.2469618320465088, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.002388205798342824, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.002388205798342824, "signal/frontier_coverage_15/centered_abs_mean": 0.19105646312236785, "signal/frontier_coverage_15/group_std_mean": 0.2469618320465088, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002388205798342824, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002388205798342824, "signal/frontier_coverage_20/centered_abs_mean": 0.19105646312236785, "signal/frontier_coverage_20/group_std_mean": 0.2469618320465088, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002388205798342824, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002388205798342824, "signal/frontier_coverage_25/centered_abs_mean": 0.19105646312236785, "signal/frontier_coverage_25/group_std_mean": 0.2469618320465088, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.002388205798342824, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.002388205798342824, "signal/frontier_ece_reward/centered_abs_mean": 0.03633599579334259, "signal/frontier_ece_reward/group_std_mean": 0.05160667449235916, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0036335998214781285, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0036335998214781285, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2626646161079407, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34101126790046693, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026266463100910187, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026266463100910187, "step": 100 }, { "epoch": 0.32, "eval_calibration/aurc": 0.5154191519482911, "eval_calibration/batch_distribution_entropy": 0.9199682016541648, "eval_calibration/buffer_distribution_entropy": 0.9862273887528408, "eval_calibration/confidence_entropy": 0.4870277605332142, "eval_calibration/coverage@0%": 0.046875, "eval_calibration/coverage@1%": 0.046875, "eval_calibration/coverage@10%": 0.0859375, "eval_calibration/coverage@15%": 0.0859375, "eval_calibration/coverage@20%": 0.125, "eval_calibration/coverage@25%": 0.1328125, "eval_calibration/coverage@30%": 0.140625, "eval_calibration/coverage@5%": 0.046875, "eval_calibration/ece": 0.26012049031776796, "eval_calibration/mean_confidence": 0.45733527972941607, "eval_completions/clipped_ratio": 0.0, "eval_completions/max_length": 397.0, "eval_completions/max_terminated_length": 397.0, "eval_completions/mean_length": 202.18406677246094, "eval_completions/mean_terminated_length": 202.18406677246094, "eval_completions/min_length": 117.0, "eval_completions/min_terminated_length": 117.0, "eval_loss": 0.0, "eval_num_tokens": 333533824.0, "eval_reward": 0.7840017229318619, "eval_reward_std": 0.2287086918950081, "eval_rewards/accuracy_reward": 0.404296875, "eval_rewards/batch_coverage_0": 0.14562061801552773, "eval_rewards/batch_coverage_1": 0.14562061801552773, "eval_rewards/batch_coverage_5": 0.14562061801552773, "eval_rewards/brier_reward": 0.7628249675035477, "eval_rewards/confidence_uniqueness_reward": 0.899658203125, "eval_rewards/format_reward": 1.0, "eval_rewards/frontier_aurc_reward": -0.0035637450055219233, "eval_rewards/frontier_coverage_10": 0.17291036993265152, "eval_rewards/frontier_coverage_15": 0.17291036993265152, "eval_rewards/frontier_coverage_20": 0.17291036993265152, "eval_rewards/frontier_coverage_25": 0.17291036993265152, "eval_rewards/frontier_ece_reward": 0.01543213170953095, "eval_rewards/frontier_entropy_batch_reward": -1.0, "eval_runtime": 20.0137, "eval_samples_per_second": 24.983, "eval_signal/accuracy_reward/centered_abs_mean": 0.4666748046875, "eval_signal/accuracy_reward/group_std_mean": 0.49021974951028824, "eval_signal/accuracy_reward/group_zero_std_frac": 0.0, "eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.23333740234375, "eval_signal/accuracy_reward/weight": 0.5, "eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.23333740234375, "eval_signal/advantage_abs_mean": 0.21138876676559448, "eval_signal/advantage_pre_scale_abs_mean": 0.21138876676559448, "eval_signal/advantage_pre_scale_std": 0.22633719071745872, "eval_signal/advantage_std": 0.22633719071745872, "eval_signal/batch_coverage_0/centered_abs_mean": 0.2982206717133522, "eval_signal/batch_coverage_0/group_std_mean": 0.3763407841324806, "eval_signal/batch_coverage_0/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0037277585361152887, "eval_signal/batch_coverage_0/weight": 0.012500000186264515, "eval_signal/batch_coverage_0/weighted_centered_abs_mean": 0.0037277585361152887, "eval_signal/batch_coverage_1/centered_abs_mean": 0.2982206717133522, "eval_signal/batch_coverage_1/group_std_mean": 0.3763407841324806, "eval_signal/batch_coverage_1/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0037277585361152887, "eval_signal/batch_coverage_1/weight": 0.012500000186264515, "eval_signal/batch_coverage_1/weighted_centered_abs_mean": 0.0037277585361152887, "eval_signal/batch_coverage_5/centered_abs_mean": 0.2982206717133522, "eval_signal/batch_coverage_5/group_std_mean": 0.3763407841324806, "eval_signal/batch_coverage_5/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0037277585361152887, "eval_signal/batch_coverage_5/weight": 0.012500000186264515, "eval_signal/batch_coverage_5/weighted_centered_abs_mean": 0.0037277585361152887, "eval_signal/brier_reward/centered_abs_mean": 0.2208237424492836, "eval_signal/brier_reward/group_std_mean": 0.2668558470904827, "eval_signal/brier_reward/group_zero_std_frac": 0.0, "eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.022082374896854162, "eval_signal/brier_reward/weight": 0.10000000149011612, "eval_signal/brier_reward/weighted_centered_abs_mean": 0.022082374896854162, "eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.038482666015625, "eval_signal/confidence_uniqueness_reward/group_std_mean": 0.04386392701417208, "eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.003848266671411693, "eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.003848266671411693, "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_aurc_reward/centered_abs_mean": 0.0035641847061924636, "eval_signal/frontier_aurc_reward/group_std_mean": 0.005475352983921766, "eval_signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 4.455230737221427e-05, "eval_signal/frontier_aurc_reward/weight": 0.012500000186264515, "eval_signal/frontier_aurc_reward/weighted_centered_abs_mean": 4.455230737221427e-05, "eval_signal/frontier_coverage_10/centered_abs_mean": 0.352456696331501, "eval_signal/frontier_coverage_10/group_std_mean": 0.43984775245189667, "eval_signal/frontier_coverage_10/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.004405708867125213, "eval_signal/frontier_coverage_10/weight": 0.012500000186264515, "eval_signal/frontier_coverage_10/weighted_centered_abs_mean": 0.004405708867125213, "eval_signal/frontier_coverage_15/centered_abs_mean": 0.352456696331501, "eval_signal/frontier_coverage_15/group_std_mean": 0.43984775245189667, "eval_signal/frontier_coverage_15/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.004405708867125213, "eval_signal/frontier_coverage_15/weight": 0.012500000186264515, "eval_signal/frontier_coverage_15/weighted_centered_abs_mean": 0.004405708867125213, "eval_signal/frontier_coverage_20/centered_abs_mean": 0.352456696331501, "eval_signal/frontier_coverage_20/group_std_mean": 0.43984775245189667, "eval_signal/frontier_coverage_20/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.004405708867125213, "eval_signal/frontier_coverage_20/weight": 0.012500000186264515, "eval_signal/frontier_coverage_20/weighted_centered_abs_mean": 0.004405708867125213, "eval_signal/frontier_coverage_25/centered_abs_mean": 0.352456696331501, "eval_signal/frontier_coverage_25/group_std_mean": 0.43984775245189667, "eval_signal/frontier_coverage_25/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.004405708867125213, "eval_signal/frontier_coverage_25/weight": 0.012500000186264515, "eval_signal/frontier_coverage_25/weighted_centered_abs_mean": 0.004405708867125213, "eval_signal/frontier_ece_reward/centered_abs_mean": 0.0408102422952652, "eval_signal/frontier_ece_reward/group_std_mean": 0.062478141859173775, "eval_signal/frontier_ece_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0040810241480357945, "eval_signal/frontier_ece_reward/weight": 0.10000000149011612, "eval_signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0040810241480357945, "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": 0.2, "step": 100 }, { "calibration/aurc": 0.3196450382672963, "calibration/batch_distribution_entropy": 0.9867421912284042, "calibration/buffer_distribution_entropy": 0.9881853265880236, "calibration/confidence_entropy": 0.4995684288989851, "calibration/coverage@0%": 0.0054902413283024355, "calibration/coverage@1%": 0.0054902413283024355, "calibration/coverage@10%": 0.0054902413283024355, "calibration/coverage@15%": 0.024329209986631913, "calibration/coverage@20%": 0.1426748102838374, "calibration/coverage@25%": 0.41238806871982964, "calibration/coverage@30%": 0.5439935222484529, "calibration/coverage@5%": 0.0054902413283024355, "calibration/ece": 0.16247002163844593, "calibration/mean_confidence": 0.48427799440014196, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 745.2, "completions/max_terminated_length": 745.2, "completions/mean_length": 203.75234375, "completions/mean_terminated_length": 204.2887756347656, "completions/min_length": 0.0, "completions/min_terminated_length": 92.0, "epoch": 0.336, "grad_norm": 0.0011075808433815837, "learning_rate": 1e-06, "loss": -0.0021, "num_tokens": 350342680.0, "reward": 0.9391863346099854, "reward_std": 0.10431983917951584, "rewards/accuracy_reward": 0.53974609375, "rewards/batch_coverage_0": 0.3115852832794189, "rewards/batch_coverage_1": 0.3115852832794189, "rewards/batch_coverage_5": 0.3388616740703583, "rewards/brier_reward": 0.7595332026481628, "rewards/confidence_uniqueness_reward": 0.9517552733421326, "rewards/format_reward": 0.99736328125, "rewards/frontier_aurc_reward": -0.0026606473140418528, "rewards/frontier_coverage_10": 0.07436705008149147, "rewards/frontier_coverage_15": 0.07436705008149147, "rewards/frontier_coverage_20": 0.07436705008149147, "rewards/frontier_coverage_25": 0.07436705008149147, "rewards/frontier_ece_reward": 0.015066038630902768, "rewards/frontier_entropy_batch_reward": -0.17714300751686096, "signal/accuracy_reward/centered_abs_mean": 0.109368896484375, "signal/accuracy_reward/group_std_mean": 0.14659990668296813, "signal/accuracy_reward/group_zero_std_frac": 0.56875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0546844482421875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0546844482421875, "signal/advantage_abs_mean": 0.0785999283194542, "signal/advantage_pre_scale_abs_mean": 0.0785999283194542, "signal/advantage_pre_scale_std": 0.12327169775962829, "signal/advantage_std": 0.12327169775962829, "signal/batch_coverage_0/centered_abs_mean": 0.19352388978004456, "signal/batch_coverage_0/group_std_mean": 0.2433212637901306, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0024190485943108796, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0024190485943108796, "signal/batch_coverage_1/centered_abs_mean": 0.19352388978004456, "signal/batch_coverage_1/group_std_mean": 0.2433212637901306, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0024190485943108796, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0024190485943108796, "signal/batch_coverage_5/centered_abs_mean": 0.20151999592781067, "signal/batch_coverage_5/group_std_mean": 0.25335530936717987, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0025189999490976334, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0025189999490976334, "signal/brier_reward/centered_abs_mean": 0.1666221112012863, "signal/brier_reward/group_std_mean": 0.2118913471698761, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01666221134364605, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01666221134364605, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.015501018427312375, "signal/confidence_uniqueness_reward/group_std_mean": 0.026365644484758376, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0015501018846407534, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0015501018846407534, "signal/format_reward/centered_abs_mean": 0.005059814453125, "signal/format_reward/group_std_mean": 0.013570299744606018, "signal/format_reward/group_zero_std_frac": 0.928125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0025299072265625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0025299072265625, "signal/frontier_aurc_reward/centered_abs_mean": 0.0021155539900064467, "signal/frontier_aurc_reward/group_std_mean": 0.003189340280368924, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.644442611199338e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.644442611199338e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.20592671036720275, "signal/frontier_coverage_10/group_std_mean": 0.26229742765426634, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.002574084047228098, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.002574084047228098, "signal/frontier_coverage_15/centered_abs_mean": 0.20592671036720275, "signal/frontier_coverage_15/group_std_mean": 0.26229742765426634, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002574084047228098, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002574084047228098, "signal/frontier_coverage_20/centered_abs_mean": 0.20592671036720275, "signal/frontier_coverage_20/group_std_mean": 0.26229742765426634, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002574084047228098, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002574084047228098, "signal/frontier_coverage_25/centered_abs_mean": 0.20592671036720275, "signal/frontier_coverage_25/group_std_mean": 0.26229742765426634, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.002574084047228098, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.002574084047228098, "signal/frontier_ece_reward/centered_abs_mean": 0.03176134303212166, "signal/frontier_ece_reward/group_std_mean": 0.04571407660841942, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.00317613435909152, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.00317613435909152, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.25965495109558107, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3369144558906555, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02596549578011036, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02596549578011036, "step": 105 }, { "calibration/aurc": 0.3456084353470958, "calibration/batch_distribution_entropy": 0.9675796049757068, "calibration/buffer_distribution_entropy": 0.9935659326649713, "calibration/confidence_entropy": 0.47231927794527173, "calibration/coverage@0%": 0.017634126813251876, "calibration/coverage@1%": 0.017634126813251876, "calibration/coverage@10%": 0.10894962711313769, "calibration/coverage@15%": 0.22369224825210784, "calibration/coverage@20%": 0.31617563018071965, "calibration/coverage@25%": 0.4019131642542839, "calibration/coverage@30%": 0.47789819483398627, "calibration/coverage@5%": 0.04511601532012811, "calibration/ece": 0.1263553567669657, "calibration/mean_confidence": 0.4420099973149906, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0021484375, "completions/max_length": 782.6, "completions/max_terminated_length": 782.6, "completions/mean_length": 204.94638671875, "completions/mean_terminated_length": 205.38886413574218, "completions/min_length": 0.0, "completions/min_terminated_length": 99.6, "epoch": 0.352, "grad_norm": 0.0009076938149519265, "learning_rate": 1e-06, "loss": -0.0015, "num_tokens": 367701747.0, "reward": 0.9082619547843933, "reward_std": 0.102626533806324, "rewards/accuracy_reward": 0.47099609375, "rewards/batch_coverage_0": 0.32009387016296387, "rewards/batch_coverage_1": 0.32009387016296387, "rewards/batch_coverage_5": 0.34040191769599915, "rewards/brier_reward": 0.769482684135437, "rewards/confidence_uniqueness_reward": 0.9515717029571533, "rewards/format_reward": 0.99775390625, "rewards/frontier_aurc_reward": -0.00305192144587636, "rewards/frontier_coverage_10": 0.1323698341846466, "rewards/frontier_coverage_15": 0.1323698341846466, "rewards/frontier_coverage_20": 0.1323698341846466, "rewards/frontier_coverage_25": 0.1323698341846466, "rewards/frontier_ece_reward": 0.012802435457706452, "rewards/frontier_entropy_batch_reward": -0.18336443305015565, "signal/accuracy_reward/centered_abs_mean": 0.112298583984375, "signal/accuracy_reward/group_std_mean": 0.14630261957645416, "signal/accuracy_reward/group_zero_std_frac": 0.5875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0561492919921875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0561492919921875, "signal/advantage_abs_mean": 0.07844830304384232, "signal/advantage_pre_scale_abs_mean": 0.07844830304384232, "signal/advantage_pre_scale_std": 0.12251064479351044, "signal/advantage_std": 0.12251064479351044, "signal/batch_coverage_0/centered_abs_mean": 0.1968519389629364, "signal/batch_coverage_0/group_std_mean": 0.24579598307609557, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002460649237036705, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002460649237036705, "signal/batch_coverage_1/centered_abs_mean": 0.1968519389629364, "signal/batch_coverage_1/group_std_mean": 0.24579598307609557, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002460649237036705, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002460649237036705, "signal/batch_coverage_5/centered_abs_mean": 0.2032012462615967, "signal/batch_coverage_5/group_std_mean": 0.25404067039489747, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0025400156155228615, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0025400156155228615, "signal/brier_reward/centered_abs_mean": 0.159521421790123, "signal/brier_reward/group_std_mean": 0.20316518545150758, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01595214232802391, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01595214232802391, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.015370776690542699, "signal/confidence_uniqueness_reward/group_std_mean": 0.02492622844874859, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0015370776876807214, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0015370776876807214, "signal/format_reward/centered_abs_mean": 0.004290771484375, "signal/format_reward/group_std_mean": 0.01130137830041349, "signal/format_reward/group_zero_std_frac": 0.940625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0021453857421875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0021453857421875, "signal/frontier_aurc_reward/centered_abs_mean": 0.002293165912851691, "signal/frontier_aurc_reward/group_std_mean": 0.0034823329653590916, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.8664574347203598e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.8664574347203598e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.20986687242984772, "signal/frontier_coverage_10/group_std_mean": 0.26480966806411743, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0026233359705656765, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0026233359705656765, "signal/frontier_coverage_15/centered_abs_mean": 0.20986687242984772, "signal/frontier_coverage_15/group_std_mean": 0.26480966806411743, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0026233359705656765, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0026233359705656765, "signal/frontier_coverage_20/centered_abs_mean": 0.20986687242984772, "signal/frontier_coverage_20/group_std_mean": 0.26480966806411743, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0026233359705656765, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0026233359705656765, "signal/frontier_coverage_25/centered_abs_mean": 0.20986687242984772, "signal/frontier_coverage_25/group_std_mean": 0.26480966806411743, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0026233359705656765, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0026233359705656765, "signal/frontier_ece_reward/centered_abs_mean": 0.02856634072959423, "signal/frontier_ece_reward/group_std_mean": 0.04047100394964218, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.002856634184718132, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.002856634184718132, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2592151463031769, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3351094901561737, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.025921515375375747, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025921515375375747, "step": 110 }, { "calibration/aurc": 0.37720919222191196, "calibration/batch_distribution_entropy": 0.9823951198253725, "calibration/buffer_distribution_entropy": 0.997470796334059, "calibration/confidence_entropy": 0.474846575946022, "calibration/coverage@0%": 0.011334240459882582, "calibration/coverage@1%": 0.011334240459882582, "calibration/coverage@10%": 0.013287365459882582, "calibration/coverage@15%": 0.05704195205479452, "calibration/coverage@20%": 0.23087695694716243, "calibration/coverage@25%": 0.3285530821917808, "calibration/coverage@30%": 0.3817147749510763, "calibration/coverage@5%": 0.011334240459882582, "calibration/ece": 0.14286153959476197, "calibration/mean_confidence": 0.492683542229572, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 1186.6, "completions/max_terminated_length": 1186.6, "completions/mean_length": 208.47265625, "completions/mean_terminated_length": 209.0223388671875, "completions/min_length": 0.0, "completions/min_terminated_length": 95.4, "epoch": 0.368, "grad_norm": 0.00111673166975379, "learning_rate": 1e-06, "loss": -0.0026, "num_tokens": 384901979.0, "reward": 0.9184874653816223, "reward_std": 0.09868537932634354, "rewards/accuracy_reward": 0.49404296875, "rewards/batch_coverage_0": 0.32936405539512636, "rewards/batch_coverage_1": 0.32936405539512636, "rewards/batch_coverage_5": 0.34511337280273435, "rewards/brier_reward": 0.7676645040512085, "rewards/confidence_uniqueness_reward": 0.9514286518096924, "rewards/format_reward": 0.99736328125, "rewards/frontier_aurc_reward": -0.003175344597548246, "rewards/frontier_coverage_10": 0.11608664691448212, "rewards/frontier_coverage_15": 0.11608664691448212, "rewards/frontier_coverage_20": 0.11608664691448212, "rewards/frontier_coverage_25": 0.11608664691448212, "rewards/frontier_ece_reward": 0.01332780383527279, "rewards/frontier_entropy_batch_reward": -0.18770430684089662, "signal/accuracy_reward/centered_abs_mean": 0.102899169921875, "signal/accuracy_reward/group_std_mean": 0.13341209292411804, "signal/accuracy_reward/group_zero_std_frac": 0.628125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0514495849609375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0514495849609375, "signal/advantage_abs_mean": 0.07502751648426056, "signal/advantage_pre_scale_abs_mean": 0.07502751648426056, "signal/advantage_pre_scale_std": 0.11870786696672439, "signal/advantage_std": 0.11870786696672439, "signal/batch_coverage_0/centered_abs_mean": 0.19275737106800078, "signal/batch_coverage_0/group_std_mean": 0.24188868403434755, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0024094672873616217, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0024094672873616217, "signal/batch_coverage_1/centered_abs_mean": 0.19275737106800078, "signal/batch_coverage_1/group_std_mean": 0.24188868403434755, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0024094672873616217, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0024094672873616217, "signal/batch_coverage_5/centered_abs_mean": 0.19718607068061828, "signal/batch_coverage_5/group_std_mean": 0.24712843894958497, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002464825985953212, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002464825985953212, "signal/brier_reward/centered_abs_mean": 0.15777615308761597, "signal/brier_reward/group_std_mean": 0.20168555974960328, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.015777615644037724, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.015777615644037724, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.015588437393307685, "signal/confidence_uniqueness_reward/group_std_mean": 0.026337483897805214, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0015588437439873815, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0015588437439873815, "signal/format_reward/centered_abs_mean": 0.004986572265625, "signal/format_reward/group_std_mean": 0.013306216523051261, "signal/format_reward/group_zero_std_frac": 0.928125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0024932861328125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0024932861328125, "signal/frontier_aurc_reward/centered_abs_mean": 0.002676937123760581, "signal/frontier_aurc_reward/group_std_mean": 0.00409984439611435, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.346171288285404e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.346171288285404e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.203335765004158, "signal/frontier_coverage_10/group_std_mean": 0.25854618549346925, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0025416971184313297, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0025416971184313297, "signal/frontier_coverage_15/centered_abs_mean": 0.203335765004158, "signal/frontier_coverage_15/group_std_mean": 0.25854618549346925, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0025416971184313297, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0025416971184313297, "signal/frontier_coverage_20/centered_abs_mean": 0.203335765004158, "signal/frontier_coverage_20/group_std_mean": 0.25854618549346925, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0025416971184313297, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0025416971184313297, "signal/frontier_coverage_25/centered_abs_mean": 0.203335765004158, "signal/frontier_coverage_25/group_std_mean": 0.25854618549346925, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0025416971184313297, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0025416971184313297, "signal/frontier_ece_reward/centered_abs_mean": 0.02644171565771103, "signal/frontier_ece_reward/group_std_mean": 0.036624275147914886, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0026441716589033605, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0026441716589033605, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2689514756202698, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34790114164352415, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02689514830708504, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02689514830708504, "step": 115 }, { "calibration/aurc": 0.3325915670720264, "calibration/batch_distribution_entropy": 0.9719514587979404, "calibration/buffer_distribution_entropy": 0.9990019013780438, "calibration/confidence_entropy": 0.4560380915126088, "calibration/coverage@0%": 0.00863132449578901, "calibration/coverage@1%": 0.00863132449578901, "calibration/coverage@10%": 0.14859227300038932, "calibration/coverage@15%": 0.25219043535923175, "calibration/coverage@20%": 0.2922035902295982, "calibration/coverage@25%": 0.34434828264649014, "calibration/coverage@30%": 0.40516351974060116, "calibration/coverage@5%": 0.08349798348815307, "calibration/ece": 0.12469228202365668, "calibration/mean_confidence": 0.4547253329616255, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 1144.4, "completions/max_terminated_length": 1144.4, "completions/mean_length": 212.9341796875, "completions/mean_terminated_length": 213.5219268798828, "completions/min_length": 0.0, "completions/min_terminated_length": 102.4, "epoch": 0.384, "grad_norm": 0.00121228804346174, "learning_rate": 1e-06, "loss": -0.0019, "num_tokens": 401938937.0, "reward": 0.9359744548797607, "reward_std": 0.10185678750276565, "rewards/accuracy_reward": 0.5279296875, "rewards/batch_coverage_0": 0.3254628241062164, "rewards/batch_coverage_1": 0.3254628241062164, "rewards/batch_coverage_5": 0.3614440143108368, "rewards/brier_reward": 0.7808833003044129, "rewards/confidence_uniqueness_reward": 0.9508043766021729, "rewards/format_reward": 0.99716796875, "rewards/frontier_aurc_reward": -0.0028614962473511698, "rewards/frontier_coverage_10": 0.10814944505691529, "rewards/frontier_coverage_15": 0.10814944505691529, "rewards/frontier_coverage_20": 0.10814944505691529, "rewards/frontier_coverage_25": 0.10814944505691529, "rewards/frontier_ece_reward": 0.014707298018038273, "rewards/frontier_entropy_batch_reward": -0.19240202009677887, "signal/accuracy_reward/centered_abs_mean": 0.10535888671875, "signal/accuracy_reward/group_std_mean": 0.1491241306066513, "signal/accuracy_reward/group_zero_std_frac": 0.53125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.052679443359375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.052679443359375, "signal/advantage_abs_mean": 0.07327564358711243, "signal/advantage_pre_scale_abs_mean": 0.07327564358711243, "signal/advantage_pre_scale_std": 0.11974472105503083, "signal/advantage_std": 0.11974472105503083, "signal/batch_coverage_0/centered_abs_mean": 0.18728042542934417, "signal/batch_coverage_0/group_std_mean": 0.2371540904045105, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002341005392372608, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002341005392372608, "signal/batch_coverage_1/centered_abs_mean": 0.18728042542934417, "signal/batch_coverage_1/group_std_mean": 0.2371540904045105, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002341005392372608, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002341005392372608, "signal/batch_coverage_5/centered_abs_mean": 0.19866327345371246, "signal/batch_coverage_5/group_std_mean": 0.2515545904636383, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002483290946111083, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002483290946111083, "signal/brier_reward/centered_abs_mean": 0.1508100986480713, "signal/brier_reward/group_std_mean": 0.19526297152042388, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.015081010200083256, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.015081010200083256, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.01601494774222374, "signal/confidence_uniqueness_reward/group_std_mean": 0.027539638802409172, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.00160149484872818, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.00160149484872818, "signal/format_reward/centered_abs_mean": 0.005401611328125, "signal/format_reward/group_std_mean": 0.014524403773248196, "signal/format_reward/group_zero_std_frac": 0.921875, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0027008056640625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0027008056640625, "signal/frontier_aurc_reward/centered_abs_mean": 0.0028855432756245136, "signal/frontier_aurc_reward/group_std_mean": 0.004477264592424035, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.606929167290218e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.606929167290218e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1959839254617691, "signal/frontier_coverage_10/group_std_mean": 0.25251251459121704, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0024497990496456625, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0024497990496456625, "signal/frontier_coverage_15/centered_abs_mean": 0.1959839254617691, "signal/frontier_coverage_15/group_std_mean": 0.25251251459121704, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0024497990496456625, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0024497990496456625, "signal/frontier_coverage_20/centered_abs_mean": 0.1959839254617691, "signal/frontier_coverage_20/group_std_mean": 0.25251251459121704, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0024497990496456625, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0024497990496456625, "signal/frontier_coverage_25/centered_abs_mean": 0.1959839254617691, "signal/frontier_coverage_25/group_std_mean": 0.25251251459121704, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0024497990496456625, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0024497990496456625, "signal/frontier_ece_reward/centered_abs_mean": 0.02398340478539467, "signal/frontier_ece_reward/group_std_mean": 0.032227625325322154, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0023983404971659184, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0023983404971659184, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.25935113728046416, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3366460561752319, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.025935114547610282, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025935114547610282, "step": 120 }, { "calibration/aurc": 0.4278979652849483, "calibration/batch_distribution_entropy": 0.9818957947468752, "calibration/buffer_distribution_entropy": 0.9988738769794742, "calibration/confidence_entropy": 0.49665681491171726, "calibration/coverage@0%": 0.0023498714554314876, "calibration/coverage@1%": 0.0023498714554314876, "calibration/coverage@10%": 0.0023498714554314876, "calibration/coverage@15%": 0.0023498714554314876, "calibration/coverage@20%": 0.016439891024903112, "calibration/coverage@25%": 0.04835241124090067, "calibration/coverage@30%": 0.16939441805740696, "calibration/coverage@5%": 0.0023498714554314876, "calibration/ece": 0.17177305286252811, "calibration/mean_confidence": 0.5198353890958622, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0021484375, "completions/max_length": 982.4, "completions/max_terminated_length": 982.4, "completions/mean_length": 217.64072265625, "completions/mean_terminated_length": 218.10805969238282, "completions/min_length": 0.0, "completions/min_terminated_length": 101.6, "epoch": 0.4, "grad_norm": 0.000913388910703361, "learning_rate": 1e-06, "loss": -0.0018, "num_tokens": 419204026.0, "reward": 0.9218825459480285, "reward_std": 0.1091567412018776, "rewards/accuracy_reward": 0.5072265625, "rewards/batch_coverage_0": 0.31388933062553404, "rewards/batch_coverage_1": 0.31388933062553404, "rewards/batch_coverage_5": 0.3319485247135162, "rewards/brier_reward": 0.7653324604034424, "rewards/confidence_uniqueness_reward": 0.950584864616394, "rewards/format_reward": 0.9978515625, "rewards/frontier_aurc_reward": -0.0037335056811571123, "rewards/frontier_coverage_10": 0.10363066233694554, "rewards/frontier_coverage_15": 0.10363066233694554, "rewards/frontier_coverage_20": 0.10363066233694554, "rewards/frontier_coverage_25": 0.10363066233694554, "rewards/frontier_ece_reward": 0.011668909713625908, "rewards/frontier_entropy_batch_reward": -0.20546606183052063, "signal/accuracy_reward/centered_abs_mean": 0.12281494140625, "signal/accuracy_reward/group_std_mean": 0.16203629672527314, "signal/accuracy_reward/group_zero_std_frac": 0.5375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.061407470703125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.061407470703125, "signal/advantage_abs_mean": 0.08268684893846512, "signal/advantage_pre_scale_abs_mean": 0.08268684893846512, "signal/advantage_pre_scale_std": 0.13070999681949616, "signal/advantage_std": 0.13070999681949616, "signal/batch_coverage_0/centered_abs_mean": 0.1767641067504883, "signal/batch_coverage_0/group_std_mean": 0.22425232231616973, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0022095513995736837, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0022095513995736837, "signal/batch_coverage_1/centered_abs_mean": 0.1767641067504883, "signal/batch_coverage_1/group_std_mean": 0.22425232231616973, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0022095513995736837, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0022095513995736837, "signal/batch_coverage_5/centered_abs_mean": 0.182526159286499, "signal/batch_coverage_5/group_std_mean": 0.2314590573310852, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0022815770469605924, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0022815770469605924, "signal/brier_reward/centered_abs_mean": 0.15626563429832457, "signal/brier_reward/group_std_mean": 0.19891900420188904, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.015626564063131808, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.015626564063131808, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.01565089784562588, "signal/confidence_uniqueness_reward/group_std_mean": 0.025914452597498895, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0015650898218154906, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0015650898218154906, "signal/format_reward/centered_abs_mean": 0.004150390625, "signal/format_reward/group_std_mean": 0.011817089281976223, "signal/format_reward/group_zero_std_frac": 0.934375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0020751953125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0020751953125, "signal/frontier_aurc_reward/centered_abs_mean": 0.0038944244384765624, "signal/frontier_aurc_reward/group_std_mean": 0.006205868348479271, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 4.868030737270601e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 4.868030737270601e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.18902657628059388, "signal/frontier_coverage_10/group_std_mean": 0.24355286359786987, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.002362832240760326, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.002362832240760326, "signal/frontier_coverage_15/centered_abs_mean": 0.18902657628059388, "signal/frontier_coverage_15/group_std_mean": 0.24355286359786987, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002362832240760326, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002362832240760326, "signal/frontier_coverage_20/centered_abs_mean": 0.18902657628059388, "signal/frontier_coverage_20/group_std_mean": 0.24355286359786987, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002362832240760326, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002362832240760326, "signal/frontier_coverage_25/centered_abs_mean": 0.18902657628059388, "signal/frontier_coverage_25/group_std_mean": 0.24355286359786987, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.002362832240760326, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.002362832240760326, "signal/frontier_ece_reward/centered_abs_mean": 0.020860131829977036, "signal/frontier_ece_reward/group_std_mean": 0.027483052387833596, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.002086013229563832, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.002086013229563832, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.27581494450569155, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3534397900104523, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02758149467408657, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02758149467408657, "step": 125 }, { "calibration/aurc": 0.32992934616857483, "calibration/batch_distribution_entropy": 0.9775496438307186, "calibration/buffer_distribution_entropy": 0.998693995931802, "calibration/confidence_entropy": 0.5022215416641719, "calibration/coverage@0%": 0.011020628480682214, "calibration/coverage@1%": 0.011020628480682214, "calibration/coverage@10%": 0.02397572768012531, "calibration/coverage@15%": 0.0667927598184244, "calibration/coverage@20%": 0.11978734307768255, "calibration/coverage@25%": 0.3183175270797303, "calibration/coverage@30%": 0.4109735812665221, "calibration/coverage@5%": 0.01926056350794756, "calibration/ece": 0.1159850406260741, "calibration/mean_confidence": 0.5264952678032878, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 974.6, "completions/max_terminated_length": 974.6, "completions/mean_length": 225.54384765625, "completions/mean_terminated_length": 226.36394348144532, "completions/min_length": 0.0, "completions/min_terminated_length": 103.0, "epoch": 0.416, "grad_norm": 0.0009895272087305784, "learning_rate": 1e-06, "loss": -0.0029, "num_tokens": 436394779.0, "reward": 0.9275412321090698, "reward_std": 0.11238719075918198, "rewards/accuracy_reward": 0.512890625, "rewards/batch_coverage_0": 0.32113032341003417, "rewards/batch_coverage_1": 0.32113032341003417, "rewards/batch_coverage_5": 0.3340096831321716, "rewards/brier_reward": 0.7752441763877869, "rewards/confidence_uniqueness_reward": 0.9506540894508362, "rewards/format_reward": 0.99619140625, "rewards/frontier_aurc_reward": -0.003275211388245225, "rewards/frontier_coverage_10": 0.10501094162464142, "rewards/frontier_coverage_15": 0.10501094162464142, "rewards/frontier_coverage_20": 0.10501094162464142, "rewards/frontier_coverage_25": 0.10501094162464142, "rewards/frontier_ece_reward": 0.010579920560121536, "rewards/frontier_entropy_batch_reward": -0.18060575425624847, "signal/accuracy_reward/centered_abs_mean": 0.12030029296875, "signal/accuracy_reward/group_std_mean": 0.15878662317991257, "signal/accuracy_reward/group_zero_std_frac": 0.55, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.060150146484375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.060150146484375, "signal/advantage_abs_mean": 0.08285276293754577, "signal/advantage_pre_scale_abs_mean": 0.08285276293754577, "signal/advantage_pre_scale_std": 0.13275846391916274, "signal/advantage_std": 0.13275846391916274, "signal/batch_coverage_0/centered_abs_mean": 0.1770572006702423, "signal/batch_coverage_0/group_std_mean": 0.2223646879196167, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002213215082883835, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002213215082883835, "signal/batch_coverage_1/centered_abs_mean": 0.1770572006702423, "signal/batch_coverage_1/group_std_mean": 0.2223646879196167, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002213215082883835, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002213215082883835, "signal/batch_coverage_5/centered_abs_mean": 0.18228316605091094, "signal/batch_coverage_5/group_std_mean": 0.22852807939052583, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.00227853967808187, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.00227853967808187, "signal/brier_reward/centered_abs_mean": 0.15505231022834778, "signal/brier_reward/group_std_mean": 0.19803432524204254, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.0155052300542593, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.0155052300542593, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.016809065267443656, "signal/confidence_uniqueness_reward/group_std_mean": 0.03164429664611816, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0016809065360575914, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0016809065360575914, "signal/format_reward/centered_abs_mean": 0.007330322265625, "signal/format_reward/group_std_mean": 0.020199426077306272, "signal/format_reward/group_zero_std_frac": 0.890625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0036651611328125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0036651611328125, "signal/frontier_aurc_reward/centered_abs_mean": 0.003497358039021492, "signal/frontier_aurc_reward/group_std_mean": 0.005551520269364119, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 4.3716976506402716e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 4.3716976506402716e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1960183411836624, "signal/frontier_coverage_10/group_std_mean": 0.24909888207912445, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0024502291809767483, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0024502291809767483, "signal/frontier_coverage_15/centered_abs_mean": 0.1960183411836624, "signal/frontier_coverage_15/group_std_mean": 0.24909888207912445, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0024502291809767483, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0024502291809767483, "signal/frontier_coverage_20/centered_abs_mean": 0.1960183411836624, "signal/frontier_coverage_20/group_std_mean": 0.24909888207912445, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0024502291809767483, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0024502291809767483, "signal/frontier_coverage_25/centered_abs_mean": 0.1960183411836624, "signal/frontier_coverage_25/group_std_mean": 0.24909888207912445, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0024502291809767483, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0024502291809767483, "signal/frontier_ece_reward/centered_abs_mean": 0.018052129819989205, "signal/frontier_ece_reward/group_std_mean": 0.023573490604758263, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0018052129540592431, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0018052129540592431, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2584738940000534, "signal/frontier_entropy_batch_reward/group_std_mean": 0.33377833366394044, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.025847389921545984, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025847389921545984, "step": 130 }, { "calibration/aurc": 0.27968053052772357, "calibration/batch_distribution_entropy": 0.9848971742652356, "calibration/buffer_distribution_entropy": 0.9988645266187446, "calibration/confidence_entropy": 0.47566178531825526, "calibration/coverage@0%": 0.01338518709745388, "calibration/coverage@1%": 0.01338518709745388, "calibration/coverage@10%": 0.08573433426219641, "calibration/coverage@15%": 0.22350804632463345, "calibration/coverage@20%": 0.32388128989133136, "calibration/coverage@25%": 0.42643376747635103, "calibration/coverage@30%": 0.5482591689862415, "calibration/coverage@5%": 0.029102522098245154, "calibration/ece": 0.11881250045624898, "calibration/mean_confidence": 0.5352764260355155, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 1043.6, "completions/max_terminated_length": 1043.6, "completions/mean_length": 225.38427734375, "completions/mean_terminated_length": 226.55670776367188, "completions/min_length": 0.0, "completions/min_terminated_length": 107.8, "epoch": 0.432, "grad_norm": 0.0011306490050628781, "learning_rate": 1e-06, "loss": -0.0045, "num_tokens": 453717050.0, "reward": 0.9454703330993652, "reward_std": 0.10661281198263169, "rewards/accuracy_reward": 0.5482421875, "rewards/batch_coverage_0": 0.3483874201774597, "rewards/batch_coverage_1": 0.3483874201774597, "rewards/batch_coverage_5": 0.36769152879714967, "rewards/brier_reward": 0.7850048184394837, "rewards/confidence_uniqueness_reward": 0.9492023110389709, "rewards/format_reward": 0.9947265625, "rewards/frontier_aurc_reward": -0.00277026635594666, "rewards/frontier_coverage_10": 0.09695247001945972, "rewards/frontier_coverage_15": 0.09695247001945972, "rewards/frontier_coverage_20": 0.09695249237120152, "rewards/frontier_coverage_25": 0.09685924928635359, "rewards/frontier_ece_reward": 0.011067810282111169, "rewards/frontier_entropy_batch_reward": -0.18659218847751619, "signal/accuracy_reward/centered_abs_mean": 0.108447265625, "signal/accuracy_reward/group_std_mean": 0.14264771342277527, "signal/accuracy_reward/group_zero_std_frac": 0.590625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0542236328125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0542236328125, "signal/advantage_abs_mean": 0.07824351340532303, "signal/advantage_pre_scale_abs_mean": 0.07824351340532303, "signal/advantage_pre_scale_std": 0.12824587225914003, "signal/advantage_std": 0.12824587225914003, "signal/batch_coverage_0/centered_abs_mean": 0.18021714091300964, "signal/batch_coverage_0/group_std_mean": 0.2254747837781906, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0022527142893522976, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0022527142893522976, "signal/batch_coverage_1/centered_abs_mean": 0.18021714091300964, "signal/batch_coverage_1/group_std_mean": 0.2254747837781906, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0022527142893522976, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0022527142893522976, "signal/batch_coverage_5/centered_abs_mean": 0.18509128093719482, "signal/batch_coverage_5/group_std_mean": 0.23147269189357758, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.00231364113278687, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.00231364113278687, "signal/brier_reward/centered_abs_mean": 0.14622356593608857, "signal/brier_reward/group_std_mean": 0.1903170496225357, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.014622356556355954, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.014622356556355954, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.019582652673125266, "signal/confidence_uniqueness_reward/group_std_mean": 0.03780309110879898, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0019582653418183326, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0019582653418183326, "signal/format_reward/centered_abs_mean": 0.01007080078125, "signal/format_reward/group_std_mean": 0.02665354125201702, "signal/format_reward/group_zero_std_frac": 0.859375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.005035400390625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.005035400390625, "signal/frontier_aurc_reward/centered_abs_mean": 0.0032149428967386483, "signal/frontier_aurc_reward/group_std_mean": 0.00542918024584651, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 4.0186786645790565e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 4.0186786645790565e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1889925181865692, "signal/frontier_coverage_10/group_std_mean": 0.2408166140317917, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.00236240653321147, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.00236240653321147, "signal/frontier_coverage_15/centered_abs_mean": 0.1889925181865692, "signal/frontier_coverage_15/group_std_mean": 0.2408166140317917, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.00236240653321147, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.00236240653321147, "signal/frontier_coverage_20/centered_abs_mean": 0.188992440700531, "signal/frontier_coverage_20/group_std_mean": 0.24081650972366334, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0023624055553227662, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0023624055553227662, "signal/frontier_coverage_25/centered_abs_mean": 0.18836949467658998, "signal/frontier_coverage_25/group_std_mean": 0.2400407761335373, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0023546186741441487, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0023546186741441487, "signal/frontier_ece_reward/centered_abs_mean": 0.016344958171248435, "signal/frontier_ece_reward/group_std_mean": 0.021385031566023827, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0016344958916306497, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0016344958916306497, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26241403818130493, "signal/frontier_entropy_batch_reward/group_std_mean": 0.339809650182724, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026241404190659522, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026241404190659522, "step": 135 }, { "calibration/aurc": 0.3075347700092078, "calibration/batch_distribution_entropy": 0.9778138852229581, "calibration/buffer_distribution_entropy": 0.9988175536741257, "calibration/confidence_entropy": 0.5052049944715488, "calibration/coverage@0%": 0.007061170998831215, "calibration/coverage@1%": 0.007061170998831215, "calibration/coverage@10%": 0.08996883308134594, "calibration/coverage@15%": 0.13111866629389196, "calibration/coverage@20%": 0.18968445334104383, "calibration/coverage@25%": 0.2360316931525437, "calibration/coverage@30%": 0.39268369018996413, "calibration/coverage@5%": 0.01059751677486265, "calibration/ece": 0.12882414226166722, "calibration/mean_confidence": 0.5439760984325032, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00390625, "completions/max_length": 1135.8, "completions/max_terminated_length": 1135.8, "completions/mean_length": 230.3421875, "completions/mean_terminated_length": 231.24584350585937, "completions/min_length": 0.0, "completions/min_terminated_length": 109.4, "epoch": 0.448, "grad_norm": 0.0010944586247205734, "learning_rate": 1e-06, "loss": -0.0034, "num_tokens": 471028554.0, "reward": 0.9289451956748962, "reward_std": 0.10082367062568665, "rewards/accuracy_reward": 0.51337890625, "rewards/batch_coverage_0": 0.34634156823158263, "rewards/batch_coverage_1": 0.34634156823158263, "rewards/batch_coverage_5": 0.3614945590496063, "rewards/brier_reward": 0.7845125198364258, "rewards/confidence_uniqueness_reward": 0.9499234199523926, "rewards/format_reward": 0.99599609375, "rewards/frontier_aurc_reward": -0.0029740627855062486, "rewards/frontier_coverage_10": 0.11677841991186141, "rewards/frontier_coverage_15": 0.11677840203046799, "rewards/frontier_coverage_20": 0.11677816808223725, "rewards/frontier_coverage_25": 0.11593354344367982, "rewards/frontier_ece_reward": 0.009015307389199733, "rewards/frontier_entropy_batch_reward": -0.1905584901571274, "signal/accuracy_reward/centered_abs_mean": 0.100823974609375, "signal/accuracy_reward/group_std_mean": 0.13174283653497695, "signal/accuracy_reward/group_zero_std_frac": 0.63125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0504119873046875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0504119873046875, "signal/advantage_abs_mean": 0.0752875879406929, "signal/advantage_pre_scale_abs_mean": 0.0752875879406929, "signal/advantage_pre_scale_std": 0.12117603570222854, "signal/advantage_std": 0.12117603570222854, "signal/batch_coverage_0/centered_abs_mean": 0.1856759399175644, "signal/batch_coverage_0/group_std_mean": 0.2310125172138214, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002320949314162135, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002320949314162135, "signal/batch_coverage_1/centered_abs_mean": 0.1856759399175644, "signal/batch_coverage_1/group_std_mean": 0.2310125172138214, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002320949314162135, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002320949314162135, "signal/batch_coverage_5/centered_abs_mean": 0.1907913237810135, "signal/batch_coverage_5/group_std_mean": 0.23715128302574157, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002384891640394926, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002384891640394926, "signal/brier_reward/centered_abs_mean": 0.14696918427944183, "signal/brier_reward/group_std_mean": 0.18939995765686035, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.014696918241679669, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.014696918241679669, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.017957666888833046, "signal/confidence_uniqueness_reward/group_std_mean": 0.031488869711756705, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0017957667354494334, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0017957667354494334, "signal/format_reward/centered_abs_mean": 0.007562255859375, "signal/format_reward/group_std_mean": 0.018984930217266084, "signal/format_reward/group_zero_std_frac": 0.903125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0037811279296875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0037811279296875, "signal/frontier_aurc_reward/centered_abs_mean": 0.003044320410117507, "signal/frontier_aurc_reward/group_std_mean": 0.005033304076641798, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.805400483543053e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.805400483543053e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.19509401619434358, "signal/frontier_coverage_10/group_std_mean": 0.2471470683813095, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.002438675286248326, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.002438675286248326, "signal/frontier_coverage_15/centered_abs_mean": 0.1950937807559967, "signal/frontier_coverage_15/group_std_mean": 0.24714677929878234, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0024386723060160877, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0024386723060160877, "signal/frontier_coverage_20/centered_abs_mean": 0.19508895874023438, "signal/frontier_coverage_20/group_std_mean": 0.24714082777500151, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.00243861204944551, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.00243861204944551, "signal/frontier_coverage_25/centered_abs_mean": 0.18899599015712737, "signal/frontier_coverage_25/group_std_mean": 0.23964630663394929, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0023624499328434466, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0023624499328434466, "signal/frontier_ece_reward/centered_abs_mean": 0.014707912318408489, "signal/frontier_ece_reward/group_std_mean": 0.01931345872581005, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0014707912458106875, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0014707912458106875, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26829124689102174, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3444648444652557, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026829124987125398, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026829124987125398, "step": 140 }, { "calibration/aurc": 0.42251264064761107, "calibration/batch_distribution_entropy": 0.983004542921255, "calibration/buffer_distribution_entropy": 0.9986651457602921, "calibration/confidence_entropy": 0.5162808328423759, "calibration/coverage@0%": 0.00864131900304326, "calibration/coverage@1%": 0.00864131900304326, "calibration/coverage@10%": 0.027828628450599767, "calibration/coverage@15%": 0.03683519902565004, "calibration/coverage@20%": 0.04780025162364647, "calibration/coverage@25%": 0.14439116206807828, "calibration/coverage@30%": 0.2237301829793467, "calibration/coverage@5%": 0.009427173619939135, "calibration/ece": 0.13667226734688462, "calibration/mean_confidence": 0.48400141300242777, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00556640625, "completions/max_length": 1049.8, "completions/max_terminated_length": 1049.8, "completions/mean_length": 234.3, "completions/mean_terminated_length": 235.61791381835937, "completions/min_length": 0.0, "completions/min_terminated_length": 110.2, "epoch": 0.464, "grad_norm": 0.0008999215206131339, "learning_rate": 1e-06, "loss": -0.0048, "num_tokens": 488598602.0, "reward": 0.9000043392181396, "reward_std": 0.10157125294208527, "rewards/accuracy_reward": 0.465234375, "rewards/batch_coverage_0": 0.3196440637111664, "rewards/batch_coverage_1": 0.3196440637111664, "rewards/batch_coverage_5": 0.3420140087604523, "rewards/brier_reward": 0.7600026369094849, "rewards/confidence_uniqueness_reward": 0.9480797171592712, "rewards/format_reward": 0.9943359375, "rewards/frontier_aurc_reward": -0.003716340381652117, "rewards/frontier_coverage_10": 0.1241892084479332, "rewards/frontier_coverage_15": 0.12418920248746872, "rewards/frontier_coverage_20": 0.12418830581009388, "rewards/frontier_coverage_25": 0.12012989968061447, "rewards/frontier_ece_reward": 0.006515136826783419, "rewards/frontier_entropy_batch_reward": -0.19619101881980897, "signal/accuracy_reward/centered_abs_mean": 0.08768310546875, "signal/accuracy_reward/group_std_mean": 0.12859813421964644, "signal/accuracy_reward/group_zero_std_frac": 0.5875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.043841552734375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.043841552734375, "signal/advantage_abs_mean": 0.07088839709758758, "signal/advantage_pre_scale_abs_mean": 0.07088839709758758, "signal/advantage_pre_scale_std": 0.11998113542795182, "signal/advantage_std": 0.11998113542795182, "signal/batch_coverage_0/centered_abs_mean": 0.17328222393989562, "signal/batch_coverage_0/group_std_mean": 0.2200840651988983, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.00216602785512805, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.00216602785512805, "signal/batch_coverage_1/centered_abs_mean": 0.17328222393989562, "signal/batch_coverage_1/group_std_mean": 0.2200840651988983, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.00216602785512805, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.00216602785512805, "signal/batch_coverage_5/centered_abs_mean": 0.17990256249904632, "signal/batch_coverage_5/group_std_mean": 0.22849339842796326, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002248782105743885, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002248782105743885, "signal/brier_reward/centered_abs_mean": 0.1525415748357773, "signal/brier_reward/group_std_mean": 0.19790156781673432, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.015254157595336437, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.015254157595336437, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.020429689064621927, "signal/confidence_uniqueness_reward/group_std_mean": 0.037467952072620395, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.002042968920432031, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.002042968920432031, "signal/format_reward/centered_abs_mean": 0.01065673828125, "signal/format_reward/group_std_mean": 0.02590869925916195, "signal/format_reward/group_zero_std_frac": 0.871875, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.005328369140625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.005328369140625, "signal/frontier_aurc_reward/centered_abs_mean": 0.003307991800829768, "signal/frontier_aurc_reward/group_std_mean": 0.005511091928929091, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 4.134989867452532e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 4.134989867452532e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.18691303730010986, "signal/frontier_coverage_10/group_std_mean": 0.24188108146190643, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0023364129941910506, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0023364129941910506, "signal/frontier_coverage_15/centered_abs_mean": 0.18691303133964537, "signal/frontier_coverage_15/group_std_mean": 0.24188107550144194, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0023364129476249216, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0023364129476249216, "signal/frontier_coverage_20/centered_abs_mean": 0.18691181242465973, "signal/frontier_coverage_20/group_std_mean": 0.24187954962253572, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002336397673934698, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002336397673934698, "signal/frontier_coverage_25/centered_abs_mean": 0.1797201693058014, "signal/frontier_coverage_25/group_std_mean": 0.23255808353424073, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0022465021349489687, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0022465021349489687, "signal/frontier_ece_reward/centered_abs_mean": 0.013078332878649235, "signal/frontier_ece_reward/group_std_mean": 0.017391672357916832, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0013078333344310523, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0013078333344310523, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2678228974342346, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3448445379734039, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026782290264964103, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026782290264964103, "step": 145 }, { "calibration/aurc": 0.29854593729506496, "calibration/batch_distribution_entropy": 0.9785048140385607, "calibration/buffer_distribution_entropy": 0.9985097089757359, "calibration/confidence_entropy": 0.4883856841168012, "calibration/coverage@0%": 0.00823014043878775, "calibration/coverage@1%": 0.00823014043878775, "calibration/coverage@10%": 0.07507574687443132, "calibration/coverage@15%": 0.18631105009225307, "calibration/coverage@20%": 0.2744652157035149, "calibration/coverage@25%": 0.36036190759351516, "calibration/coverage@30%": 0.47275760831243013, "calibration/coverage@5%": 0.00823014043878775, "calibration/ece": 0.13808430776716613, "calibration/mean_confidence": 0.5168411793960647, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 960.2, "completions/max_terminated_length": 960.2, "completions/mean_length": 237.40068359375, "completions/mean_terminated_length": 238.7991180419922, "completions/min_length": 0.0, "completions/min_terminated_length": 117.8, "epoch": 0.48, "grad_norm": 0.0010402585612609982, "learning_rate": 1e-06, "loss": -0.0046, "num_tokens": 506077617.0, "reward": 0.9272619962692261, "reward_std": 0.10827716141939163, "rewards/accuracy_reward": 0.51328125, "rewards/batch_coverage_0": 0.33160064220428465, "rewards/batch_coverage_1": 0.33160064220428465, "rewards/batch_coverage_5": 0.35568556785583494, "rewards/brier_reward": 0.7722211837768554, "rewards/confidence_uniqueness_reward": 0.9483663082122803, "rewards/format_reward": 0.99404296875, "rewards/frontier_aurc_reward": -0.003292375709861517, "rewards/frontier_coverage_10": 0.10929831266403198, "rewards/frontier_coverage_15": 0.10929830670356751, "rewards/frontier_coverage_20": 0.10927678644657135, "rewards/frontier_coverage_25": 0.10621259808540344, "rewards/frontier_ece_reward": 0.007439652178436518, "rewards/frontier_entropy_batch_reward": -0.1732386976480484, "signal/accuracy_reward/centered_abs_mean": 0.1136962890625, "signal/accuracy_reward/group_std_mean": 0.15411219894886016, "signal/accuracy_reward/group_zero_std_frac": 0.540625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.05684814453125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.05684814453125, "signal/advantage_abs_mean": 0.07793455421924592, "signal/advantage_pre_scale_abs_mean": 0.07793455421924592, "signal/advantage_pre_scale_std": 0.12913145422935485, "signal/advantage_std": 0.12913145422935485, "signal/batch_coverage_0/centered_abs_mean": 0.17881105542182923, "signal/batch_coverage_0/group_std_mean": 0.22615096271038054, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002235138276591897, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002235138276591897, "signal/batch_coverage_1/centered_abs_mean": 0.17881105542182923, "signal/batch_coverage_1/group_std_mean": 0.22615096271038054, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002235138276591897, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002235138276591897, "signal/batch_coverage_5/centered_abs_mean": 0.18588078916072845, "signal/batch_coverage_5/group_std_mean": 0.23523793816566468, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0023235098458826544, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0023235098458826544, "signal/brier_reward/centered_abs_mean": 0.1502091884613037, "signal/brier_reward/group_std_mean": 0.19434000849723815, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.015020918473601342, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.015020918473601342, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.020434437319636345, "signal/confidence_uniqueness_reward/group_std_mean": 0.03880791217088699, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0020434437319636345, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0020434437319636345, "signal/format_reward/centered_abs_mean": 0.011273193359375, "signal/format_reward/group_std_mean": 0.02798873223364353, "signal/format_reward/group_zero_std_frac": 0.859375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0056365966796875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0056365966796875, "signal/frontier_aurc_reward/centered_abs_mean": 0.003298774268478155, "signal/frontier_aurc_reward/group_std_mean": 0.005438829120248556, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 4.123467952013016e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 4.123467952013016e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.19652977883815764, "signal/frontier_coverage_10/group_std_mean": 0.25134665369987486, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0024566222913563252, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0024566222913563252, "signal/frontier_coverage_15/centered_abs_mean": 0.19652974903583526, "signal/frontier_coverage_15/group_std_mean": 0.25134662389755247, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0024566218722611665, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0024566218722611665, "signal/frontier_coverage_20/centered_abs_mean": 0.19649703204631805, "signal/frontier_coverage_20/group_std_mean": 0.2513046681880951, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0024562129285186527, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0024562129285186527, "signal/frontier_coverage_25/centered_abs_mean": 0.18423069417476653, "signal/frontier_coverage_25/group_std_mean": 0.23614015877246858, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.002302883705124259, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.002302883705124259, "signal/frontier_ece_reward/centered_abs_mean": 0.012983166612684727, "signal/frontier_ece_reward/group_std_mean": 0.01702551692724228, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0012983167078346014, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0012983167078346014, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.24633871614933014, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3230536639690399, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.024633872136473656, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.024633872136473656, "step": 150 }, { "epoch": 0.48, "eval_calibration/aurc": 0.4432238858374702, "eval_calibration/batch_distribution_entropy": 0.9139135703169725, "eval_calibration/buffer_distribution_entropy": 0.9983479443854384, "eval_calibration/confidence_entropy": 0.5095747919766502, "eval_calibration/coverage@0%": 0.047379032258064516, "eval_calibration/coverage@1%": 0.047379032258064516, "eval_calibration/coverage@10%": 0.047379032258064516, "eval_calibration/coverage@15%": 0.10206653225806452, "eval_calibration/coverage@20%": 0.1489415322580645, "eval_calibration/coverage@25%": 0.21194556451612903, "eval_calibration/coverage@30%": 0.2752016129032258, "eval_calibration/coverage@5%": 0.047379032258064516, "eval_calibration/ece": 0.20855649847263277, "eval_calibration/mean_confidence": 0.514187655429143, "eval_completions/clipped_ratio": 0.00390625, "eval_completions/max_length": 535.5, "eval_completions/max_terminated_length": 535.5, "eval_completions/mean_length": 239.47103881835938, "eval_completions/mean_terminated_length": 240.39476013183594, "eval_completions/min_length": 67.5, "eval_completions/min_terminated_length": 135.0, "eval_loss": 0.0, "eval_num_tokens": 506077617.0, "eval_reward": 0.7855269461870193, "eval_reward_std": 0.23481816053390503, "eval_rewards/accuracy_reward": 0.412109375, "eval_rewards/batch_coverage_0": 0.14985975436866283, "eval_rewards/batch_coverage_1": 0.14985975436866283, "eval_rewards/batch_coverage_5": 0.14985975436866283, "eval_rewards/brier_reward": 0.7721172869205475, "eval_rewards/confidence_uniqueness_reward": 0.892664760351181, "eval_rewards/format_reward": 0.99609375, "eval_rewards/frontier_aurc_reward": -0.004516357439570129, "eval_rewards/frontier_coverage_10": 0.17051639407873154, "eval_rewards/frontier_coverage_15": 0.17051635310053825, "eval_rewards/frontier_coverage_20": 0.17021949589252472, "eval_rewards/frontier_coverage_25": 0.1506239641457796, "eval_rewards/frontier_ece_reward": 0.007197963073849678, "eval_rewards/frontier_entropy_batch_reward": -0.99609375, "eval_runtime": 40.6167, "eval_samples_per_second": 12.31, "eval_signal/accuracy_reward/centered_abs_mean": 0.4674072265625, "eval_signal/accuracy_reward/group_std_mean": 0.4906700849533081, "eval_signal/accuracy_reward/group_zero_std_frac": 0.0, "eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.23370361328125, "eval_signal/accuracy_reward/weight": 0.5, "eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.23370361328125, "eval_signal/advantage_abs_mean": 0.2164582908153534, "eval_signal/advantage_pre_scale_abs_mean": 0.2164582908153534, "eval_signal/advantage_pre_scale_std": 0.23250839859247208, "eval_signal/advantage_std": 0.23250839859247208, "eval_signal/batch_coverage_0/centered_abs_mean": 0.2798871770501137, "eval_signal/batch_coverage_0/group_std_mean": 0.3548194020986557, "eval_signal/batch_coverage_0/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0034985897946171463, "eval_signal/batch_coverage_0/weight": 0.012500000186264515, "eval_signal/batch_coverage_0/weighted_centered_abs_mean": 0.0034985897946171463, "eval_signal/batch_coverage_1/centered_abs_mean": 0.2798871770501137, "eval_signal/batch_coverage_1/group_std_mean": 0.3548194020986557, "eval_signal/batch_coverage_1/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0034985897946171463, "eval_signal/batch_coverage_1/weight": 0.012500000186264515, "eval_signal/batch_coverage_1/weighted_centered_abs_mean": 0.0034985897946171463, "eval_signal/batch_coverage_5/centered_abs_mean": 0.2798871770501137, "eval_signal/batch_coverage_5/group_std_mean": 0.3548194020986557, "eval_signal/batch_coverage_5/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0034985897946171463, "eval_signal/batch_coverage_5/weight": 0.012500000186264515, "eval_signal/batch_coverage_5/weighted_centered_abs_mean": 0.0034985897946171463, "eval_signal/brier_reward/centered_abs_mean": 0.20697294175624847, "eval_signal/brier_reward/group_std_mean": 0.2567337602376938, "eval_signal/brier_reward/group_zero_std_frac": 0.0, "eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.020697294268757105, "eval_signal/brier_reward/weight": 0.10000000149011612, "eval_signal/brier_reward/weighted_centered_abs_mean": 0.020697294268757105, "eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.04560753609985113, "eval_signal/confidence_uniqueness_reward/group_std_mean": 0.06350228935480118, "eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.004560753470286727, "eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004560753470286727, "eval_signal/format_reward/centered_abs_mean": 0.007568359375, "eval_signal/format_reward/group_std_mean": 0.022097086533904076, "eval_signal/format_reward/group_zero_std_frac": 0.875, "eval_signal/format_reward/scaled_weighted_centered_abs_mean": 0.0037841796875, "eval_signal/format_reward/weight": 0.5, "eval_signal/format_reward/weighted_centered_abs_mean": 0.0037841796875, "eval_signal/frontier_aurc_reward/centered_abs_mean": 0.0057645016349852085, "eval_signal/frontier_aurc_reward/group_std_mean": 0.010412166360765696, "eval_signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 7.205627298390027e-05, "eval_signal/frontier_aurc_reward/weight": 0.012500000186264515, "eval_signal/frontier_aurc_reward/weighted_centered_abs_mean": 7.205627298390027e-05, "eval_signal/frontier_coverage_10/centered_abs_mean": 0.32263241708278656, "eval_signal/frontier_coverage_10/group_std_mean": 0.4069429710507393, "eval_signal/frontier_coverage_10/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.004032905271742493, "eval_signal/frontier_coverage_10/weight": 0.012500000186264515, "eval_signal/frontier_coverage_10/weighted_centered_abs_mean": 0.004032905271742493, "eval_signal/frontier_coverage_15/centered_abs_mean": 0.3226323649287224, "eval_signal/frontier_coverage_15/group_std_mean": 0.4069429188966751, "eval_signal/frontier_coverage_15/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.004032904573250562, "eval_signal/frontier_coverage_15/weight": 0.012500000186264515, "eval_signal/frontier_coverage_15/weighted_centered_abs_mean": 0.004032904573250562, "eval_signal/frontier_coverage_20/centered_abs_mean": 0.322020560503006, "eval_signal/frontier_coverage_20/group_std_mean": 0.4062218815088272, "eval_signal/frontier_coverage_20/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.00402525719255209, "eval_signal/frontier_coverage_20/weight": 0.012500000186264515, "eval_signal/frontier_coverage_20/weighted_centered_abs_mean": 0.00402525719255209, "eval_signal/frontier_coverage_25/centered_abs_mean": 0.28987917304039, "eval_signal/frontier_coverage_25/group_std_mean": 0.36807454377412796, "eval_signal/frontier_coverage_25/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0036234896979294717, "eval_signal/frontier_coverage_25/weight": 0.012500000186264515, "eval_signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0036234896979294717, "eval_signal/frontier_ece_reward/centered_abs_mean": 0.0154415019787848, "eval_signal/frontier_ece_reward/group_std_mean": 0.020852980203926563, "eval_signal/frontier_ece_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.001544150203699246, "eval_signal/frontier_ece_reward/weight": 0.10000000149011612, "eval_signal/frontier_ece_reward/weighted_centered_abs_mean": 0.001544150203699246, "eval_signal/frontier_entropy_batch_reward/centered_abs_mean": 0.007568359375, "eval_signal/frontier_entropy_batch_reward/group_std_mean": 0.022097086533904076, "eval_signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.875, "eval_signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.0007568359724245965, "eval_signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "eval_signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.0007568359724245965, "eval_steps_per_second": 0.098, "step": 150 }, { "calibration/aurc": 0.3985256133510302, "calibration/batch_distribution_entropy": 0.9846702862712318, "calibration/buffer_distribution_entropy": 0.9982634110224419, "calibration/confidence_entropy": 0.48897491181205177, "calibration/coverage@0%": 0.006277618690383277, "calibration/coverage@1%": 0.006277618690383277, "calibration/coverage@10%": 0.08878158833271177, "calibration/coverage@15%": 0.11353600876493182, "calibration/coverage@20%": 0.1512493409961207, "calibration/coverage@25%": 0.2231488959492347, "calibration/coverage@30%": 0.2874845611670941, "calibration/coverage@5%": 0.006277618690383277, "calibration/ece": 0.1460139418519541, "calibration/mean_confidence": 0.5192476820572416, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0041015625, "completions/max_length": 938.8, "completions/max_terminated_length": 938.8, "completions/mean_length": 237.85556640625, "completions/mean_terminated_length": 238.83727111816407, "completions/min_length": 0.0, "completions/min_terminated_length": 112.6, "epoch": 0.496, "grad_norm": 0.000891564937774092, "learning_rate": 1e-06, "loss": -0.0033, "num_tokens": 523821098.0, "reward": 0.9410536408424377, "reward_std": 0.10202527791261673, "rewards/accuracy_reward": 0.5447265625, "rewards/batch_coverage_0": 0.3223016023635864, "rewards/batch_coverage_1": 0.3223016023635864, "rewards/batch_coverage_5": 0.35361250042915343, "rewards/brier_reward": 0.7729079246520996, "rewards/confidence_uniqueness_reward": 0.9500503778457642, "rewards/format_reward": 0.9958984375, "rewards/frontier_aurc_reward": -0.0032522575929760933, "rewards/frontier_coverage_10": 0.08487666845321655, "rewards/frontier_coverage_15": 0.08487665057182311, "rewards/frontier_coverage_20": 0.08486807942390442, "rewards/frontier_coverage_25": 0.07269335314631462, "rewards/frontier_ece_reward": 0.006895613670349121, "rewards/frontier_entropy_batch_reward": -0.1877274215221405, "signal/accuracy_reward/centered_abs_mean": 0.0974853515625, "signal/accuracy_reward/group_std_mean": 0.13464866876602172, "signal/accuracy_reward/group_zero_std_frac": 0.59375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.04874267578125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.04874267578125, "signal/advantage_abs_mean": 0.07363522797822952, "signal/advantage_pre_scale_abs_mean": 0.07363522797822952, "signal/advantage_pre_scale_std": 0.12118377834558487, "signal/advantage_std": 0.12118377834558487, "signal/batch_coverage_0/centered_abs_mean": 0.1734347939491272, "signal/batch_coverage_0/group_std_mean": 0.2191330075263977, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.00216793492436409, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.00216793492436409, "signal/batch_coverage_1/centered_abs_mean": 0.1734347939491272, "signal/batch_coverage_1/group_std_mean": 0.2191330075263977, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.00216793492436409, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.00216793492436409, "signal/batch_coverage_5/centered_abs_mean": 0.18376100659370423, "signal/batch_coverage_5/group_std_mean": 0.2322361171245575, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0022970126010477543, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0022970126010477543, "signal/brier_reward/centered_abs_mean": 0.14599907994270325, "signal/brier_reward/group_std_mean": 0.19045807123184205, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.014599908515810967, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.014599908515810967, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.01802073922008276, "signal/confidence_uniqueness_reward/group_std_mean": 0.03199348896741867, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.001802074001170695, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001802074001170695, "signal/format_reward/centered_abs_mean": 0.0077880859375, "signal/format_reward/group_std_mean": 0.019661218300461768, "signal/format_reward/group_zero_std_frac": 0.9, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00389404296875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00389404296875, "signal/frontier_aurc_reward/centered_abs_mean": 0.00322837196290493, "signal/frontier_aurc_reward/group_std_mean": 0.005327490624040365, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 4.035465099150315e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 4.035465099150315e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.178617399930954, "signal/frontier_coverage_10/group_std_mean": 0.23152371048927306, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0022327175829559565, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0022327175829559565, "signal/frontier_coverage_15/centered_abs_mean": 0.17861737310886383, "signal/frontier_coverage_15/group_std_mean": 0.23152367174625396, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0022327172104269267, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0022327172104269267, "signal/frontier_coverage_20/centered_abs_mean": 0.1782741516828537, "signal/frontier_coverage_20/group_std_mean": 0.2310851037502289, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0022284269332885744, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0022284269332885744, "signal/frontier_coverage_25/centered_abs_mean": 0.15444442629814148, "signal/frontier_coverage_25/group_std_mean": 0.20094362497329712, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0019305553054437042, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0019305553054437042, "signal/frontier_ece_reward/centered_abs_mean": 0.01217555645853281, "signal/frontier_ece_reward/group_std_mean": 0.015891730785369873, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.001217555650509894, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.001217555650509894, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.266143387556076, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3460667371749878, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026614339277148246, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026614339277148246, "step": 155 }, { "calibration/aurc": 0.3292563617036605, "calibration/batch_distribution_entropy": 0.984055561246965, "calibration/buffer_distribution_entropy": 0.9980414670886384, "calibration/confidence_entropy": 0.5075650399393163, "calibration/coverage@0%": 0.03193880410666863, "calibration/coverage@1%": 0.04771789680883826, "calibration/coverage@10%": 0.14278693033941026, "calibration/coverage@15%": 0.22444913887200868, "calibration/coverage@20%": 0.3302268356190303, "calibration/coverage@25%": 0.419236938773882, "calibration/coverage@30%": 0.48505080806227274, "calibration/coverage@5%": 0.09308278832757592, "calibration/ece": 0.14539898940259627, "calibration/mean_confidence": 0.5135734566733303, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.004296875, "completions/max_length": 950.0, "completions/max_terminated_length": 950.0, "completions/mean_length": 233.966015625, "completions/mean_terminated_length": 234.9716003417969, "completions/min_length": 0.0, "completions/min_terminated_length": 115.8, "epoch": 0.512, "grad_norm": 0.0010575206251814961, "learning_rate": 1e-06, "loss": -0.0035, "num_tokens": 541362574.0, "reward": 0.9411205768585205, "reward_std": 0.1072201356291771, "rewards/accuracy_reward": 0.53818359375, "rewards/batch_coverage_0": 0.3274132668972015, "rewards/batch_coverage_1": 0.3274132668972015, "rewards/batch_coverage_5": 0.363258159160614, "rewards/brier_reward": 0.7896447658538819, "rewards/confidence_uniqueness_reward": 0.9500052094459533, "rewards/format_reward": 0.9955078125, "rewards/frontier_aurc_reward": -0.0027379343286156655, "rewards/frontier_coverage_10": 0.10332640781998634, "rewards/frontier_coverage_15": 0.10325794667005539, "rewards/frontier_coverage_20": 0.10302141606807709, "rewards/frontier_coverage_25": 0.08712001293897628, "rewards/frontier_ece_reward": 0.0076949925161898134, "rewards/frontier_entropy_batch_reward": -0.1811053991317749, "signal/accuracy_reward/centered_abs_mean": 0.107037353515625, "signal/accuracy_reward/group_std_mean": 0.14526507556438445, "signal/accuracy_reward/group_zero_std_frac": 0.5625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0535186767578125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0535186767578125, "signal/advantage_abs_mean": 0.07822389602661133, "signal/advantage_pre_scale_abs_mean": 0.07822389602661133, "signal/advantage_pre_scale_std": 0.12810295820236206, "signal/advantage_std": 0.12810295820236206, "signal/batch_coverage_0/centered_abs_mean": 0.1684813231229782, "signal/batch_coverage_0/group_std_mean": 0.21425497829914092, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002106016594916582, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002106016594916582, "signal/batch_coverage_1/centered_abs_mean": 0.1684813231229782, "signal/batch_coverage_1/group_std_mean": 0.21425497829914092, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002106016594916582, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002106016594916582, "signal/batch_coverage_5/centered_abs_mean": 0.17979929745197296, "signal/batch_coverage_5/group_std_mean": 0.22917997539043428, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0022474912460893393, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0022474912460893393, "signal/brier_reward/centered_abs_mean": 0.14208460748195648, "signal/brier_reward/group_std_mean": 0.18435995876789094, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01420846078544855, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01420846078544855, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.01836678050458431, "signal/confidence_uniqueness_reward/group_std_mean": 0.033973584324121474, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.00183667812962085, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.00183667812962085, "signal/format_reward/centered_abs_mean": 0.008544921875, "signal/format_reward/group_std_mean": 0.022174911201000215, "signal/format_reward/group_zero_std_frac": 0.884375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0042724609375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0042724609375, "signal/frontier_aurc_reward/centered_abs_mean": 0.003006741916760802, "signal/frontier_aurc_reward/group_std_mean": 0.004906453611329198, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.758427446882706e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.758427446882706e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.17261604964733124, "signal/frontier_coverage_10/group_std_mean": 0.22368737757205964, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0021577006205916404, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0021577006205916404, "signal/frontier_coverage_15/centered_abs_mean": 0.17251891195774077, "signal/frontier_coverage_15/group_std_mean": 0.22356524765491487, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002156486501917243, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002156486501917243, "signal/frontier_coverage_20/centered_abs_mean": 0.17205561995506286, "signal/frontier_coverage_20/group_std_mean": 0.22298562824726104, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002150695282034576, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002150695282034576, "signal/frontier_coverage_25/centered_abs_mean": 0.13892405182123185, "signal/frontier_coverage_25/group_std_mean": 0.18110643625259398, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.001736550685018301, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.001736550685018301, "signal/frontier_ece_reward/centered_abs_mean": 0.011249086447060108, "signal/frontier_ece_reward/group_std_mean": 0.014791714400053025, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0011249086586758495, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0011249086586758495, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2635261654853821, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34212314486503603, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026352616772055627, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026352616772055627, "step": 160 }, { "calibration/aurc": 0.2321895035080715, "calibration/batch_distribution_entropy": 0.9888145428142521, "calibration/buffer_distribution_entropy": 0.9981974317146705, "calibration/confidence_entropy": 0.4900033387425869, "calibration/coverage@0%": 0.016876803791043043, "calibration/coverage@1%": 0.016876803791043043, "calibration/coverage@10%": 0.16484903447325677, "calibration/coverage@15%": 0.3029497689952191, "calibration/coverage@20%": 0.5038255087719687, "calibration/coverage@25%": 0.6276035428164943, "calibration/coverage@30%": 0.7384158368597464, "calibration/coverage@5%": 0.05818321984859032, "calibration/ece": 0.11771511137981974, "calibration/mean_confidence": 0.5037801664672854, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 1018.8, "completions/max_terminated_length": 1018.8, "completions/mean_length": 235.12431640625, "completions/mean_terminated_length": 236.11756286621093, "completions/min_length": 0.0, "completions/min_terminated_length": 111.4, "epoch": 0.528, "grad_norm": 0.0009307150030508637, "learning_rate": 1e-06, "loss": -0.0032, "num_tokens": 558799783.0, "reward": 0.9391830682754516, "reward_std": 0.10343104153871537, "rewards/accuracy_reward": 0.53525390625, "rewards/batch_coverage_0": 0.34922213554382325, "rewards/batch_coverage_1": 0.34922213554382325, "rewards/batch_coverage_5": 0.36895452737808226, "rewards/brier_reward": 0.7850141882896423, "rewards/confidence_uniqueness_reward": 0.9499168276786805, "rewards/format_reward": 0.995703125, "rewards/frontier_aurc_reward": -0.002615780150517821, "rewards/frontier_coverage_10": 0.11019408032298088, "rewards/frontier_coverage_15": 0.11013314723968506, "rewards/frontier_coverage_20": 0.1097121961414814, "rewards/frontier_coverage_25": 0.08989185839891434, "rewards/frontier_ece_reward": 0.00693745082244277, "rewards/frontier_entropy_batch_reward": -0.19041245877742768, "signal/accuracy_reward/centered_abs_mean": 0.112689208984375, "signal/accuracy_reward/group_std_mean": 0.1498502403497696, "signal/accuracy_reward/group_zero_std_frac": 0.565625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0563446044921875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0563446044921875, "signal/advantage_abs_mean": 0.0762990415096283, "signal/advantage_pre_scale_abs_mean": 0.0762990415096283, "signal/advantage_pre_scale_std": 0.1256231039762497, "signal/advantage_std": 0.1256231039762497, "signal/batch_coverage_0/centered_abs_mean": 0.16967527866363524, "signal/batch_coverage_0/group_std_mean": 0.21518825590610505, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0021209409926086662, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0021209409926086662, "signal/batch_coverage_1/centered_abs_mean": 0.16967527866363524, "signal/batch_coverage_1/group_std_mean": 0.21518825590610505, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0021209409926086662, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0021209409926086662, "signal/batch_coverage_5/centered_abs_mean": 0.17531973421573638, "signal/batch_coverage_5/group_std_mean": 0.2226909279823303, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002191496593877673, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002191496593877673, "signal/brier_reward/centered_abs_mean": 0.13886926770210267, "signal/brier_reward/group_std_mean": 0.17950203120708466, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01388692706823349, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01388692706823349, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.018081923201680183, "signal/confidence_uniqueness_reward/group_std_mean": 0.03330075740814209, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0018081923946738244, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0018081923946738244, "signal/format_reward/centered_abs_mean": 0.00819091796875, "signal/format_reward/group_std_mean": 0.02146557830274105, "signal/format_reward/group_zero_std_frac": 0.8875, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.004095458984375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.004095458984375, "signal/frontier_aurc_reward/centered_abs_mean": 0.0027651645243167875, "signal/frontier_aurc_reward/group_std_mean": 0.0047821936197578905, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.456455815467052e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.456455815467052e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.189670592546463, "signal/frontier_coverage_10/group_std_mean": 0.24423516988754274, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0023708823602646588, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0023708823602646588, "signal/frontier_coverage_15/centered_abs_mean": 0.18952684104442596, "signal/frontier_coverage_15/group_std_mean": 0.2440561443567276, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0023690855130553247, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0023690855130553247, "signal/frontier_coverage_20/centered_abs_mean": 0.18865520060062407, "signal/frontier_coverage_20/group_std_mean": 0.2429677039384842, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002358190016821027, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002358190016821027, "signal/frontier_coverage_25/centered_abs_mean": 0.14281097054481506, "signal/frontier_coverage_25/group_std_mean": 0.18492644131183625, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0017851371318101882, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0017851371318101882, "signal/frontier_ece_reward/centered_abs_mean": 0.010328078456223012, "signal/frontier_ece_reward/group_std_mean": 0.013512607477605344, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0010328078875318169, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0010328078875318169, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2657435178756714, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3406790018081665, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026574351638555527, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026574351638555527, "step": 165 }, { "calibration/aurc": 0.24466753067109157, "calibration/batch_distribution_entropy": 0.9845321085764469, "calibration/buffer_distribution_entropy": 0.9983160910516826, "calibration/confidence_entropy": 0.48649404268242924, "calibration/coverage@0%": 0.03985215875733855, "calibration/coverage@1%": 0.03985215875733855, "calibration/coverage@10%": 0.14969040484344423, "calibration/coverage@15%": 0.24544168909001957, "calibration/coverage@20%": 0.33300895914872797, "calibration/coverage@25%": 0.5855117111056751, "calibration/coverage@30%": 0.7062989236790607, "calibration/coverage@5%": 0.08129586594911938, "calibration/ece": 0.09074984738892926, "calibration/mean_confidence": 0.5056297555414678, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00234375, "completions/max_length": 770.8, "completions/max_terminated_length": 770.8, "completions/mean_length": 233.02626953125, "completions/mean_terminated_length": 233.5727294921875, "completions/min_length": 0.0, "completions/min_terminated_length": 106.8, "epoch": 0.544, "grad_norm": 0.0009329225867986679, "learning_rate": 1e-06, "loss": -0.0018, "num_tokens": 576349556.0, "reward": 0.9510281801223754, "reward_std": 0.10431258380413055, "rewards/accuracy_reward": 0.5623046875, "rewards/batch_coverage_0": 0.30469314455986024, "rewards/batch_coverage_1": 0.30469314455986024, "rewards/batch_coverage_5": 0.3353682577610016, "rewards/brier_reward": 0.776526153087616, "rewards/confidence_uniqueness_reward": 0.9517050504684448, "rewards/format_reward": 0.99765625, "rewards/frontier_aurc_reward": -0.002814192697405815, "rewards/frontier_coverage_10": 0.07425818145275116, "rewards/frontier_coverage_15": 0.07425158321857453, "rewards/frontier_coverage_20": 0.07402418851852417, "rewards/frontier_coverage_25": 0.05917001962661743, "rewards/frontier_ece_reward": 0.005444050068035722, "rewards/frontier_entropy_batch_reward": -0.1761537969112396, "signal/accuracy_reward/centered_abs_mean": 0.11986083984375, "signal/accuracy_reward/group_std_mean": 0.16211077272892, "signal/accuracy_reward/group_zero_std_frac": 0.521875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.059930419921875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.059930419921875, "signal/advantage_abs_mean": 0.07677890509366989, "signal/advantage_pre_scale_abs_mean": 0.07677890509366989, "signal/advantage_pre_scale_std": 0.12326882630586625, "signal/advantage_std": 0.12326882630586625, "signal/batch_coverage_0/centered_abs_mean": 0.17479489743709564, "signal/batch_coverage_0/group_std_mean": 0.22143566012382507, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002184936311095953, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002184936311095953, "signal/batch_coverage_1/centered_abs_mean": 0.17479489743709564, "signal/batch_coverage_1/group_std_mean": 0.22143566012382507, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002184936311095953, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002184936311095953, "signal/batch_coverage_5/centered_abs_mean": 0.18566111028194426, "signal/batch_coverage_5/group_std_mean": 0.23521003425121306, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0023207639344036577, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0023207639344036577, "signal/brier_reward/centered_abs_mean": 0.14487112760543824, "signal/brier_reward/group_std_mean": 0.18507781624794006, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.014487113244831562, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.014487113244831562, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.015338851325213908, "signal/confidence_uniqueness_reward/group_std_mean": 0.02584679201245308, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0015338851138949395, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0015338851138949395, "signal/format_reward/centered_abs_mean": 0.0045166015625, "signal/format_reward/group_std_mean": 0.012585635203868151, "signal/format_reward/group_zero_std_frac": 0.93125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00225830078125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00225830078125, "signal/frontier_aurc_reward/centered_abs_mean": 0.0028497647028416393, "signal/frontier_aurc_reward/group_std_mean": 0.0046899310313165184, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.5622059658635405e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.5622059658635405e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.19139614403247834, "signal/frontier_coverage_10/group_std_mean": 0.2440494030714035, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0023924518376588822, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0023924518376588822, "signal/frontier_coverage_15/centered_abs_mean": 0.19127858579158782, "signal/frontier_coverage_15/group_std_mean": 0.2439014732837677, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002390982350334525, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002390982350334525, "signal/frontier_coverage_20/centered_abs_mean": 0.19005476236343383, "signal/frontier_coverage_20/group_std_mean": 0.2423454135656357, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002375684538856149, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002375684538856149, "signal/frontier_coverage_25/centered_abs_mean": 0.1342945471405983, "signal/frontier_coverage_25/group_std_mean": 0.1726694792509079, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0016786818858236074, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0016786818858236074, "signal/frontier_ece_reward/centered_abs_mean": 0.01035686582326889, "signal/frontier_ece_reward/group_std_mean": 0.013562301360070705, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.001035686582326889, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.001035686582326889, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.25433451533317564, "signal/frontier_entropy_batch_reward/group_std_mean": 0.32907217741012573, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.025433451309800147, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025433451309800147, "step": 170 }, { "calibration/aurc": 0.2782831999501868, "calibration/batch_distribution_entropy": 0.9857212500452277, "calibration/buffer_distribution_entropy": 0.9983133988619481, "calibration/confidence_entropy": 0.4963047268790629, "calibration/coverage@0%": 0.05841757379321592, "calibration/coverage@1%": 0.09488816202851004, "calibration/coverage@10%": 0.21953571968170832, "calibration/coverage@15%": 0.3016993334148728, "calibration/coverage@20%": 0.3607402764187867, "calibration/coverage@25%": 0.4166768590998043, "calibration/coverage@30%": 0.49097205234833663, "calibration/coverage@5%": 0.16273129928341198, "calibration/ece": 0.13001094236149396, "calibration/mean_confidence": 0.4828596168261151, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 917.0, "completions/max_terminated_length": 917.0, "completions/mean_length": 229.246484375, "completions/mean_terminated_length": 229.69794921875, "completions/min_length": 0.0, "completions/min_terminated_length": 112.0, "epoch": 0.56, "grad_norm": 0.0008990481146611273, "learning_rate": 1e-06, "loss": -0.0011, "num_tokens": 593518448.0, "reward": 0.9355271816253662, "reward_std": 0.0944584622979164, "rewards/accuracy_reward": 0.5240234375, "rewards/batch_coverage_0": 0.34526376724243163, "rewards/batch_coverage_1": 0.34526376724243163, "rewards/batch_coverage_5": 0.37272083163261416, "rewards/brier_reward": 0.7936571717262269, "rewards/confidence_uniqueness_reward": 0.9517397046089172, "rewards/format_reward": 0.998046875, "rewards/frontier_aurc_reward": -0.002639968553557992, "rewards/frontier_coverage_10": 0.11518731117248535, "rewards/frontier_coverage_15": 0.11509384959936142, "rewards/frontier_coverage_20": 0.11313305795192719, "rewards/frontier_coverage_25": 0.08338004499673843, "rewards/frontier_ece_reward": 0.0063049433752894405, "rewards/frontier_entropy_batch_reward": -0.1927069067955017, "signal/accuracy_reward/centered_abs_mean": 0.096240234375, "signal/accuracy_reward/group_std_mean": 0.13344694525003434, "signal/accuracy_reward/group_zero_std_frac": 0.59375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0481201171875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0481201171875, "signal/advantage_abs_mean": 0.06985260993242264, "signal/advantage_pre_scale_abs_mean": 0.06985260993242264, "signal/advantage_pre_scale_std": 0.11402361243963241, "signal/advantage_std": 0.11402361243963241, "signal/batch_coverage_0/centered_abs_mean": 0.16477873027324677, "signal/batch_coverage_0/group_std_mean": 0.20702188909053804, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0020597341470420362, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0020597341470420362, "signal/batch_coverage_1/centered_abs_mean": 0.16477873027324677, "signal/batch_coverage_1/group_std_mean": 0.20702188909053804, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0020597341470420362, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0020597341470420362, "signal/batch_coverage_5/centered_abs_mean": 0.1740093231201172, "signal/batch_coverage_5/group_std_mean": 0.21892915964126586, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0021751166321337225, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0021751166321337225, "signal/brier_reward/centered_abs_mean": 0.13172113746404648, "signal/brier_reward/group_std_mean": 0.16949404776096344, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01317211389541626, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01317211389541626, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.014627835154533387, "signal/confidence_uniqueness_reward/group_std_mean": 0.023868880793452264, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.001462783571332693, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001462783571332693, "signal/format_reward/centered_abs_mean": 0.003759765625, "signal/format_reward/group_std_mean": 0.010375926457345485, "signal/format_reward/group_zero_std_frac": 0.94375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0018798828125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0018798828125, "signal/frontier_aurc_reward/centered_abs_mean": 0.00260388795286417, "signal/frontier_aurc_reward/group_std_mean": 0.004329502303153277, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.254860057495534e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.254860057495534e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1756815105676651, "signal/frontier_coverage_10/group_std_mean": 0.22533594071865082, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0021960189566016196, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0021960189566016196, "signal/frontier_coverage_15/centered_abs_mean": 0.17555386424064637, "signal/frontier_coverage_15/group_std_mean": 0.2251744419336319, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0021944233682006597, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0021944233682006597, "signal/frontier_coverage_20/centered_abs_mean": 0.1729944258928299, "signal/frontier_coverage_20/group_std_mean": 0.22197224497795104, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002162430388852954, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002162430388852954, "signal/frontier_coverage_25/centered_abs_mean": 0.11791795790195465, "signal/frontier_coverage_25/group_std_mean": 0.1522868365049362, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0014739745063707232, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0014739745063707232, "signal/frontier_ece_reward/centered_abs_mean": 0.009257416240870953, "signal/frontier_ece_reward/group_std_mean": 0.012104857712984085, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0009257416473701596, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0009257416473701596, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2655459940433502, "signal/frontier_entropy_batch_reward/group_std_mean": 0.340226149559021, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026554600149393082, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026554600149393082, "step": 175 }, { "calibration/aurc": 0.30473855504090847, "calibration/batch_distribution_entropy": 0.983214428348024, "calibration/buffer_distribution_entropy": 0.9985082234932466, "calibration/confidence_entropy": 0.48482509428397236, "calibration/coverage@0%": 0.015254280222362918, "calibration/coverage@1%": 0.015254280222362918, "calibration/coverage@10%": 0.131897663174859, "calibration/coverage@15%": 0.21946114404665976, "calibration/coverage@20%": 0.2941362932159165, "calibration/coverage@25%": 0.3864018612582019, "calibration/coverage@30%": 0.4606986852960363, "calibration/coverage@5%": 0.09160286203522505, "calibration/ece": 0.0937741018744215, "calibration/mean_confidence": 0.49879996019084827, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 1097.6, "completions/max_terminated_length": 1097.6, "completions/mean_length": 223.659765625, "completions/mean_terminated_length": 224.10057067871094, "completions/min_length": 0.0, "completions/min_terminated_length": 107.0, "epoch": 0.576, "grad_norm": 0.000906144268810749, "learning_rate": 1e-06, "loss": -0.0012, "num_tokens": 610995348.0, "reward": 0.9302504658699036, "reward_std": 0.09146282821893692, "rewards/accuracy_reward": 0.51396484375, "rewards/batch_coverage_0": 0.34967092871665956, "rewards/batch_coverage_1": 0.34967092871665956, "rewards/batch_coverage_5": 0.378152197599411, "rewards/brier_reward": 0.7866646647453308, "rewards/confidence_uniqueness_reward": 0.9515843629837036, "rewards/format_reward": 0.998046875, "rewards/frontier_aurc_reward": -0.0033160723280161617, "rewards/frontier_coverage_10": 0.11806410551071167, "rewards/frontier_coverage_15": 0.1180365115404129, "rewards/frontier_coverage_20": 0.11751026064157485, "rewards/frontier_coverage_25": 0.08601402789354325, "rewards/frontier_ece_reward": 0.005975698307156563, "rewards/frontier_entropy_batch_reward": -0.19100389182567595, "signal/accuracy_reward/centered_abs_mean": 0.090216064453125, "signal/accuracy_reward/group_std_mean": 0.124189093708992, "signal/accuracy_reward/group_zero_std_frac": 0.625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0451080322265625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0451080322265625, "signal/advantage_abs_mean": 0.06790656447410584, "signal/advantage_pre_scale_abs_mean": 0.06790656447410584, "signal/advantage_pre_scale_std": 0.11092653423547745, "signal/advantage_std": 0.11092653423547745, "signal/batch_coverage_0/centered_abs_mean": 0.16153261661529542, "signal/batch_coverage_0/group_std_mean": 0.20566450357437133, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002019157726317644, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002019157726317644, "signal/batch_coverage_1/centered_abs_mean": 0.16153261661529542, "signal/batch_coverage_1/group_std_mean": 0.20566450357437133, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002019157726317644, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002019157726317644, "signal/batch_coverage_5/centered_abs_mean": 0.1695544809103012, "signal/batch_coverage_5/group_std_mean": 0.2166013538837433, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002119430992752314, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002119430992752314, "signal/brier_reward/centered_abs_mean": 0.13086027354002, "signal/brier_reward/group_std_mean": 0.17123542428016664, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.013086027652025222, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.013086027652025222, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.014868937805294991, "signal/confidence_uniqueness_reward/group_std_mean": 0.023424912616610526, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0014868937898427248, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014868937898427248, "signal/format_reward/centered_abs_mean": 0.0037353515625, "signal/format_reward/group_std_mean": 0.009703309601172804, "signal/format_reward/group_zero_std_frac": 0.95, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00186767578125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00186767578125, "signal/frontier_aurc_reward/centered_abs_mean": 0.003207807056605816, "signal/frontier_aurc_reward/group_std_mean": 0.005212449654936791, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 4.009758813481312e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 4.009758813481312e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1703052341938019, "signal/frontier_coverage_10/group_std_mean": 0.2200066089630127, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0021288154646754267, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0021288154646754267, "signal/frontier_coverage_15/centered_abs_mean": 0.17020165026187897, "signal/frontier_coverage_15/group_std_mean": 0.21987622380256652, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0021275205770507457, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0021275205770507457, "signal/frontier_coverage_20/centered_abs_mean": 0.16779713034629823, "signal/frontier_coverage_20/group_std_mean": 0.21683943569660186, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002097464120015502, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002097464120015502, "signal/frontier_coverage_25/centered_abs_mean": 0.11331845223903655, "signal/frontier_coverage_25/group_std_mean": 0.14725278913974763, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.001416480727493763, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.001416480727493763, "signal/frontier_ece_reward/centered_abs_mean": 0.009245606325566768, "signal/frontier_ece_reward/group_std_mean": 0.012007570825517178, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0009245606837794184, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0009245606837794184, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26394750475883483, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3392052948474884, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02639475092291832, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02639475092291832, "step": 180 }, { "calibration/aurc": 0.2972691411574999, "calibration/batch_distribution_entropy": 0.9841643978420889, "calibration/buffer_distribution_entropy": 0.9985284494056017, "calibration/confidence_entropy": 0.4912927121382157, "calibration/coverage@0%": 0.029368180398824762, "calibration/coverage@1%": 0.029368180398824762, "calibration/coverage@10%": 0.14094621014262745, "calibration/coverage@15%": 0.29977546088055346, "calibration/coverage@20%": 0.44268938948958086, "calibration/coverage@25%": 0.5401266888518939, "calibration/coverage@30%": 0.6141165295721391, "calibration/coverage@5%": 0.08455892023561315, "calibration/ece": 0.1257793231552548, "calibration/mean_confidence": 0.49902772450595806, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00166015625, "completions/max_length": 1040.4, "completions/max_terminated_length": 1040.4, "completions/mean_length": 220.47158203125, "completions/mean_terminated_length": 220.83885192871094, "completions/min_length": 0.0, "completions/min_terminated_length": 109.0, "epoch": 0.592, "grad_norm": 0.0009568691602908075, "learning_rate": 1e-06, "loss": -0.0017, "num_tokens": 628420689.0, "reward": 0.9329164266586304, "reward_std": 0.09069040715694428, "rewards/accuracy_reward": 0.5224609375, "rewards/batch_coverage_0": 0.364839643239975, "rewards/batch_coverage_1": 0.364839643239975, "rewards/batch_coverage_5": 0.38795469999313353, "rewards/brier_reward": 0.7838621497154236, "rewards/confidence_uniqueness_reward": 0.9513613104820251, "rewards/format_reward": 0.99833984375, "rewards/frontier_aurc_reward": -0.0031355343759059905, "rewards/frontier_coverage_10": 0.11191404610872269, "rewards/frontier_coverage_15": 0.11182275265455247, "rewards/frontier_coverage_20": 0.11066430807113647, "rewards/frontier_coverage_25": 0.0756689727306366, "rewards/frontier_ece_reward": 0.005897596012800932, "rewards/frontier_entropy_batch_reward": -0.20653201639652252, "signal/accuracy_reward/centered_abs_mean": 0.0947509765625, "signal/accuracy_reward/group_std_mean": 0.12710948586463927, "signal/accuracy_reward/group_zero_std_frac": 0.61875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.04737548828125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.04737548828125, "signal/advantage_abs_mean": 0.06895003989338874, "signal/advantage_pre_scale_abs_mean": 0.06895003989338874, "signal/advantage_pre_scale_std": 0.11086848676204682, "signal/advantage_std": 0.11086848676204682, "signal/batch_coverage_0/centered_abs_mean": 0.1667444735765457, "signal/batch_coverage_0/group_std_mean": 0.21070497035980223, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0020843059290200474, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0020843059290200474, "signal/batch_coverage_1/centered_abs_mean": 0.1667444735765457, "signal/batch_coverage_1/group_std_mean": 0.21070497035980223, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0020843059290200474, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0020843059290200474, "signal/batch_coverage_5/centered_abs_mean": 0.17382386028766633, "signal/batch_coverage_5/group_std_mean": 0.2197079986333847, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002172798290848732, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002172798290848732, "signal/brier_reward/centered_abs_mean": 0.12995279878377913, "signal/brier_reward/group_std_mean": 0.16773780882358552, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01299527995288372, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01299527995288372, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.015028782933950425, "signal/confidence_uniqueness_reward/group_std_mean": 0.022947072982788086, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0015028782887384295, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0015028782887384295, "signal/format_reward/centered_abs_mean": 0.003179931640625, "signal/format_reward/group_std_mean": 0.008382336515933276, "signal/format_reward/group_zero_std_frac": 0.95625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0015899658203125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0015899658203125, "signal/frontier_aurc_reward/centered_abs_mean": 0.003012676211073995, "signal/frontier_aurc_reward/group_std_mean": 0.004799708258360624, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.7658453948097304e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.7658453948097304e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.17378068566322327, "signal/frontier_coverage_10/group_std_mean": 0.22225480973720552, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0021722585894167422, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0021722585894167422, "signal/frontier_coverage_15/centered_abs_mean": 0.17353184819221495, "signal/frontier_coverage_15/group_std_mean": 0.22194352149963378, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0021691481582820416, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0021691481582820416, "signal/frontier_coverage_20/centered_abs_mean": 0.1701044738292694, "signal/frontier_coverage_20/group_std_mean": 0.2176322251558304, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0021263059228658677, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0021263059228658677, "signal/frontier_coverage_25/centered_abs_mean": 0.1100577712059021, "signal/frontier_coverage_25/group_std_mean": 0.14178918600082396, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.001375722186639905, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.001375722186639905, "signal/frontier_ece_reward/centered_abs_mean": 0.008959782123565675, "signal/frontier_ece_reward/group_std_mean": 0.011550957150757313, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0008959782309830189, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0008959782309830189, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2795942187309265, "signal/frontier_entropy_batch_reward/group_std_mean": 0.35300915241241454, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.027959422022104264, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027959422022104264, "step": 185 }, { "calibration/aurc": 0.24945990415868188, "calibration/batch_distribution_entropy": 0.9687325263864215, "calibration/buffer_distribution_entropy": 0.998500267559729, "calibration/confidence_entropy": 0.47891183703288415, "calibration/coverage@0%": 0.02578660102739726, "calibration/coverage@1%": 0.02578660102739726, "calibration/coverage@10%": 0.1931086411448141, "calibration/coverage@15%": 0.28423281555772995, "calibration/coverage@20%": 0.4874036815068493, "calibration/coverage@25%": 0.5874457252935421, "calibration/coverage@30%": 0.6878952115949118, "calibration/coverage@5%": 0.13134555406066534, "calibration/ece": 0.10097432861399339, "calibration/mean_confidence": 0.47559856022697106, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 924.0, "completions/max_terminated_length": 924.0, "completions/mean_length": 218.85302734375, "completions/mean_terminated_length": 219.4324981689453, "completions/min_length": 0.0, "completions/min_terminated_length": 108.8, "epoch": 0.608, "grad_norm": 0.0007343413890339434, "learning_rate": 1e-06, "loss": -0.0021, "num_tokens": 645661232.0, "reward": 0.9350929617881775, "reward_std": 0.09022950083017349, "rewards/accuracy_reward": 0.52041015625, "rewards/batch_coverage_0": 0.3865961074829102, "rewards/batch_coverage_1": 0.3865961074829102, "rewards/batch_coverage_5": 0.40012803077697756, "rewards/brier_reward": 0.8028019905090332, "rewards/confidence_uniqueness_reward": 0.9500757455825806, "rewards/format_reward": 0.997265625, "rewards/frontier_aurc_reward": -0.0025759967509657146, "rewards/frontier_coverage_10": 0.13697629123926164, "rewards/frontier_coverage_15": 0.13676386922597886, "rewards/frontier_coverage_20": 0.13333353400230408, "rewards/frontier_coverage_25": 0.08915815353393555, "rewards/frontier_ece_reward": 0.006305753719061613, "rewards/frontier_entropy_batch_reward": -0.20500501990318298, "signal/accuracy_reward/centered_abs_mean": 0.094952392578125, "signal/accuracy_reward/group_std_mean": 0.1252264618873596, "signal/accuracy_reward/group_zero_std_frac": 0.64375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0474761962890625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0474761962890625, "signal/advantage_abs_mean": 0.0668793685734272, "signal/advantage_pre_scale_abs_mean": 0.0668793685734272, "signal/advantage_pre_scale_std": 0.11068283170461654, "signal/advantage_std": 0.11068283170461654, "signal/batch_coverage_0/centered_abs_mean": 0.1667961061000824, "signal/batch_coverage_0/group_std_mean": 0.20975518226623535, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002084951288998127, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002084951288998127, "signal/batch_coverage_1/centered_abs_mean": 0.1667961061000824, "signal/batch_coverage_1/group_std_mean": 0.20975518226623535, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002084951288998127, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002084951288998127, "signal/batch_coverage_5/centered_abs_mean": 0.17114523947238922, "signal/batch_coverage_5/group_std_mean": 0.21545170545578002, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0021393154747784138, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0021393154747784138, "signal/brier_reward/centered_abs_mean": 0.12462074160575867, "signal/brier_reward/group_std_mean": 0.16169702708721162, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01246207468211651, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01246207468211651, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.016549999825656414, "signal/confidence_uniqueness_reward/group_std_mean": 0.027828745916485785, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0016549999825656414, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0016549999825656414, "signal/format_reward/centered_abs_mean": 0.0052490234375, "signal/format_reward/group_std_mean": 0.014122726581990718, "signal/format_reward/group_zero_std_frac": 0.925, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00262451171875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00262451171875, "signal/frontier_aurc_reward/centered_abs_mean": 0.002489195205271244, "signal/frontier_aurc_reward/group_std_mean": 0.004241161420941353, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.1114940793486315e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.1114940793486315e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1814405381679535, "signal/frontier_coverage_10/group_std_mean": 0.228531876206398, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.002268006792291999, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.002268006792291999, "signal/frontier_coverage_15/centered_abs_mean": 0.18110425770282745, "signal/frontier_coverage_15/group_std_mean": 0.22810145616531372, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0022638033144176005, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0022638033144176005, "signal/frontier_coverage_20/centered_abs_mean": 0.17451958656311034, "signal/frontier_coverage_20/group_std_mean": 0.21984116435050965, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002181494818069041, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002181494818069041, "signal/frontier_coverage_25/centered_abs_mean": 0.1069264367222786, "signal/frontier_coverage_25/group_std_mean": 0.1353532314300537, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0013365804916247725, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0013365804916247725, "signal/frontier_ece_reward/centered_abs_mean": 0.008543512038886546, "signal/frontier_ece_reward/group_std_mean": 0.010987267456948757, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0008543511619791388, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0008543511619791388, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2726089894771576, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34626291394233705, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.027260898798704147, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027260898798704147, "step": 190 }, { "calibration/aurc": 0.25792269774253024, "calibration/batch_distribution_entropy": 0.9889070394848243, "calibration/buffer_distribution_entropy": 0.9984948809542626, "calibration/confidence_entropy": 0.5006283280367128, "calibration/coverage@0%": 0.006654438850687623, "calibration/coverage@1%": 0.006654438850687623, "calibration/coverage@10%": 0.14122513506876228, "calibration/coverage@15%": 0.27191490667976426, "calibration/coverage@20%": 0.3962525250154089, "calibration/coverage@25%": 0.5214066081368697, "calibration/coverage@30%": 0.6310585743383798, "calibration/coverage@5%": 0.02543667116895874, "calibration/ece": 0.09748939364720823, "calibration/mean_confidence": 0.511776033266302, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00234375, "completions/max_length": 816.6, "completions/max_terminated_length": 816.6, "completions/mean_length": 215.85966796875, "completions/mean_terminated_length": 216.3705810546875, "completions/min_length": 0.0, "completions/min_terminated_length": 107.4, "epoch": 0.624, "grad_norm": 0.0009215656318701804, "learning_rate": 1e-06, "loss": -0.0018, "num_tokens": 663215539.0, "reward": 0.939211106300354, "reward_std": 0.09588029682636261, "rewards/accuracy_reward": 0.52861328125, "rewards/batch_coverage_0": 0.35543850660324094, "rewards/batch_coverage_1": 0.35543850660324094, "rewards/batch_coverage_5": 0.369016969203949, "rewards/brier_reward": 0.7932912945747376, "rewards/confidence_uniqueness_reward": 0.951915180683136, "rewards/format_reward": 0.99755859375, "rewards/frontier_aurc_reward": -0.002853636536747217, "rewards/frontier_coverage_10": 0.11610840111970902, "rewards/frontier_coverage_15": 0.11587588340044022, "rewards/frontier_coverage_20": 0.10823035538196564, "rewards/frontier_coverage_25": 0.07128367722034454, "rewards/frontier_ece_reward": 0.005040967091917992, "rewards/frontier_entropy_batch_reward": -0.1750631809234619, "signal/accuracy_reward/centered_abs_mean": 0.104876708984375, "signal/accuracy_reward/group_std_mean": 0.1374049127101898, "signal/accuracy_reward/group_zero_std_frac": 0.60625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0524383544921875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0524383544921875, "signal/advantage_abs_mean": 0.07228044420480728, "signal/advantage_pre_scale_abs_mean": 0.07228044420480728, "signal/advantage_pre_scale_std": 0.11648347973823547, "signal/advantage_std": 0.11648347973823547, "signal/batch_coverage_0/centered_abs_mean": 0.16825419962406157, "signal/batch_coverage_0/group_std_mean": 0.21355200409889222, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.002103177597746253, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.002103177597746253, "signal/batch_coverage_1/centered_abs_mean": 0.16825419962406157, "signal/batch_coverage_1/group_std_mean": 0.21355200409889222, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.002103177597746253, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.002103177597746253, "signal/batch_coverage_5/centered_abs_mean": 0.17341194152832032, "signal/batch_coverage_5/group_std_mean": 0.22024895250797272, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0021676492411643266, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0021676492411643266, "signal/brier_reward/centered_abs_mean": 0.1269794300198555, "signal/brier_reward/group_std_mean": 0.16681278944015504, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.012697943300008774, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.012697943300008774, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.015193746611475944, "signal/confidence_uniqueness_reward/group_std_mean": 0.025495481491088868, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.001519374712370336, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001519374712370336, "signal/format_reward/centered_abs_mean": 0.004681396484375, "signal/format_reward/group_std_mean": 0.01246544560417533, "signal/format_reward/group_zero_std_frac": 0.934375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0023406982421875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0023406982421875, "signal/frontier_aurc_reward/centered_abs_mean": 0.0026586991967633366, "signal/frontier_aurc_reward/group_std_mean": 0.00445527695119381, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.3233740032301286e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.3233740032301286e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.17277007400989533, "signal/frontier_coverage_10/group_std_mean": 0.22225799560546874, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0021596259437501432, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0021596259437501432, "signal/frontier_coverage_15/centered_abs_mean": 0.1722703605890274, "signal/frontier_coverage_15/group_std_mean": 0.22162049114704133, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0021533796098083258, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0021533796098083258, "signal/frontier_coverage_20/centered_abs_mean": 0.15647023022174836, "signal/frontier_coverage_20/group_std_mean": 0.20173780024051666, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.001955877966247499, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.001955877966247499, "signal/frontier_coverage_25/centered_abs_mean": 0.09244705587625504, "signal/frontier_coverage_25/group_std_mean": 0.12020283043384553, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0011555882403627038, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0011555882403627038, "signal/frontier_ece_reward/centered_abs_mean": 0.007718043029308319, "signal/frontier_ece_reward/group_std_mean": 0.010004553012549877, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0007718043169006705, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0007718043169006705, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2552692711353302, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3307402491569519, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.025526927784085274, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025526927784085274, "step": 195 }, { "calibration/aurc": 0.2702763095676272, "calibration/batch_distribution_entropy": 0.9778551292507259, "calibration/buffer_distribution_entropy": 0.9985339737071609, "calibration/confidence_entropy": 0.49651567230638005, "calibration/coverage@0%": 0.05596419289465745, "calibration/coverage@1%": 0.09079785238585118, "calibration/coverage@10%": 0.2846709260159882, "calibration/coverage@15%": 0.3327476138829158, "calibration/coverage@20%": 0.36832047958741676, "calibration/coverage@25%": 0.42420890302185904, "calibration/coverage@30%": 0.5790313117588932, "calibration/coverage@5%": 0.22057631985160464, "calibration/ece": 0.158789997284181, "calibration/mean_confidence": 0.5409215097972188, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 782.0, "completions/max_terminated_length": 782.0, "completions/mean_length": 214.81396484375, "completions/mean_terminated_length": 215.23322448730468, "completions/min_length": 0.0, "completions/min_terminated_length": 108.8, "epoch": 0.64, "grad_norm": 0.0008913954952731729, "learning_rate": 1e-06, "loss": -0.0011, "num_tokens": 680757922.0, "reward": 0.9550224304199219, "reward_std": 0.08659601956605911, "rewards/accuracy_reward": 0.56474609375, "rewards/batch_coverage_0": 0.3767550468444824, "rewards/batch_coverage_1": 0.3767550468444824, "rewards/batch_coverage_5": 0.3962052702903748, "rewards/brier_reward": 0.8009827017784119, "rewards/confidence_uniqueness_reward": 0.9508432745933533, "rewards/format_reward": 0.99794921875, "rewards/frontier_aurc_reward": -0.0025522151263430715, "rewards/frontier_coverage_10": 0.09268663451075554, "rewards/frontier_coverage_15": 0.09238817989826202, "rewards/frontier_coverage_20": 0.08323745727539063, "rewards/frontier_coverage_25": 0.0569609597325325, "rewards/frontier_ece_reward": 0.005287410318851471, "rewards/frontier_entropy_batch_reward": -0.20442000031471252, "signal/accuracy_reward/centered_abs_mean": 0.081475830078125, "signal/accuracy_reward/group_std_mean": 0.11334783136844635, "signal/accuracy_reward/group_zero_std_frac": 0.653125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0407379150390625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0407379150390625, "signal/advantage_abs_mean": 0.06465212777256965, "signal/advantage_pre_scale_abs_mean": 0.06465212777256965, "signal/advantage_pre_scale_std": 0.1069864958524704, "signal/advantage_std": 0.1069864958524704, "signal/batch_coverage_0/centered_abs_mean": 0.15731403231620789, "signal/batch_coverage_0/group_std_mean": 0.19781863391399385, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0019664254738017916, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0019664254738017916, "signal/batch_coverage_1/centered_abs_mean": 0.15731403231620789, "signal/batch_coverage_1/group_std_mean": 0.19781863391399385, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0019664254738017916, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0019664254738017916, "signal/batch_coverage_5/centered_abs_mean": 0.16203798949718476, "signal/batch_coverage_5/group_std_mean": 0.20434709787368774, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0020254747942090034, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0020254747942090034, "signal/brier_reward/centered_abs_mean": 0.11717531681060792, "signal/brier_reward/group_std_mean": 0.15107038319110871, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011717531643807889, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011717531643807889, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.015049685537815095, "signal/confidence_uniqueness_reward/group_std_mean": 0.022032597661018373, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0015049685724079608, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0015049685724079608, "signal/format_reward/centered_abs_mean": 0.003717041015625, "signal/format_reward/group_std_mean": 0.00804602806456387, "signal/format_reward/group_zero_std_frac": 0.9625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0018585205078125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0018585205078125, "signal/frontier_aurc_reward/centered_abs_mean": 0.002512537548318505, "signal/frontier_aurc_reward/group_std_mean": 0.004136429913341999, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.14067212457303e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.14067212457303e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1491511881351471, "signal/frontier_coverage_10/group_std_mean": 0.19108016192913055, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0018643899355083704, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0018643899355083704, "signal/frontier_coverage_15/centered_abs_mean": 0.14845846295356752, "signal/frontier_coverage_15/group_std_mean": 0.19019553065299988, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0018557308241724968, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0018557308241724968, "signal/frontier_coverage_20/centered_abs_mean": 0.12964392602443695, "signal/frontier_coverage_20/group_std_mean": 0.16609353721141815, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0016205491730943323, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0016205491730943323, "signal/frontier_coverage_25/centered_abs_mean": 0.07533989399671555, "signal/frontier_coverage_25/group_std_mean": 0.0967397689819336, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0009417486959137023, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0009417486959137023, "signal/frontier_ece_reward/centered_abs_mean": 0.0075281173922121525, "signal/frontier_ece_reward/group_std_mean": 0.009740599431097507, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0007528117392212152, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0007528117392212152, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2746058821678162, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34960067868232725, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02746058814227581, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02746058814227581, "step": 200 }, { "epoch": 0.64, "eval_calibration/aurc": 0.4948504337971422, "eval_calibration/batch_distribution_entropy": 0.9408090328185059, "eval_calibration/buffer_distribution_entropy": 0.9985579358282551, "eval_calibration/confidence_entropy": 0.4876156693507687, "eval_calibration/coverage@0%": 0.0625, "eval_calibration/coverage@1%": 0.0625, "eval_calibration/coverage@10%": 0.0625, "eval_calibration/coverage@15%": 0.0859375, "eval_calibration/coverage@20%": 0.1171875, "eval_calibration/coverage@25%": 0.1328125, "eval_calibration/coverage@30%": 0.15625, "eval_calibration/coverage@5%": 0.0625, "eval_calibration/ece": 0.21838366368587264, "eval_calibration/mean_confidence": 0.4980459764374663, "eval_completions/clipped_ratio": 0.0, "eval_completions/max_length": 461.25, "eval_completions/max_terminated_length": 461.25, "eval_completions/mean_length": 215.08809280395508, "eval_completions/mean_terminated_length": 215.08809280395508, "eval_completions/min_length": 125.5, "eval_completions/min_terminated_length": 125.5, "eval_loss": 0.0, "eval_num_tokens": 680757922.0, "eval_reward": 0.7935310304164886, "eval_reward_std": 0.23132389411330223, "eval_rewards/accuracy_reward": 0.421875, "eval_rewards/batch_coverage_0": 0.15543591976165771, "eval_rewards/batch_coverage_1": 0.15543591976165771, "eval_rewards/batch_coverage_5": 0.15543591976165771, "eval_rewards/brier_reward": 0.7904356867074966, "eval_rewards/confidence_uniqueness_reward": 0.8994140625, "eval_rewards/format_reward": 1.0, "eval_rewards/frontier_aurc_reward": -0.003559257194865495, "eval_rewards/frontier_coverage_10": 0.17605121806263924, "eval_rewards/frontier_coverage_15": 0.17528558522462845, "eval_rewards/frontier_coverage_20": 0.15015807002782822, "eval_rewards/frontier_coverage_25": 0.08468825556337833, "eval_rewards/frontier_ece_reward": 0.004968958091922104, "eval_rewards/frontier_entropy_batch_reward": -1.0, "eval_runtime": 22.5081, "eval_samples_per_second": 22.214, "eval_signal/accuracy_reward/centered_abs_mean": 0.46923828125, "eval_signal/accuracy_reward/group_std_mean": 0.4916032701730728, "eval_signal/accuracy_reward/group_zero_std_frac": 0.0, "eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.234619140625, "eval_signal/accuracy_reward/weight": 0.5, "eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.234619140625, "eval_signal/advantage_abs_mean": 0.21653173863887787, "eval_signal/advantage_pre_scale_abs_mean": 0.21653173863887787, "eval_signal/advantage_pre_scale_std": 0.22882291302084923, "eval_signal/advantage_std": 0.22882291302084923, "eval_signal/batch_coverage_0/centered_abs_mean": 0.2712183967232704, "eval_signal/batch_coverage_0/group_std_mean": 0.3371654972434044, "eval_signal/batch_coverage_0/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.003390230005607009, "eval_signal/batch_coverage_0/weight": 0.012500000186264515, "eval_signal/batch_coverage_0/weighted_centered_abs_mean": 0.003390230005607009, "eval_signal/batch_coverage_1/centered_abs_mean": 0.2712183967232704, "eval_signal/batch_coverage_1/group_std_mean": 0.3371654972434044, "eval_signal/batch_coverage_1/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.003390230005607009, "eval_signal/batch_coverage_1/weight": 0.012500000186264515, "eval_signal/batch_coverage_1/weighted_centered_abs_mean": 0.003390230005607009, "eval_signal/batch_coverage_5/centered_abs_mean": 0.2712183967232704, "eval_signal/batch_coverage_5/group_std_mean": 0.3371654972434044, "eval_signal/batch_coverage_5/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.003390230005607009, "eval_signal/batch_coverage_5/weight": 0.012500000186264515, "eval_signal/batch_coverage_5/weighted_centered_abs_mean": 0.003390230005607009, "eval_signal/brier_reward/centered_abs_mean": 0.19362923502922058, "eval_signal/brier_reward/group_std_mean": 0.2464425042271614, "eval_signal/brier_reward/group_zero_std_frac": 0.0, "eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01936292415484786, "eval_signal/brier_reward/weight": 0.10000000149011612, "eval_signal/brier_reward/weighted_centered_abs_mean": 0.01936292415484786, "eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.0402374267578125, "eval_signal/confidence_uniqueness_reward/group_std_mean": 0.047222621738910675, "eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.004023742862045765, "eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004023742862045765, "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_aurc_reward/centered_abs_mean": 0.004474427842069417, "eval_signal/frontier_aurc_reward/group_std_mean": 0.0076356109930202365, "eval_signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 5.5930348025867715e-05, "eval_signal/frontier_aurc_reward/weight": 0.012500000186264515, "eval_signal/frontier_aurc_reward/weighted_centered_abs_mean": 5.5930348025867715e-05, "eval_signal/frontier_coverage_10/centered_abs_mean": 0.3317065164446831, "eval_signal/frontier_coverage_10/group_std_mean": 0.4069000855088234, "eval_signal/frontier_coverage_10/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.00414633157197386, "eval_signal/frontier_coverage_10/weight": 0.012500000186264515, "eval_signal/frontier_coverage_10/weighted_centered_abs_mean": 0.00414633157197386, "eval_signal/frontier_coverage_15/centered_abs_mean": 0.3300493136048317, "eval_signal/frontier_coverage_15/group_std_mean": 0.4049970954656601, "eval_signal/frontier_coverage_15/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.004125616571400315, "eval_signal/frontier_coverage_15/weight": 0.012500000186264515, "eval_signal/frontier_coverage_15/weighted_centered_abs_mean": 0.004125616571400315, "eval_signal/frontier_coverage_20/centered_abs_mean": 0.2780879884958267, "eval_signal/frontier_coverage_20/group_std_mean": 0.3443576619029045, "eval_signal/frontier_coverage_20/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.003476099926047027, "eval_signal/frontier_coverage_20/weight": 0.012500000186264515, "eval_signal/frontier_coverage_20/weighted_centered_abs_mean": 0.003476099926047027, "eval_signal/frontier_coverage_25/centered_abs_mean": 0.14350834116339684, "eval_signal/frontier_coverage_25/group_std_mean": 0.18634271249175072, "eval_signal/frontier_coverage_25/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0017938542587216944, "eval_signal/frontier_coverage_25/weight": 0.012500000186264515, "eval_signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0017938542587216944, "eval_signal/frontier_ece_reward/centered_abs_mean": 0.009417804423719645, "eval_signal/frontier_ece_reward/group_std_mean": 0.012488778913393617, "eval_signal/frontier_ece_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0009417804831173271, "eval_signal/frontier_ece_reward/weight": 0.10000000149011612, "eval_signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0009417804831173271, "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": 0.178, "step": 200 }, { "calibration/aurc": 0.42494955635137527, "calibration/batch_distribution_entropy": 0.9768851260244288, "calibration/buffer_distribution_entropy": 0.9986297248084577, "calibration/confidence_entropy": 0.5317504895441105, "calibration/coverage@0%": 0.0003929273084479371, "calibration/coverage@1%": 0.0003929273084479371, "calibration/coverage@10%": 0.01213239808447937, "calibration/coverage@15%": 0.01252302308447937, "calibration/coverage@20%": 0.055147194253438114, "calibration/coverage@25%": 0.18974091355599215, "calibration/coverage@30%": 0.2944974828094303, "calibration/coverage@5%": 0.0003929273084479371, "calibration/ece": 0.10751880502496983, "calibration/mean_confidence": 0.48554695880518084, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 901.6, "completions/max_terminated_length": 901.6, "completions/mean_length": 210.169921875, "completions/mean_terminated_length": 210.35592651367188, "completions/min_length": 0.0, "completions/min_terminated_length": 100.0, "epoch": 0.656, "grad_norm": 0.0008849771111272275, "learning_rate": 1e-06, "loss": -0.001, "num_tokens": 697766606.0, "reward": 0.9191736698150634, "reward_std": 0.09234474301338196, "rewards/accuracy_reward": 0.49609375, "rewards/batch_coverage_0": 0.3357621252536774, "rewards/batch_coverage_1": 0.3357621252536774, "rewards/batch_coverage_5": 0.3529069125652313, "rewards/brier_reward": 0.7803008675575256, "rewards/confidence_uniqueness_reward": 0.9525490641593933, "rewards/format_reward": 0.99912109375, "rewards/frontier_aurc_reward": -0.0033357749227434395, "rewards/frontier_coverage_10": 0.11503563970327377, "rewards/frontier_coverage_15": 0.11460700035095214, "rewards/frontier_coverage_20": 0.0995073989033699, "rewards/frontier_coverage_25": 0.061080900579690935, "rewards/frontier_ece_reward": 0.0038642494007945062, "rewards/frontier_entropy_batch_reward": -0.19746759831905364, "signal/accuracy_reward/centered_abs_mean": 0.09580078125, "signal/accuracy_reward/group_std_mean": 0.1299314647912979, "signal/accuracy_reward/group_zero_std_frac": 0.615625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.047900390625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.047900390625, "signal/advantage_abs_mean": 0.07032258063554764, "signal/advantage_pre_scale_abs_mean": 0.07032258063554764, "signal/advantage_pre_scale_std": 0.1121922954916954, "signal/advantage_std": 0.1121922954916954, "signal/batch_coverage_0/centered_abs_mean": 0.15262713134288788, "signal/batch_coverage_0/group_std_mean": 0.1938321501016617, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.00190783916041255, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.00190783916041255, "signal/batch_coverage_1/centered_abs_mean": 0.15262713134288788, "signal/batch_coverage_1/group_std_mean": 0.1938321501016617, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.00190783916041255, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.00190783916041255, "signal/batch_coverage_5/centered_abs_mean": 0.1583049476146698, "signal/batch_coverage_5/group_std_mean": 0.2015283912420273, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.001978811901062727, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.001978811901062727, "signal/brier_reward/centered_abs_mean": 0.1268072634935379, "signal/brier_reward/group_std_mean": 0.16388813257217408, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.012680726498365403, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.012680726498365403, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013236395642161369, "signal/confidence_uniqueness_reward/group_std_mean": 0.018511437252163886, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013236395316198468, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013236395316198468, "signal/format_reward/centered_abs_mean": 0.001666259765625, "signal/format_reward/group_std_mean": 0.004240014730021358, "signal/format_reward/group_zero_std_frac": 0.978125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0008331298828125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0008331298828125, "signal/frontier_aurc_reward/centered_abs_mean": 0.002925736829638481, "signal/frontier_aurc_reward/group_std_mean": 0.004920898843556643, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.6571711825672536e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.6571711825672536e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.16363621950149537, "signal/frontier_coverage_10/group_std_mean": 0.210546612739563, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0020454526878893376, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0020454526878893376, "signal/frontier_coverage_15/centered_abs_mean": 0.16295842528343202, "signal/frontier_coverage_15/group_std_mean": 0.20967310667037964, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002036980330012739, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002036980330012739, "signal/frontier_coverage_20/centered_abs_mean": 0.13940106183290482, "signal/frontier_coverage_20/group_std_mean": 0.17956289649009705, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.001742513314820826, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.001742513314820826, "signal/frontier_coverage_25/centered_abs_mean": 0.07964335083961487, "signal/frontier_coverage_25/group_std_mean": 0.10371702909469604, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0009955418994650246, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0009955418994650246, "signal/frontier_ece_reward/centered_abs_mean": 0.0067257765680551525, "signal/frontier_ece_reward/group_std_mean": 0.008921916224062442, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0006725776824168861, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0006725776824168861, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2692209482192993, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3426955759525299, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026922095194458963, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026922095194458963, "step": 205 }, { "calibration/aurc": 0.2947798454459553, "calibration/batch_distribution_entropy": 0.980755180334483, "calibration/buffer_distribution_entropy": 0.9988806016632459, "calibration/confidence_entropy": 0.49828694150715175, "calibration/coverage@0%": 0.02816475048923679, "calibration/coverage@1%": 0.02816475048923679, "calibration/coverage@10%": 0.15565144936399217, "calibration/coverage@15%": 0.19669230063600782, "calibration/coverage@20%": 0.27609390288649704, "calibration/coverage@25%": 0.3613617600293542, "calibration/coverage@30%": 0.4399622370352251, "calibration/coverage@5%": 0.0970087756849315, "calibration/ece": 0.13286553096207648, "calibration/mean_confidence": 0.4940963419111613, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00126953125, "completions/max_length": 811.0, "completions/max_terminated_length": 811.0, "completions/mean_length": 207.251171875, "completions/mean_terminated_length": 207.51486206054688, "completions/min_length": 0.0, "completions/min_terminated_length": 98.6, "epoch": 0.672, "grad_norm": 0.0008939718245528638, "learning_rate": 1e-06, "loss": -0.0007, "num_tokens": 714802298.0, "reward": 0.931768274307251, "reward_std": 0.08992264568805694, "rewards/accuracy_reward": 0.5173828125, "rewards/batch_coverage_0": 0.3593066930770874, "rewards/batch_coverage_1": 0.3593066930770874, "rewards/batch_coverage_5": 0.37718488574028014, "rewards/brier_reward": 0.7885993242263794, "rewards/confidence_uniqueness_reward": 0.9514999389648438, "rewards/format_reward": 0.9986328125, "rewards/frontier_aurc_reward": -0.002757534384727478, "rewards/frontier_coverage_10": 0.12478849291801453, "rewards/frontier_coverage_15": 0.12443659901618957, "rewards/frontier_coverage_20": 0.10814662724733352, "rewards/frontier_coverage_25": 0.07036265432834625, "rewards/frontier_ece_reward": 0.004431106615811586, "rewards/frontier_entropy_batch_reward": -0.19702277779579164, "signal/accuracy_reward/centered_abs_mean": 0.10213623046875, "signal/accuracy_reward/group_std_mean": 0.13383425176143646, "signal/accuracy_reward/group_zero_std_frac": 0.61875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.051068115234375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.051068115234375, "signal/advantage_abs_mean": 0.0680392861366272, "signal/advantage_pre_scale_abs_mean": 0.0680392861366272, "signal/advantage_pre_scale_std": 0.10944953560829163, "signal/advantage_std": 0.10944953560829163, "signal/batch_coverage_0/centered_abs_mean": 0.16303878426551818, "signal/batch_coverage_0/group_std_mean": 0.20606594383716584, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0020379848312586544, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0020379848312586544, "signal/batch_coverage_1/centered_abs_mean": 0.16303878426551818, "signal/batch_coverage_1/group_std_mean": 0.20606594383716584, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0020379848312586544, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0020379848312586544, "signal/batch_coverage_5/centered_abs_mean": 0.16831139624118804, "signal/batch_coverage_5/group_std_mean": 0.21294715702533723, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0021038925042375922, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0021038925042375922, "signal/brier_reward/centered_abs_mean": 0.12908470034599304, "signal/brier_reward/group_std_mean": 0.16681250929832458, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.012908470630645753, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.012908470630645753, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.014231539145112038, "signal/confidence_uniqueness_reward/group_std_mean": 0.02164352424442768, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0014231539331376552, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014231539331376552, "signal/format_reward/centered_abs_mean": 0.00263671875, "signal/format_reward/group_std_mean": 0.0073976716957986355, "signal/format_reward/group_zero_std_frac": 0.959375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.001318359375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.001318359375, "signal/frontier_aurc_reward/centered_abs_mean": 0.0024111053673550485, "signal/frontier_aurc_reward/group_std_mean": 0.003920221840962768, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.0138817965053023e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.0138817965053023e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.18596895039081573, "signal/frontier_coverage_10/group_std_mean": 0.23547858297824859, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0023246118798851967, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0023246118798851967, "signal/frontier_coverage_15/centered_abs_mean": 0.18507969081401826, "signal/frontier_coverage_15/group_std_mean": 0.23437669575214387, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.002313496125862002, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.002313496125862002, "signal/frontier_coverage_20/centered_abs_mean": 0.15426360368728637, "signal/frontier_coverage_20/group_std_mean": 0.19625768065452576, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0019282949855551124, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0019282949855551124, "signal/frontier_coverage_25/centered_abs_mean": 0.09017275720834732, "signal/frontier_coverage_25/group_std_mean": 0.11547116488218308, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.00112715947907418, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.00112715947907418, "signal/frontier_ece_reward/centered_abs_mean": 0.007089004293084145, "signal/frontier_ece_reward/group_std_mean": 0.009215687029063701, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0007089004619047045, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0007089004619047045, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26512801051139834, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34174053072929383, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02651280015707016, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02651280015707016, "step": 210 }, { "calibration/aurc": 0.32921692199037833, "calibration/batch_distribution_entropy": 0.9850321423711392, "calibration/buffer_distribution_entropy": 0.9989667783325187, "calibration/confidence_entropy": 0.4977429598395493, "calibration/coverage@0%": 0.021108274217221133, "calibration/coverage@1%": 0.021108274217221133, "calibration/coverage@10%": 0.12594254525440313, "calibration/coverage@15%": 0.18580831090998043, "calibration/coverage@20%": 0.3199662120841487, "calibration/coverage@25%": 0.5040690741193738, "calibration/coverage@30%": 0.5967167624755382, "calibration/coverage@5%": 0.029311399217221135, "calibration/ece": 0.12032595926826466, "calibration/mean_confidence": 0.5061838544399405, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0013671875, "completions/max_length": 866.8, "completions/max_terminated_length": 866.8, "completions/mean_length": 206.01044921875, "completions/mean_terminated_length": 206.29424743652345, "completions/min_length": 0.0, "completions/min_terminated_length": 99.4, "epoch": 0.688, "grad_norm": 0.0011067946907132864, "learning_rate": 1e-06, "loss": -0.0012, "num_tokens": 731865765.0, "reward": 0.9411412000656127, "reward_std": 0.09213570952415466, "rewards/accuracy_reward": 0.53974609375, "rewards/batch_coverage_0": 0.34631012082099916, "rewards/batch_coverage_1": 0.34631012082099916, "rewards/batch_coverage_5": 0.36942777037620544, "rewards/brier_reward": 0.7906100749969482, "rewards/confidence_uniqueness_reward": 0.9518609046936035, "rewards/format_reward": 0.9986328125, "rewards/frontier_aurc_reward": -0.0028737538959831, "rewards/frontier_coverage_10": 0.10888413116335868, "rewards/frontier_coverage_15": 0.10826522558927536, "rewards/frontier_coverage_20": 0.08880022019147873, "rewards/frontier_coverage_25": 0.05820474661886692, "rewards/frontier_ece_reward": 0.004444253398105502, "rewards/frontier_entropy_batch_reward": -0.20531384348869325, "signal/accuracy_reward/centered_abs_mean": 0.101202392578125, "signal/accuracy_reward/group_std_mean": 0.13667802959680558, "signal/accuracy_reward/group_zero_std_frac": 0.6, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0506011962890625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0506011962890625, "signal/advantage_abs_mean": 0.0687965989112854, "signal/advantage_pre_scale_abs_mean": 0.0687965989112854, "signal/advantage_pre_scale_std": 0.11202323138713836, "signal/advantage_std": 0.11202323138713836, "signal/batch_coverage_0/centered_abs_mean": 0.1571878671646118, "signal/batch_coverage_0/group_std_mean": 0.19790867567062378, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0019648483488708735, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0019648483488708735, "signal/batch_coverage_1/centered_abs_mean": 0.1571878671646118, "signal/batch_coverage_1/group_std_mean": 0.19790867567062378, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0019648483488708735, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0019648483488708735, "signal/batch_coverage_5/centered_abs_mean": 0.16310351490974426, "signal/batch_coverage_5/group_std_mean": 0.2058473527431488, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002038793987594545, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002038793987594545, "signal/brier_reward/centered_abs_mean": 0.1244299903512001, "signal/brier_reward/group_std_mean": 0.16156683564186097, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.012442999146878719, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.012442999146878719, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.014019639603793621, "signal/confidence_uniqueness_reward/group_std_mean": 0.021429548785090447, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0014019639231264592, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014019639231264592, "signal/format_reward/centered_abs_mean": 0.00263671875, "signal/format_reward/group_std_mean": 0.007397671835497022, "signal/format_reward/group_zero_std_frac": 0.959375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.001318359375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.001318359375, "signal/frontier_aurc_reward/centered_abs_mean": 0.0025706546381115913, "signal/frontier_aurc_reward/group_std_mean": 0.004091716231778264, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.2133183412952346e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.2133183412952346e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.17212996184825896, "signal/frontier_coverage_10/group_std_mean": 0.22016197443008423, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0021516245789825915, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0021516245789825915, "signal/frontier_coverage_15/centered_abs_mean": 0.171197247505188, "signal/frontier_coverage_15/group_std_mean": 0.2189592659473419, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0021399656776338815, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0021399656776338815, "signal/frontier_coverage_20/centered_abs_mean": 0.13726199567317962, "signal/frontier_coverage_20/group_std_mean": 0.1761687994003296, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0017157749505713582, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0017157749505713582, "signal/frontier_coverage_25/centered_abs_mean": 0.08083412945270538, "signal/frontier_coverage_25/group_std_mean": 0.10405371636152268, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0010104266460984946, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0010104266460984946, "signal/frontier_ece_reward/centered_abs_mean": 0.007018454652279615, "signal/frontier_ece_reward/group_std_mean": 0.009031379781663417, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0007018454722128809, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0007018454722128809, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.277674013376236, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3492076277732849, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.027767401933670045, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027767401933670045, "step": 215 }, { "calibration/aurc": 0.2663918025465445, "calibration/batch_distribution_entropy": 0.962494517311927, "calibration/buffer_distribution_entropy": 0.9988609051367868, "calibration/confidence_entropy": 0.46960854051372475, "calibration/coverage@0%": 0.00352097602739726, "calibration/coverage@1%": 0.00352097602739726, "calibration/coverage@10%": 0.06456167441291585, "calibration/coverage@15%": 0.13887903620352252, "calibration/coverage@20%": 0.4034078400195694, "calibration/coverage@25%": 0.5276770425636007, "calibration/coverage@30%": 0.6698943554305284, "calibration/coverage@5%": 0.00352097602739726, "calibration/ece": 0.09837505505222606, "calibration/mean_confidence": 0.5480954701382965, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 711.6, "completions/max_terminated_length": 711.6, "completions/mean_length": 204.13212890625, "completions/mean_terminated_length": 204.2326690673828, "completions/min_length": 20.2, "completions/min_terminated_length": 106.4, "epoch": 0.704, "grad_norm": 0.0010399462189525366, "learning_rate": 1e-06, "loss": -0.0001, "num_tokens": 748822222.0, "reward": 0.943814480304718, "reward_std": 0.08291405737400055, "rewards/accuracy_reward": 0.54033203125, "rewards/batch_coverage_0": 0.3860515356063843, "rewards/batch_coverage_1": 0.3860515356063843, "rewards/batch_coverage_5": 0.4067423105239868, "rewards/brier_reward": 0.8020105719566345, "rewards/confidence_uniqueness_reward": 0.9514900326728821, "rewards/format_reward": 0.99951171875, "rewards/frontier_aurc_reward": -0.003184229042381048, "rewards/frontier_coverage_10": 0.11281085014343262, "rewards/frontier_coverage_15": 0.1123992770910263, "rewards/frontier_coverage_20": 0.09370657950639724, "rewards/frontier_coverage_25": 0.06226547360420227, "rewards/frontier_ece_reward": 0.004809339996427298, "rewards/frontier_entropy_batch_reward": -0.21398924589157103, "signal/accuracy_reward/centered_abs_mean": 0.085333251953125, "signal/accuracy_reward/group_std_mean": 0.11319491118192673, "signal/accuracy_reward/group_zero_std_frac": 0.66875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0426666259765625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0426666259765625, "signal/advantage_abs_mean": 0.0640787810087204, "signal/advantage_pre_scale_abs_mean": 0.0640787810087204, "signal/advantage_pre_scale_std": 0.10339965671300888, "signal/advantage_std": 0.10339965671300888, "signal/batch_coverage_0/centered_abs_mean": 0.1537727326154709, "signal/batch_coverage_0/group_std_mean": 0.1936441868543625, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0019221591297537088, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0019221591297537088, "signal/batch_coverage_1/centered_abs_mean": 0.1537727326154709, "signal/batch_coverage_1/group_std_mean": 0.1936441868543625, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0019221591297537088, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0019221591297537088, "signal/batch_coverage_5/centered_abs_mean": 0.16157048046588898, "signal/batch_coverage_5/group_std_mean": 0.2035224348306656, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0020196310244500635, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0020196310244500635, "signal/brier_reward/centered_abs_mean": 0.11933209598064423, "signal/brier_reward/group_std_mean": 0.15175741016864777, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011933209747076035, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011933209747076035, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013200141116976738, "signal/confidence_uniqueness_reward/group_std_mean": 0.01806243620812893, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013200141256675124, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013200141256675124, "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": 0.0004730224609375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0004730224609375, "signal/frontier_aurc_reward/centered_abs_mean": 0.003110066894441843, "signal/frontier_aurc_reward/group_std_mean": 0.0051496901083737615, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.8875835889484735e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.8875835889484735e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.15503488183021547, "signal/frontier_coverage_10/group_std_mean": 0.19720603227615358, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0019379360601305962, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0019379360601305962, "signal/frontier_coverage_15/centered_abs_mean": 0.15407584607601166, "signal/frontier_coverage_15/group_std_mean": 0.19602100551128387, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0019259481457993388, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0019259481457993388, "signal/frontier_coverage_20/centered_abs_mean": 0.12004280239343643, "signal/frontier_coverage_20/group_std_mean": 0.1533791482448578, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0015005350345745684, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0015005350345745684, "signal/frontier_coverage_25/centered_abs_mean": 0.07125528007745743, "signal/frontier_coverage_25/group_std_mean": 0.09153106212615966, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0008906910195946693, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0008906910195946693, "signal/frontier_ece_reward/centered_abs_mean": 0.006723399553447962, "signal/frontier_ece_reward/group_std_mean": 0.008735590614378452, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0006723399739712477, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0006723399739712477, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2687765061855316, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34292990565299986, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02687765210866928, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02687765210866928, "step": 220 }, { "calibration/aurc": 0.24640589202422897, "calibration/batch_distribution_entropy": 0.989703520015053, "calibration/buffer_distribution_entropy": 0.9988122389662992, "calibration/confidence_entropy": 0.4812027376027485, "calibration/coverage@0%": 0.028916187622309197, "calibration/coverage@1%": 0.028916187622309197, "calibration/coverage@10%": 0.28049015410958905, "calibration/coverage@15%": 0.3390930772994129, "calibration/coverage@20%": 0.4125420437866928, "calibration/coverage@25%": 0.5302340692270059, "calibration/coverage@30%": 0.6037426614481409, "calibration/coverage@5%": 0.18360980308219177, "calibration/ece": 0.13885218172722796, "calibration/mean_confidence": 0.5259228307403994, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 765.8, "completions/max_terminated_length": 765.8, "completions/mean_length": 200.5015625, "completions/mean_terminated_length": 200.61971740722657, "completions/min_length": 20.0, "completions/min_terminated_length": 99.6, "epoch": 0.72, "grad_norm": 0.000825917290057987, "learning_rate": 1e-06, "loss": -0.0003, "num_tokens": 765885214.0, "reward": 0.9578583598136902, "reward_std": 0.08148396462202072, "rewards/accuracy_reward": 0.56396484375, "rewards/batch_coverage_0": 0.3776911795139313, "rewards/batch_coverage_1": 0.3776911795139313, "rewards/batch_coverage_5": 0.4024718165397644, "rewards/brier_reward": 0.8012660622596741, "rewards/confidence_uniqueness_reward": 0.9531724333763123, "rewards/format_reward": 0.9994140625, "rewards/frontier_aurc_reward": -0.002669902052730322, "rewards/frontier_coverage_10": 0.09937943145632744, "rewards/frontier_coverage_15": 0.09878300726413727, "rewards/frontier_coverage_20": 0.08317163214087486, "rewards/frontier_coverage_25": 0.05664384588599205, "rewards/frontier_ece_reward": 0.004179801791906357, "rewards/frontier_entropy_batch_reward": -0.18357448577880858, "signal/accuracy_reward/centered_abs_mean": 0.082489013671875, "signal/accuracy_reward/group_std_mean": 0.11604515761137009, "signal/accuracy_reward/group_zero_std_frac": 0.646875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0412445068359375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0412445068359375, "signal/advantage_abs_mean": 0.0600632943212986, "signal/advantage_pre_scale_abs_mean": 0.0600632943212986, "signal/advantage_pre_scale_std": 0.10013380497694016, "signal/advantage_std": 0.10013380497694016, "signal/batch_coverage_0/centered_abs_mean": 0.14879767894744872, "signal/batch_coverage_0/group_std_mean": 0.18876244127750397, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0018599710194393993, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0018599710194393993, "signal/batch_coverage_1/centered_abs_mean": 0.14879767894744872, "signal/batch_coverage_1/group_std_mean": 0.18876244127750397, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0018599710194393993, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0018599710194393993, "signal/batch_coverage_5/centered_abs_mean": 0.15661481320858, "signal/batch_coverage_5/group_std_mean": 0.1996169149875641, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0019576851977035404, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0019576851977035404, "signal/brier_reward/centered_abs_mean": 0.11505196541547776, "signal/brier_reward/group_std_mean": 0.14892061650753022, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011505196802318096, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011505196802318096, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.012641343101859093, "signal/confidence_uniqueness_reward/group_std_mean": 0.01764179114252329, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0012641343753784895, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012641343753784895, "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": 0.000567626953125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.000567626953125, "signal/frontier_aurc_reward/centered_abs_mean": 0.0024686852935701608, "signal/frontier_aurc_reward/group_std_mean": 0.003913819417357444, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.085856442339718e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.085856442339718e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.15635860562324524, "signal/frontier_coverage_10/group_std_mean": 0.20113556683063508, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.001954482588917017, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.001954482588917017, "signal/frontier_coverage_15/centered_abs_mean": 0.1550187051296234, "signal/frontier_coverage_15/group_std_mean": 0.1994163453578949, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0019377337768673897, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0019377337768673897, "signal/frontier_coverage_20/centered_abs_mean": 0.11751220971345902, "signal/frontier_coverage_20/group_std_mean": 0.15130072832107544, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0014689026633277536, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0014689026633277536, "signal/frontier_coverage_25/centered_abs_mean": 0.06875089928507805, "signal/frontier_coverage_25/group_std_mean": 0.08829759359359741, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0008593862527050078, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0008593862527050078, "signal/frontier_ece_reward/centered_abs_mean": 0.0064884194172918795, "signal/frontier_ece_reward/group_std_mean": 0.008342665713280439, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0006488419487141073, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0006488419487141073, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2562240809202194, "signal/frontier_entropy_batch_reward/group_std_mean": 0.33134799003601073, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.025622408464550973, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025622408464550973, "step": 225 }, { "calibration/aurc": 0.2581623679389343, "calibration/batch_distribution_entropy": 0.9861055553539831, "calibration/buffer_distribution_entropy": 0.9989521982707135, "calibration/confidence_entropy": 0.49198630078262334, "calibration/coverage@0%": 0.006640625, "calibration/coverage@1%": 0.006640625, "calibration/coverage@10%": 0.11328125, "calibration/coverage@15%": 0.193359375, "calibration/coverage@20%": 0.394921875, "calibration/coverage@25%": 0.56875, "calibration/coverage@30%": 0.6734375, "calibration/coverage@5%": 0.062109375, "calibration/ece": 0.14015479720227256, "calibration/mean_confidence": 0.5086107207051811, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 617.6, "completions/max_terminated_length": 617.6, "completions/mean_length": 200.63984375, "completions/mean_terminated_length": 200.8167694091797, "completions/min_length": 0.0, "completions/min_terminated_length": 98.8, "epoch": 0.736, "grad_norm": 0.0009157548192888498, "learning_rate": 1e-06, "loss": -0.0005, "num_tokens": 782879350.0, "reward": 0.9553258299827576, "reward_std": 0.07961441427469254, "rewards/accuracy_reward": 0.5615234375, "rewards/batch_coverage_0": 0.3924454987049103, "rewards/batch_coverage_1": 0.3924454987049103, "rewards/batch_coverage_5": 0.40428215861320493, "rewards/brier_reward": 0.795106565952301, "rewards/confidence_uniqueness_reward": 0.9532199621200561, "rewards/format_reward": 0.99912109375, "rewards/frontier_aurc_reward": -0.0023839615285396578, "rewards/frontier_coverage_10": 0.09584366902709007, "rewards/frontier_coverage_15": 0.09519965648651123, "rewards/frontier_coverage_20": 0.07984322309494019, "rewards/frontier_coverage_25": 0.055439715087413785, "rewards/frontier_ece_reward": 0.003425889555364847, "rewards/frontier_entropy_batch_reward": -0.1908560574054718, "signal/accuracy_reward/centered_abs_mean": 0.0801025390625, "signal/accuracy_reward/group_std_mean": 0.10931812822818757, "signal/accuracy_reward/group_zero_std_frac": 0.675, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.04005126953125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.04005126953125, "signal/advantage_abs_mean": 0.0602330282330513, "signal/advantage_pre_scale_abs_mean": 0.0602330282330513, "signal/advantage_pre_scale_std": 0.09761163294315338, "signal/advantage_std": 0.09761163294315338, "signal/batch_coverage_0/centered_abs_mean": 0.15604349374771118, "signal/batch_coverage_0/group_std_mean": 0.19706358909606933, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0019505437463521957, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0019505437463521957, "signal/batch_coverage_1/centered_abs_mean": 0.15604349374771118, "signal/batch_coverage_1/group_std_mean": 0.19706358909606933, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0019505437463521957, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0019505437463521957, "signal/batch_coverage_5/centered_abs_mean": 0.16017458736896514, "signal/batch_coverage_5/group_std_mean": 0.2024128943681717, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0020021823700517418, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0020021823700517418, "signal/brier_reward/centered_abs_mean": 0.11506714671850204, "signal/brier_reward/group_std_mean": 0.14947323501110077, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01150671523064375, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01150671523064375, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.01279607992619276, "signal/confidence_uniqueness_reward/group_std_mean": 0.018451326154172422, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0012796080438420177, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012796080438420177, "signal/format_reward/centered_abs_mean": 0.001690673828125, "signal/format_reward/group_std_mean": 0.004635535972192883, "signal/format_reward/group_zero_std_frac": 0.975, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0008453369140625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0008453369140625, "signal/frontier_aurc_reward/centered_abs_mean": 0.0020353728905320168, "signal/frontier_aurc_reward/group_std_mean": 0.0033360558096319436, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.5442160040256568e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.5442160040256568e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.16068104207515715, "signal/frontier_coverage_10/group_std_mean": 0.2068079262971878, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0020085131283849476, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0020085131283849476, "signal/frontier_coverage_15/centered_abs_mean": 0.15899776816368102, "signal/frontier_coverage_15/group_std_mean": 0.20465264618396758, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0019874721299856903, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0019874721299856903, "signal/frontier_coverage_20/centered_abs_mean": 0.1162015214562416, "signal/frontier_coverage_20/group_std_mean": 0.1500241219997406, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0014525190461426973, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0014525190461426973, "signal/frontier_coverage_25/centered_abs_mean": 0.0675393745303154, "signal/frontier_coverage_25/group_std_mean": 0.0872299998998642, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.000844242179300636, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.000844242179300636, "signal/frontier_ece_reward/centered_abs_mean": 0.006053383741527796, "signal/frontier_ece_reward/group_std_mean": 0.00789744695648551, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0006053383927792311, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0006053383927792311, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2693478137254715, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34564956426620486, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026934782043099403, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026934782043099403, "step": 230 }, { "calibration/aurc": 0.279803542254142, "calibration/batch_distribution_entropy": 0.974489520577562, "calibration/buffer_distribution_entropy": 0.9990645091711888, "calibration/confidence_entropy": 0.4795515053061793, "calibration/coverage@0%": 0.009378057729941292, "calibration/coverage@1%": 0.009378057729941292, "calibration/coverage@10%": 0.0953155577299413, "calibration/coverage@15%": 0.23052302470645794, "calibration/coverage@20%": 0.36346930039138947, "calibration/coverage@25%": 0.47647535469667324, "calibration/coverage@30%": 0.6070495963796477, "calibration/coverage@5%": 0.03515930772994129, "calibration/ece": 0.1178076431437709, "calibration/mean_confidence": 0.4731464528034646, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 563.0, "completions/max_terminated_length": 563.0, "completions/mean_length": 200.85322265625, "completions/mean_terminated_length": 200.97051696777345, "completions/min_length": 39.8, "completions/min_terminated_length": 100.2, "epoch": 0.752, "grad_norm": 0.0007815993158146739, "learning_rate": 1e-06, "loss": -0.0003, "num_tokens": 800163287.0, "reward": 0.9515892624855041, "reward_std": 0.08244798332452774, "rewards/accuracy_reward": 0.55732421875, "rewards/batch_coverage_0": 0.37442896962165834, "rewards/batch_coverage_1": 0.37442896962165834, "rewards/batch_coverage_5": 0.39017086029052733, "rewards/brier_reward": 0.7969243407249451, "rewards/confidence_uniqueness_reward": 0.9523859143257141, "rewards/format_reward": 0.9994140625, "rewards/frontier_aurc_reward": -0.002785804169252515, "rewards/frontier_coverage_10": 0.10273134410381317, "rewards/frontier_coverage_15": 0.1016263224184513, "rewards/frontier_coverage_20": 0.07584753520786762, "rewards/frontier_coverage_25": 0.051239912211894986, "rewards/frontier_ece_reward": 0.0035949977580457928, "rewards/frontier_entropy_batch_reward": -0.20416504740715027, "signal/accuracy_reward/centered_abs_mean": 0.080743408203125, "signal/accuracy_reward/group_std_mean": 0.11237793564796447, "signal/accuracy_reward/group_zero_std_frac": 0.65625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0403717041015625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0403717041015625, "signal/advantage_abs_mean": 0.062051520496606824, "signal/advantage_pre_scale_abs_mean": 0.062051520496606824, "signal/advantage_pre_scale_std": 0.10203461945056916, "signal/advantage_std": 0.10203461945056916, "signal/batch_coverage_0/centered_abs_mean": 0.1481984257698059, "signal/batch_coverage_0/group_std_mean": 0.18765933215618133, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0018524803454056382, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0018524803454056382, "signal/batch_coverage_1/centered_abs_mean": 0.1481984257698059, "signal/batch_coverage_1/group_std_mean": 0.18765933215618133, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0018524803454056382, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0018524803454056382, "signal/batch_coverage_5/centered_abs_mean": 0.15228593349456787, "signal/batch_coverage_5/group_std_mean": 0.19333537220954894, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.001903574145399034, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.001903574145399034, "signal/brier_reward/centered_abs_mean": 0.11209099441766739, "signal/brier_reward/group_std_mean": 0.1458438515663147, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011209099367260934, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011209099367260934, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.0131287207826972, "signal/confidence_uniqueness_reward/group_std_mean": 0.017994565702974797, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013128720689564942, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013128720689564942, "signal/format_reward/centered_abs_mean": 0.001123046875, "signal/format_reward/group_std_mean": 0.0029782545287162067, "signal/format_reward/group_zero_std_frac": 0.984375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0005615234375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0005615234375, "signal/frontier_aurc_reward/centered_abs_mean": 0.0023489481303840876, "signal/frontier_aurc_reward/group_std_mean": 0.0038049173541367053, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.936185264843516e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.936185264843516e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1500999391078949, "signal/frontier_coverage_10/group_std_mean": 0.1938435137271881, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0018762492574751378, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0018762492574751378, "signal/frontier_coverage_15/centered_abs_mean": 0.14748037457466126, "signal/frontier_coverage_15/group_std_mean": 0.19050345122814177, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0018435046542435885, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0018435046542435885, "signal/frontier_coverage_20/centered_abs_mean": 0.10476779788732529, "signal/frontier_coverage_20/group_std_mean": 0.1358924016356468, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0013095975620672107, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0013095975620672107, "signal/frontier_coverage_25/centered_abs_mean": 0.061764152348041536, "signal/frontier_coverage_25/group_std_mean": 0.07970290631055832, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007720518973655998, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007720518973655998, "signal/frontier_ece_reward/centered_abs_mean": 0.0061201823875308035, "signal/frontier_ece_reward/group_std_mean": 0.00792367635294795, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0006120182340964675, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0006120182340964675, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.27745115756988525, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3531729578971863, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.027745116874575614, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027745116874575614, "step": 235 }, { "calibration/aurc": 0.29600612717330954, "calibration/batch_distribution_entropy": 0.983338097257603, "calibration/buffer_distribution_entropy": 0.9990942841964813, "calibration/confidence_entropy": 0.499992549927008, "calibration/coverage@0%": 0.03203125, "calibration/coverage@1%": 0.03203125, "calibration/coverage@10%": 0.20859375, "calibration/coverage@15%": 0.2703125, "calibration/coverage@20%": 0.353515625, "calibration/coverage@25%": 0.43671875, "calibration/coverage@30%": 0.528125, "calibration/coverage@5%": 0.11640625, "calibration/ece": 0.1470123604244844, "calibration/mean_confidence": 0.5033313158570779, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0009765625, "completions/max_length": 806.2, "completions/max_terminated_length": 806.2, "completions/mean_length": 207.9755859375, "completions/mean_terminated_length": 208.1797088623047, "completions/min_length": 0.0, "completions/min_terminated_length": 100.0, "epoch": 0.768, "grad_norm": 0.0009092589607462287, "learning_rate": 1e-06, "loss": -0.0005, "num_tokens": 817225661.0, "reward": 0.929730761051178, "reward_std": 0.08614100962877273, "rewards/accuracy_reward": 0.5119140625, "rewards/batch_coverage_0": 0.3782623589038849, "rewards/batch_coverage_1": 0.3782623589038849, "rewards/batch_coverage_5": 0.3933583080768585, "rewards/brier_reward": 0.8018325328826904, "rewards/confidence_uniqueness_reward": 0.9518675327301025, "rewards/format_reward": 0.9990234375, "rewards/frontier_aurc_reward": -0.0028852508403360844, "rewards/frontier_coverage_10": 0.1340809166431427, "rewards/frontier_coverage_15": 0.1323210656642914, "rewards/frontier_coverage_20": 0.09949396848678589, "rewards/frontier_coverage_25": 0.0626144677400589, "rewards/frontier_ece_reward": 0.003909257845953107, "rewards/frontier_entropy_batch_reward": -0.2119278460741043, "signal/accuracy_reward/centered_abs_mean": 0.08770751953125, "signal/accuracy_reward/group_std_mean": 0.11465359032154084, "signal/accuracy_reward/group_zero_std_frac": 0.678125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.043853759765625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.043853759765625, "signal/advantage_abs_mean": 0.06641755923628807, "signal/advantage_pre_scale_abs_mean": 0.06641755923628807, "signal/advantage_pre_scale_std": 0.10704567879438401, "signal/advantage_std": 0.10704567879438401, "signal/batch_coverage_0/centered_abs_mean": 0.15254861116409302, "signal/batch_coverage_0/group_std_mean": 0.1927157759666443, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.00190685773268342, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.00190685773268342, "signal/batch_coverage_1/centered_abs_mean": 0.15254861116409302, "signal/batch_coverage_1/group_std_mean": 0.1927157759666443, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.00190685773268342, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.00190685773268342, "signal/batch_coverage_5/centered_abs_mean": 0.15637167096138, "signal/batch_coverage_5/group_std_mean": 0.19779276847839355, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0019546458963304756, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0019546458963304756, "signal/brier_reward/centered_abs_mean": 0.11479622423648835, "signal/brier_reward/group_std_mean": 0.14936064183712006, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011479622684419155, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011479622684419155, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013476397097110748, "signal/confidence_uniqueness_reward/group_std_mean": 0.01955820508301258, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013476397143676877, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013476397143676877, "signal/format_reward/centered_abs_mean": 0.0018798828125, "signal/format_reward/group_std_mean": 0.005187963135540485, "signal/format_reward/group_zero_std_frac": 0.971875, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00093994140625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00093994140625, "signal/frontier_aurc_reward/centered_abs_mean": 0.0025279912631958725, "signal/frontier_aurc_reward/group_std_mean": 0.004214816214516759, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.159988918923773e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.159988918923773e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1541954070329666, "signal/frontier_coverage_10/group_std_mean": 0.19759481847286225, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.00192744261585176, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.00192744261585176, "signal/frontier_coverage_15/centered_abs_mean": 0.15146732330322266, "signal/frontier_coverage_15/group_std_mean": 0.1941421002149582, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0018933416111394763, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0018933416111394763, "signal/frontier_coverage_20/centered_abs_mean": 0.10409807860851288, "signal/frontier_coverage_20/group_std_mean": 0.1340539425611496, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.001301226019859314, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.001301226019859314, "signal/frontier_coverage_25/centered_abs_mean": 0.06242971271276474, "signal/frontier_coverage_25/group_std_mean": 0.0802663192152977, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007803714368492364, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007803714368492364, "signal/frontier_ece_reward/centered_abs_mean": 0.005812339950352907, "signal/frontier_ece_reward/group_std_mean": 0.007544164266437292, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005812340183183551, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005812340183183551, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2772362411022186, "signal/frontier_entropy_batch_reward/group_std_mean": 0.35202175974845884, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02772362343966961, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02772362343966961, "step": 240 }, { "calibration/aurc": 0.33529252620065, "calibration/batch_distribution_entropy": 0.9840010033660377, "calibration/buffer_distribution_entropy": 0.9989733240096742, "calibration/confidence_entropy": 0.4773662194260801, "calibration/coverage@0%": 0.021101430439773795, "calibration/coverage@1%": 0.021101430439773795, "calibration/coverage@10%": 0.1504967306366242, "calibration/coverage@15%": 0.19353007217205725, "calibration/coverage@20%": 0.3306415779523707, "calibration/coverage@25%": 0.3733942403443148, "calibration/coverage@30%": 0.407906049682574, "calibration/coverage@5%": 0.10391393043977379, "calibration/ece": 0.16186048409154125, "calibration/mean_confidence": 0.5168986678234327, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.00126953125, "completions/max_length": 818.0, "completions/max_terminated_length": 818.0, "completions/mean_length": 208.98857421875, "completions/mean_terminated_length": 209.25534362792968, "completions/min_length": 20.4, "completions/min_terminated_length": 104.8, "epoch": 0.784, "grad_norm": 0.0010283656883984804, "learning_rate": 1e-06, "loss": -0.0009, "num_tokens": 834540072.0, "reward": 0.9450501680374146, "reward_std": 0.08672762364149093, "rewards/accuracy_reward": 0.54892578125, "rewards/batch_coverage_0": 0.3657715082168579, "rewards/batch_coverage_1": 0.3657715082168579, "rewards/batch_coverage_5": 0.374066036939621, "rewards/brier_reward": 0.7803781747817993, "rewards/confidence_uniqueness_reward": 0.9520835399627685, "rewards/format_reward": 0.9986328125, "rewards/frontier_aurc_reward": -0.003116954304277897, "rewards/frontier_coverage_10": 0.09036868810653687, "rewards/frontier_coverage_15": 0.08844961896538735, "rewards/frontier_coverage_20": 0.06776739954948426, "rewards/frontier_coverage_25": 0.04740912914276123, "rewards/frontier_ece_reward": 0.00269833414349705, "rewards/frontier_entropy_batch_reward": -0.19701214432716369, "signal/accuracy_reward/centered_abs_mean": 0.094317626953125, "signal/accuracy_reward/group_std_mean": 0.12160500586032867, "signal/accuracy_reward/group_zero_std_frac": 0.6625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0471588134765625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0471588134765625, "signal/advantage_abs_mean": 0.0673556350171566, "signal/advantage_pre_scale_abs_mean": 0.0673556350171566, "signal/advantage_pre_scale_std": 0.10786458998918533, "signal/advantage_std": 0.10786458998918533, "signal/batch_coverage_0/centered_abs_mean": 0.1579074203968048, "signal/batch_coverage_0/group_std_mean": 0.19955972731113433, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0019738428527489305, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0019738428527489305, "signal/batch_coverage_1/centered_abs_mean": 0.1579074203968048, "signal/batch_coverage_1/group_std_mean": 0.19955972731113433, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0019738428527489305, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0019738428527489305, "signal/batch_coverage_5/centered_abs_mean": 0.15928060710430145, "signal/batch_coverage_5/group_std_mean": 0.20109879970550537, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0019910075701773168, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0019910075701773168, "signal/brier_reward/centered_abs_mean": 0.12201150059700012, "signal/brier_reward/group_std_mean": 0.1565765470266342, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.012201149947941303, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.012201149947941303, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013731702789664268, "signal/confidence_uniqueness_reward/group_std_mean": 0.020425675064325334, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013731702696532012, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013731702696532012, "signal/format_reward/centered_abs_mean": 0.0026123046875, "signal/format_reward/group_std_mean": 0.006725054979324341, "signal/format_reward/group_zero_std_frac": 0.965625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00130615234375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00130615234375, "signal/frontier_aurc_reward/centered_abs_mean": 0.002710724249482155, "signal/frontier_aurc_reward/group_std_mean": 0.004519741144031286, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.388405530131422e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.388405530131422e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1643848717212677, "signal/frontier_coverage_10/group_std_mean": 0.2106505572795868, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0020548109197989104, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0020548109197989104, "signal/frontier_coverage_15/centered_abs_mean": 0.15967891812324525, "signal/frontier_coverage_15/group_std_mean": 0.20467236042022705, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.001995986490510404, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.001995986490510404, "signal/frontier_coverage_20/centered_abs_mean": 0.10617948025465011, "signal/frontier_coverage_20/group_std_mean": 0.1370469257235527, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0013272434938699008, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0013272434938699008, "signal/frontier_coverage_25/centered_abs_mean": 0.06253089681267739, "signal/frontier_coverage_25/group_std_mean": 0.08045676499605178, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007816361961886287, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007816361961886287, "signal/frontier_ece_reward/centered_abs_mean": 0.005935485381633044, "signal/frontier_ece_reward/group_std_mean": 0.007646645605564118, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005935485474765301, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005935485474765301, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2669118523597717, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3446871697902679, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026691185683012007, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026691185683012007, "step": 245 }, { "calibration/aurc": 0.22169967395791867, "calibration/batch_distribution_entropy": 0.9808965219487114, "calibration/buffer_distribution_entropy": 0.9989334339384272, "calibration/confidence_entropy": 0.4962449636115854, "calibration/coverage@0%": 0.01914368272994129, "calibration/coverage@1%": 0.01914368272994129, "calibration/coverage@10%": 0.1898873226516634, "calibration/coverage@15%": 0.42164337695694715, "calibration/coverage@20%": 0.557260579745597, "calibration/coverage@25%": 0.6354253302348336, "calibration/coverage@30%": 0.7085043113992172, "calibration/coverage@5%": 0.06367493272994129, "calibration/ece": 0.11412641986169644, "calibration/mean_confidence": 0.471643517823892, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 566.2, "completions/max_terminated_length": 566.2, "completions/mean_length": 210.19560546875, "completions/mean_terminated_length": 210.33868408203125, "completions/min_length": 43.2, "completions/min_terminated_length": 110.4, "epoch": 0.8, "grad_norm": 0.0011549158953130245, "learning_rate": 1e-06, "loss": -0.0006, "num_tokens": 851703035.0, "reward": 0.9630751609802246, "reward_std": 0.08390648663043976, "rewards/accuracy_reward": 0.5794921875, "rewards/batch_coverage_0": 0.3790114104747772, "rewards/batch_coverage_1": 0.3790114104747772, "rewards/batch_coverage_5": 0.39591498374938966, "rewards/brier_reward": 0.8030319094657898, "rewards/confidence_uniqueness_reward": 0.9523315668106079, "rewards/format_reward": 0.99921875, "rewards/frontier_aurc_reward": -0.002687943633645773, "rewards/frontier_coverage_10": 0.09324493408203124, "rewards/frontier_coverage_15": 0.09011918753385544, "rewards/frontier_coverage_20": 0.07055493593215942, "rewards/frontier_coverage_25": 0.05035630464553833, "rewards/frontier_ece_reward": 0.0030595229472965, "rewards/frontier_entropy_batch_reward": -0.20316683053970336, "signal/accuracy_reward/centered_abs_mean": 0.08941650390625, "signal/accuracy_reward/group_std_mean": 0.11753687709569931, "signal/accuracy_reward/group_zero_std_frac": 0.6625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.044708251953125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.044708251953125, "signal/advantage_abs_mean": 0.06446013525128365, "signal/advantage_pre_scale_abs_mean": 0.06446013525128365, "signal/advantage_pre_scale_std": 0.10524639785289765, "signal/advantage_std": 0.10524639785289765, "signal/batch_coverage_0/centered_abs_mean": 0.1497106820344925, "signal/batch_coverage_0/group_std_mean": 0.18885045349597931, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0018713835626840592, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0018713835626840592, "signal/batch_coverage_1/centered_abs_mean": 0.1497106820344925, "signal/batch_coverage_1/group_std_mean": 0.18885045349597931, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0018713835626840592, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0018713835626840592, "signal/batch_coverage_5/centered_abs_mean": 0.15395812690258026, "signal/batch_coverage_5/group_std_mean": 0.19484901428222656, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0019244766561314463, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0019244766561314463, "signal/brier_reward/centered_abs_mean": 0.10814285576343537, "signal/brier_reward/group_std_mean": 0.14126260280609132, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.010814285837113857, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.010814285837113857, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.012742668017745018, "signal/confidence_uniqueness_reward/group_std_mean": 0.017975622043013572, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.001274266792461276, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001274266792461276, "signal/format_reward/centered_abs_mean": 0.00150146484375, "signal/format_reward/group_std_mean": 0.004083108808845282, "signal/format_reward/group_zero_std_frac": 0.978125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.000750732421875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.000750732421875, "signal/frontier_aurc_reward/centered_abs_mean": 0.002264462294988334, "signal/frontier_aurc_reward/group_std_mean": 0.0035160879138857125, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.830577941494994e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.830577941494994e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.15162277817726136, "signal/frontier_coverage_10/group_std_mean": 0.19576776325702666, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0018952847458422184, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0018952847458422184, "signal/frontier_coverage_15/centered_abs_mean": 0.14256809502840043, "signal/frontier_coverage_15/group_std_mean": 0.18429208993911744, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0017821011831983923, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0017821011831983923, "signal/frontier_coverage_20/centered_abs_mean": 0.096474289894104, "signal/frontier_coverage_20/group_std_mean": 0.1258459731936455, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0012059286236763, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0012059286236763, "signal/frontier_coverage_25/centered_abs_mean": 0.056810838729143144, "signal/frontier_coverage_25/group_std_mean": 0.0737529531121254, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007101355004124343, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007101355004124343, "signal/frontier_ece_reward/centered_abs_mean": 0.005836281552910804, "signal/frontier_ece_reward/group_std_mean": 0.007533130422234535, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005836281925439835, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005836281925439835, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.271066153049469, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34618580937385557, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02710661627352238, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02710661627352238, "step": 250 }, { "epoch": 0.8, "eval_calibration/aurc": 0.43066331404991554, "eval_calibration/batch_distribution_entropy": 0.9116690358810688, "eval_calibration/buffer_distribution_entropy": 0.999053967629298, "eval_calibration/confidence_entropy": 0.5061534635021133, "eval_calibration/coverage@0%": 0.10181451612903225, "eval_calibration/coverage@1%": 0.10181451612903225, "eval_calibration/coverage@10%": 0.10181451612903225, "eval_calibration/coverage@15%": 0.10962701612903225, "eval_calibration/coverage@20%": 0.14868951612903225, "eval_calibration/coverage@25%": 0.24243951612903225, "eval_calibration/coverage@30%": 0.27368951612903225, "eval_calibration/coverage@5%": 0.10181451612903225, "eval_calibration/ece": 0.18686498157000686, "eval_calibration/mean_confidence": 0.4403038083809223, "eval_completions/clipped_ratio": 0.001953125, "eval_completions/max_length": 393.5, "eval_completions/max_terminated_length": 393.5, "eval_completions/mean_length": 214.46046447753906, "eval_completions/mean_terminated_length": 214.87223434448242, "eval_completions/min_length": 94.0, "eval_completions/min_terminated_length": 122.0, "eval_loss": 0.0, "eval_num_tokens": 851703035.0, "eval_reward": 0.7995923310518265, "eval_reward_std": 0.22984296455979347, "eval_rewards/accuracy_reward": 0.439453125, "eval_rewards/batch_coverage_0": 0.14397422783076763, "eval_rewards/batch_coverage_1": 0.14397422783076763, "eval_rewards/batch_coverage_5": 0.14397422783076763, "eval_rewards/brier_reward": 0.7886014431715012, "eval_rewards/confidence_uniqueness_reward": 0.8983666300773621, "eval_rewards/format_reward": 0.998046875, "eval_rewards/frontier_aurc_reward": -0.0031244863057509065, "eval_rewards/frontier_coverage_10": 0.17149265855550766, "eval_rewards/frontier_coverage_15": 0.15857548639178276, "eval_rewards/frontier_coverage_20": 0.10879995115101337, "eval_rewards/frontier_coverage_25": 0.05962690059095621, "eval_rewards/frontier_ece_reward": 0.003590614243876189, "eval_rewards/frontier_entropy_batch_reward": -0.998046875, "eval_runtime": 29.1507, "eval_samples_per_second": 17.152, "eval_signal/accuracy_reward/centered_abs_mean": 0.4747314453125, "eval_signal/accuracy_reward/group_std_mean": 0.49468091130256653, "eval_signal/accuracy_reward/group_zero_std_frac": 0.0, "eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.23736572265625, "eval_signal/accuracy_reward/weight": 0.5, "eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.23736572265625, "eval_signal/advantage_abs_mean": 0.21516040712594986, "eval_signal/advantage_pre_scale_abs_mean": 0.21516040712594986, "eval_signal/advantage_pre_scale_std": 0.22736676782369614, "eval_signal/advantage_std": 0.22736676782369614, "eval_signal/batch_coverage_0/centered_abs_mean": 0.29899947345256805, "eval_signal/batch_coverage_0/group_std_mean": 0.36741646379232407, "eval_signal/batch_coverage_0/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0037374935927800834, "eval_signal/batch_coverage_0/weight": 0.012500000186264515, "eval_signal/batch_coverage_0/weighted_centered_abs_mean": 0.0037374935927800834, "eval_signal/batch_coverage_1/centered_abs_mean": 0.29899947345256805, "eval_signal/batch_coverage_1/group_std_mean": 0.36741646379232407, "eval_signal/batch_coverage_1/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0037374935927800834, "eval_signal/batch_coverage_1/weight": 0.012500000186264515, "eval_signal/batch_coverage_1/weighted_centered_abs_mean": 0.0037374935927800834, "eval_signal/batch_coverage_5/centered_abs_mean": 0.29899947345256805, "eval_signal/batch_coverage_5/group_std_mean": 0.36741646379232407, "eval_signal/batch_coverage_5/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0037374935927800834, "eval_signal/batch_coverage_5/weight": 0.012500000186264515, "eval_signal/batch_coverage_5/weighted_centered_abs_mean": 0.0037374935927800834, "eval_signal/brier_reward/centered_abs_mean": 0.19343392550945282, "eval_signal/brier_reward/group_std_mean": 0.2451561614871025, "eval_signal/brier_reward/group_zero_std_frac": 0.0, "eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.019343392923474312, "eval_signal/brier_reward/weight": 0.10000000149011612, "eval_signal/brier_reward/weighted_centered_abs_mean": 0.019343392923474312, "eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.0390698816627264, "eval_signal/confidence_uniqueness_reward/group_std_mean": 0.05191366467624903, "eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.003906988305971026, "eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.003906988305971026, "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": 0.00189208984375, "eval_signal/format_reward/weight": 0.5, "eval_signal/format_reward/weighted_centered_abs_mean": 0.00189208984375, "eval_signal/frontier_aurc_reward/centered_abs_mean": 0.004009017604403198, "eval_signal/frontier_aurc_reward/group_std_mean": 0.0072981121484190226, "eval_signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 5.0112721510231495e-05, "eval_signal/frontier_aurc_reward/weight": 0.012500000186264515, "eval_signal/frontier_aurc_reward/weighted_centered_abs_mean": 5.0112721510231495e-05, "eval_signal/frontier_coverage_10/centered_abs_mean": 0.3679986968636513, "eval_signal/frontier_coverage_10/group_std_mean": 0.4471782371401787, "eval_signal/frontier_coverage_10/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.00459998392034322, "eval_signal/frontier_coverage_10/weight": 0.012500000186264515, "eval_signal/frontier_coverage_10/weighted_centered_abs_mean": 0.00459998392034322, "eval_signal/frontier_coverage_15/centered_abs_mean": 0.3412696570158005, "eval_signal/frontier_coverage_15/group_std_mean": 0.4160052463412285, "eval_signal/frontier_coverage_15/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.004265870549716055, "eval_signal/frontier_coverage_15/weight": 0.012500000186264515, "eval_signal/frontier_coverage_15/weighted_centered_abs_mean": 0.004265870549716055, "eval_signal/frontier_coverage_20/centered_abs_mean": 0.22044864669442177, "eval_signal/frontier_coverage_20/group_std_mean": 0.2757972851395607, "eval_signal/frontier_coverage_20/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0027556080603972077, "eval_signal/frontier_coverage_20/weight": 0.012500000186264515, "eval_signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0027556080603972077, "eval_signal/frontier_coverage_25/centered_abs_mean": 0.10226288437843323, "eval_signal/frontier_coverage_25/group_std_mean": 0.13231030479073524, "eval_signal/frontier_coverage_25/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0012782860139850527, "eval_signal/frontier_coverage_25/weight": 0.012500000186264515, "eval_signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0012782860139850527, "eval_signal/frontier_ece_reward/centered_abs_mean": 0.008236418245360255, "eval_signal/frontier_ece_reward/group_std_mean": 0.010502856457605958, "eval_signal/frontier_ece_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0008236418216256425, "eval_signal/frontier_ece_reward/weight": 0.10000000149011612, "eval_signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0008236418216256425, "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": 0.0003784179862122983, "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": 0.137, "step": 250 }, { "calibration/aurc": 0.22780578615691033, "calibration/batch_distribution_entropy": 0.9672760911159412, "calibration/buffer_distribution_entropy": 0.9990695365012362, "calibration/confidence_entropy": 0.4598628311556777, "calibration/coverage@0%": 0.024633836839530333, "calibration/coverage@1%": 0.024633836839530333, "calibration/coverage@10%": 0.16659277152641877, "calibration/coverage@15%": 0.23740826810176124, "calibration/coverage@20%": 0.5018116674444154, "calibration/coverage@25%": 0.7135486211500236, "calibration/coverage@30%": 0.8160683708414872, "calibration/coverage@5%": 0.09069456335616438, "calibration/ece": 0.15014448047747853, "calibration/mean_confidence": 0.5083854550712651, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0009765625, "completions/max_length": 643.2, "completions/max_terminated_length": 643.2, "completions/mean_length": 209.59443359375, "completions/mean_terminated_length": 209.79932250976563, "completions/min_length": 22.6, "completions/min_terminated_length": 109.0, "epoch": 0.816, "grad_norm": 0.0021682712249457836, "learning_rate": 1e-06, "loss": -0.0005, "num_tokens": 868948450.0, "reward": 0.9618038535118103, "reward_std": 0.08643427938222885, "rewards/accuracy_reward": 0.580078125, "rewards/batch_coverage_0": 0.3550035834312439, "rewards/batch_coverage_1": 0.3550035834312439, "rewards/batch_coverage_5": 0.3752487063407898, "rewards/brier_reward": 0.7860243082046509, "rewards/confidence_uniqueness_reward": 0.9524977564811706, "rewards/format_reward": 0.9990234375, "rewards/frontier_aurc_reward": -0.00290919526014477, "rewards/frontier_coverage_10": 0.07558527514338494, "rewards/frontier_coverage_15": 0.07363248616456985, "rewards/frontier_coverage_20": 0.06081184968352318, "rewards/frontier_coverage_25": 0.0468586653470993, "rewards/frontier_ece_reward": 0.002542387694120407, "rewards/frontier_entropy_batch_reward": -0.18593812882900237, "signal/accuracy_reward/centered_abs_mean": 0.09586181640625, "signal/accuracy_reward/group_std_mean": 0.1268088310956955, "signal/accuracy_reward/group_zero_std_frac": 0.6375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.047930908203125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.047930908203125, "signal/advantage_abs_mean": 0.06643177568912506, "signal/advantage_pre_scale_abs_mean": 0.06643177568912506, "signal/advantage_pre_scale_std": 0.11026870906352997, "signal/advantage_std": 0.11026870906352997, "signal/batch_coverage_0/centered_abs_mean": 0.14911384284496307, "signal/batch_coverage_0/group_std_mean": 0.19083908796310425, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0018639229936525225, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0018639229936525225, "signal/batch_coverage_1/centered_abs_mean": 0.14911384284496307, "signal/batch_coverage_1/group_std_mean": 0.19083908796310425, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0018639229936525225, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0018639229936525225, "signal/batch_coverage_5/centered_abs_mean": 0.15598122775554657, "signal/batch_coverage_5/group_std_mean": 0.19980722665786743, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0019497653935104608, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0019497653935104608, "signal/brier_reward/centered_abs_mean": 0.11971672624349594, "signal/brier_reward/group_std_mean": 0.1536404848098755, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01197167281061411, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01197167281061411, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013281514309346676, "signal/confidence_uniqueness_reward/group_std_mean": 0.0185428224503994, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013281514402478933, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013281514402478933, "signal/format_reward/centered_abs_mean": 0.00184326171875, "signal/format_reward/group_std_mean": 0.004456133535131812, "signal/format_reward/group_zero_std_frac": 0.978125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.000921630859375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.000921630859375, "signal/frontier_aurc_reward/centered_abs_mean": 0.002749479515478015, "signal/frontier_aurc_reward/group_std_mean": 0.0043839504010975364, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.436849401623476e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.436849401623476e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.15883970260620117, "signal/frontier_coverage_10/group_std_mean": 0.20446773767471313, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.001985496236011386, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.001985496236011386, "signal/frontier_coverage_15/centered_abs_mean": 0.14838777780532836, "signal/frontier_coverage_15/group_std_mean": 0.19125499129295348, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.001854847208596766, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.001854847208596766, "signal/frontier_coverage_20/centered_abs_mean": 0.09930170029401779, "signal/frontier_coverage_20/group_std_mean": 0.1285435825586319, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0012412712909281253, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0012412712909281253, "signal/frontier_coverage_25/centered_abs_mean": 0.05984411463141441, "signal/frontier_coverage_25/group_std_mean": 0.07677814811468124, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007480514235794544, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007480514235794544, "signal/frontier_ece_reward/centered_abs_mean": 0.0057809468358755115, "signal/frontier_ece_reward/group_std_mean": 0.0074549101293087, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005780947161838412, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005780947161838412, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2540502518415451, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3302201986312866, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.025405024737119676, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.025405024737119676, "step": 255 }, { "calibration/aurc": 0.284904839354618, "calibration/batch_distribution_entropy": 0.9718885125918584, "calibration/buffer_distribution_entropy": 0.9989393390764203, "calibration/confidence_entropy": 0.4772645858069799, "calibration/coverage@0%": 0.05513392857142857, "calibration/coverage@1%": 0.05513392857142857, "calibration/coverage@10%": 0.21495917930528377, "calibration/coverage@15%": 0.24388377568493153, "calibration/coverage@20%": 0.31344178082191776, "calibration/coverage@25%": 0.4181460983365949, "calibration/coverage@30%": 0.5056636802837573, "calibration/coverage@5%": 0.1661088857632094, "calibration/ece": 0.09845068028173118, "calibration/mean_confidence": 0.48191987600973485, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 680.2, "completions/max_terminated_length": 680.2, "completions/mean_length": 212.37978515625, "completions/mean_terminated_length": 212.46303100585936, "completions/min_length": 41.2, "completions/min_terminated_length": 104.4, "epoch": 0.832, "grad_norm": 0.001155984471552074, "learning_rate": 1e-06, "loss": -0.0001, "num_tokens": 886131571.0, "reward": 0.9464442491531372, "reward_std": 0.08505731225013732, "rewards/accuracy_reward": 0.54736328125, "rewards/batch_coverage_0": 0.39283851981163026, "rewards/batch_coverage_1": 0.39283851981163026, "rewards/batch_coverage_5": 0.4092761754989624, "rewards/brier_reward": 0.8121486306190491, "rewards/confidence_uniqueness_reward": 0.9508059620857239, "rewards/format_reward": 0.999609375, "rewards/frontier_aurc_reward": -0.002716775145381689, "rewards/frontier_coverage_10": 0.12564955800771713, "rewards/frontier_coverage_15": 0.11618602871894837, "rewards/frontier_coverage_20": 0.08242397531867027, "rewards/frontier_coverage_25": 0.059075400978326795, "rewards/frontier_ece_reward": 0.003620534110814333, "rewards/frontier_entropy_batch_reward": -0.23394226431846618, "signal/accuracy_reward/centered_abs_mean": 0.090130615234375, "signal/accuracy_reward/group_std_mean": 0.1237365499138832, "signal/accuracy_reward/group_zero_std_frac": 0.628125, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0450653076171875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0450653076171875, "signal/advantage_abs_mean": 0.06471554562449455, "signal/advantage_pre_scale_abs_mean": 0.06471554562449455, "signal/advantage_pre_scale_std": 0.1067628338932991, "signal/advantage_std": 0.1067628338932991, "signal/batch_coverage_0/centered_abs_mean": 0.1412302404642105, "signal/batch_coverage_0/group_std_mean": 0.18023532927036284, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.001765377982519567, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.001765377982519567, "signal/batch_coverage_1/centered_abs_mean": 0.1412302404642105, "signal/batch_coverage_1/group_std_mean": 0.18023532927036284, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.001765377982519567, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.001765377982519567, "signal/batch_coverage_5/centered_abs_mean": 0.1466294050216675, "signal/batch_coverage_5/group_std_mean": 0.18722580075263978, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0018328676465898751, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0018328676465898751, "signal/brier_reward/centered_abs_mean": 0.1051685631275177, "signal/brier_reward/group_std_mean": 0.13611198365688323, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.010516856797039509, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.010516856797039509, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013619576953351498, "signal/confidence_uniqueness_reward/group_std_mean": 0.01812830865383148, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013619576813653112, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013619576813653112, "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": 0.00037841796875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00037841796875, "signal/frontier_aurc_reward/centered_abs_mean": 0.002406639256514609, "signal/frontier_aurc_reward/group_std_mean": 0.003945078793913126, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.0082990087976212e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.0082990087976212e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.146014204621315, "signal/frontier_coverage_10/group_std_mean": 0.19040818512439728, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0018251775531098247, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0018251775531098247, "signal/frontier_coverage_15/centered_abs_mean": 0.13452889770269394, "signal/frontier_coverage_15/group_std_mean": 0.1758925199508667, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0016816111980006098, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0016816111980006098, "signal/frontier_coverage_20/centered_abs_mean": 0.08931544125080108, "signal/frontier_coverage_20/group_std_mean": 0.11687219142913818, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0011164430063217879, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0011164430063217879, "signal/frontier_coverage_25/centered_abs_mean": 0.057048802822828294, "signal/frontier_coverage_25/group_std_mean": 0.07306850105524063, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007131100399419665, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007131100399419665, "signal/frontier_ece_reward/centered_abs_mean": 0.005586130451411009, "signal/frontier_ece_reward/group_std_mean": 0.007241116929799318, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.000558613077737391, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.000558613077737391, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28740963339805603, "signal/frontier_entropy_batch_reward/group_std_mean": 0.36039870977401733, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.028740964084863662, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028740964084863662, "step": 260 }, { "calibration/aurc": 0.3173169574012934, "calibration/batch_distribution_entropy": 0.9774187970876994, "calibration/buffer_distribution_entropy": 0.9989279608428416, "calibration/confidence_entropy": 0.5001617869032355, "calibration/coverage@0%": 0.01251070205479452, "calibration/coverage@1%": 0.01251070205479452, "calibration/coverage@10%": 0.12639738258317026, "calibration/coverage@15%": 0.26002629647749514, "calibration/coverage@20%": 0.43505763820939336, "calibration/coverage@25%": 0.49251314823874753, "calibration/coverage@30%": 0.5327750428082192, "calibration/coverage@5%": 0.094311093444227, "calibration/ece": 0.16545728984750113, "calibration/mean_confidence": 0.5373981484322367, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 792.4, "completions/max_terminated_length": 792.4, "completions/mean_length": 209.72119140625, "completions/mean_terminated_length": 209.90668029785155, "completions/min_length": 0.0, "completions/min_terminated_length": 111.0, "epoch": 0.848, "grad_norm": 0.0014140016864985228, "learning_rate": 1e-06, "loss": -0.0005, "num_tokens": 903293484.0, "reward": 0.9379542469978333, "reward_std": 0.08483014702796936, "rewards/accuracy_reward": 0.52919921875, "rewards/batch_coverage_0": 0.3757105112075806, "rewards/batch_coverage_1": 0.3757105112075806, "rewards/batch_coverage_5": 0.38777480721473695, "rewards/brier_reward": 0.8000369906425476, "rewards/confidence_uniqueness_reward": 0.9522415280342102, "rewards/format_reward": 0.99912109375, "rewards/frontier_aurc_reward": -0.002984293410554528, "rewards/frontier_coverage_10": 0.11760776191949844, "rewards/frontier_coverage_15": 0.11307507008314133, "rewards/frontier_coverage_20": 0.07475161999464035, "rewards/frontier_coverage_25": 0.049993167817592624, "rewards/frontier_ece_reward": 0.0032593416515737774, "rewards/frontier_entropy_batch_reward": -0.2040516972541809, "signal/accuracy_reward/centered_abs_mean": 0.083062744140625, "signal/accuracy_reward/group_std_mean": 0.1137155219912529, "signal/accuracy_reward/group_zero_std_frac": 0.6625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0415313720703125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0415313720703125, "signal/advantage_abs_mean": 0.0636547863483429, "signal/advantage_pre_scale_abs_mean": 0.0636547863483429, "signal/advantage_pre_scale_std": 0.1063510000705719, "signal/advantage_std": 0.1063510000705719, "signal/batch_coverage_0/centered_abs_mean": 0.14138503670692443, "signal/batch_coverage_0/group_std_mean": 0.18060729801654815, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0017673130147159099, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0017673130147159099, "signal/batch_coverage_1/centered_abs_mean": 0.14138503670692443, "signal/batch_coverage_1/group_std_mean": 0.18060729801654815, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0017673130147159099, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0017673130147159099, "signal/batch_coverage_5/centered_abs_mean": 0.14474088847637176, "signal/batch_coverage_5/group_std_mean": 0.18487028777599335, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0018092612735927105, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0018092612735927105, "signal/brier_reward/centered_abs_mean": 0.10915502458810807, "signal/brier_reward/group_std_mean": 0.14279179871082306, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.010915502533316612, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.010915502533316612, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.0130588473752141, "signal/confidence_uniqueness_reward/group_std_mean": 0.018681207299232484, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.001305884774774313, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001305884774774313, "signal/format_reward/centered_abs_mean": 0.001690673828125, "signal/format_reward/group_std_mean": 0.004635536018759013, "signal/format_reward/group_zero_std_frac": 0.975, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0008453369140625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0008453369140625, "signal/frontier_aurc_reward/centered_abs_mean": 0.0026019166223704817, "signal/frontier_aurc_reward/group_std_mean": 0.004179058829322457, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.252395763411187e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.252395763411187e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.14312217235565186, "signal/frontier_coverage_10/group_std_mean": 0.18668304979801179, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.001789027126505971, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.001789027126505971, "signal/frontier_coverage_15/centered_abs_mean": 0.13520692884922028, "signal/frontier_coverage_15/group_std_mean": 0.1765005260705948, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0016900865593925118, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0016900865593925118, "signal/frontier_coverage_20/centered_abs_mean": 0.08786115050315857, "signal/frontier_coverage_20/group_std_mean": 0.1150710940361023, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.001098264381289482, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.001098264381289482, "signal/frontier_coverage_25/centered_abs_mean": 0.05405468940734863, "signal/frontier_coverage_25/group_std_mean": 0.06996125876903533, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0006756835966371, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0006756835966371, "signal/frontier_ece_reward/centered_abs_mean": 0.005427793879061937, "signal/frontier_ece_reward/group_std_mean": 0.0070043053478002545, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005427793716080487, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005427793716080487, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2727817714214325, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34347323179244993, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.027278177440166473, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027278177440166473, "step": 265 }, { "calibration/aurc": 0.2529951828346165, "calibration/batch_distribution_entropy": 0.9596141523779458, "calibration/buffer_distribution_entropy": 0.9989256467819031, "calibration/confidence_entropy": 0.49993889983064294, "calibration/coverage@0%": 0.02734910102739726, "calibration/coverage@1%": 0.02734910102739726, "calibration/coverage@10%": 0.14961472602739728, "calibration/coverage@15%": 0.26490566903131113, "calibration/coverage@20%": 0.36929886252446187, "calibration/coverage@25%": 0.43457528131115464, "calibration/coverage@30%": 0.6936070511252446, "calibration/coverage@5%": 0.09531785102739726, "calibration/ece": 0.1295995513550006, "calibration/mean_confidence": 0.6024682494366631, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 914.8, "completions/max_terminated_length": 914.8, "completions/mean_length": 208.97802734375, "completions/mean_terminated_length": 209.14348754882812, "completions/min_length": 42.6, "completions/min_terminated_length": 110.8, "epoch": 0.864, "grad_norm": 0.0010036705061793327, "learning_rate": 1e-06, "loss": -0.0004, "num_tokens": 920420235.0, "reward": 0.9642098307609558, "reward_std": 0.08568636775016784, "rewards/accuracy_reward": 0.59228515625, "rewards/batch_coverage_0": 0.37067938446998594, "rewards/batch_coverage_1": 0.37067938446998594, "rewards/batch_coverage_5": 0.3885198771953583, "rewards/brier_reward": 0.7951380968093872, "rewards/confidence_uniqueness_reward": 0.9504954934120178, "rewards/format_reward": 0.99921875, "rewards/frontier_aurc_reward": -0.0024885230232030153, "rewards/frontier_coverage_10": 0.07284789383411408, "rewards/frontier_coverage_15": 0.07132325246930123, "rewards/frontier_coverage_20": 0.05422785058617592, "rewards/frontier_coverage_25": 0.04602581784129143, "rewards/frontier_ece_reward": 0.002689109672792256, "rewards/frontier_entropy_batch_reward": -0.23522092700004577, "signal/accuracy_reward/centered_abs_mean": 0.090802001953125, "signal/accuracy_reward/group_std_mean": 0.12010103911161422, "signal/accuracy_reward/group_zero_std_frac": 0.659375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0454010009765625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0454010009765625, "signal/advantage_abs_mean": 0.06633162274956703, "signal/advantage_pre_scale_abs_mean": 0.06633162274956703, "signal/advantage_pre_scale_std": 0.10634591728448868, "signal/advantage_std": 0.10634591728448868, "signal/batch_coverage_0/centered_abs_mean": 0.15131303966045379, "signal/batch_coverage_0/group_std_mean": 0.19186782836914062, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0018914130749180913, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0018914130749180913, "signal/batch_coverage_1/centered_abs_mean": 0.15131303966045379, "signal/batch_coverage_1/group_std_mean": 0.19186782836914062, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0018914130749180913, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0018914130749180913, "signal/batch_coverage_5/centered_abs_mean": 0.15699659883975983, "signal/batch_coverage_5/group_std_mean": 0.19916775226593017, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0019624575041234495, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0019624575041234495, "signal/brier_reward/centered_abs_mean": 0.11428913176059723, "signal/brier_reward/group_std_mean": 0.14671537280082703, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011428913474082947, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011428913474082947, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013941565901041031, "signal/confidence_uniqueness_reward/group_std_mean": 0.0192941814661026, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013941565994173288, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013941565994173288, "signal/format_reward/centered_abs_mean": 0.0014892578125, "signal/format_reward/group_std_mean": 0.0037468004506081344, "signal/format_reward/group_zero_std_frac": 0.98125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00074462890625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00074462890625, "signal/frontier_aurc_reward/centered_abs_mean": 0.0023261834401637317, "signal/frontier_aurc_reward/group_std_mean": 0.0038332717958837748, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.9077293220325373e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.9077293220325373e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.15233877301216125, "signal/frontier_coverage_10/group_std_mean": 0.1952694833278656, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.00190423468593508, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.00190423468593508, "signal/frontier_coverage_15/centered_abs_mean": 0.14285752773284913, "signal/frontier_coverage_15/group_std_mean": 0.1831321030855179, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.001785719092004001, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.001785719092004001, "signal/frontier_coverage_20/centered_abs_mean": 0.09027667939662934, "signal/frontier_coverage_20/group_std_mean": 0.11626434624195099, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0011284585110843182, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0011284585110843182, "signal/frontier_coverage_25/centered_abs_mean": 0.05739191547036171, "signal/frontier_coverage_25/group_std_mean": 0.07379192858934402, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007173989550210535, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007173989550210535, "signal/frontier_ece_reward/centered_abs_mean": 0.005358308367431164, "signal/frontier_ece_reward/group_std_mean": 0.006918957177549601, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005358308553695679, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005358308553695679, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28787853121757506, "signal/frontier_entropy_batch_reward/group_std_mean": 0.35972235202789304, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.028787852451205253, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028787852451205253, "step": 270 }, { "calibration/aurc": 0.37163142562680884, "calibration/batch_distribution_entropy": 0.980721725063931, "calibration/buffer_distribution_entropy": 0.9987390897543371, "calibration/confidence_entropy": 0.48023975904051514, "calibration/coverage@0%": 0.01719055772994129, "calibration/coverage@1%": 0.01719055772994129, "calibration/coverage@10%": 0.07308586105675147, "calibration/coverage@15%": 0.10629586594911937, "calibration/coverage@20%": 0.1477097602739726, "calibration/coverage@25%": 0.1973466548434442, "calibration/coverage@30%": 0.2896503485812133, "calibration/coverage@5%": 0.02971119740704501, "calibration/ece": 0.13803887313861454, "calibration/mean_confidence": 0.49456336654610533, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 590.6, "completions/max_terminated_length": 590.6, "completions/mean_length": 204.8078125, "completions/mean_terminated_length": 204.92990112304688, "completions/min_length": 20.0, "completions/min_terminated_length": 108.8, "epoch": 0.88, "grad_norm": 0.0009604271035641432, "learning_rate": 1e-06, "loss": -0.0004, "num_tokens": 937664539.0, "reward": 0.9294121503829956, "reward_std": 0.0859591543674469, "rewards/accuracy_reward": 0.5115234375, "rewards/batch_coverage_0": 0.37883342504501344, "rewards/batch_coverage_1": 0.37883342504501344, "rewards/batch_coverage_5": 0.40217695832252504, "rewards/brier_reward": 0.7976484060287475, "rewards/confidence_uniqueness_reward": 0.9519205927848816, "rewards/format_reward": 0.9994140625, "rewards/frontier_aurc_reward": -0.0026464010123163463, "rewards/frontier_coverage_10": 0.1320437639951706, "rewards/frontier_coverage_15": 0.12262350916862488, "rewards/frontier_coverage_20": 0.07881927117705345, "rewards/frontier_coverage_25": 0.05345905870199204, "rewards/frontier_ece_reward": 0.003383675776422024, "rewards/frontier_entropy_batch_reward": -0.20653626024723054, "signal/accuracy_reward/centered_abs_mean": 0.09537353515625, "signal/accuracy_reward/group_std_mean": 0.12430963218212128, "signal/accuracy_reward/group_zero_std_frac": 0.646875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.047686767578125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.047686767578125, "signal/advantage_abs_mean": 0.0670087181031704, "signal/advantage_pre_scale_abs_mean": 0.0670087181031704, "signal/advantage_pre_scale_std": 0.10862193554639817, "signal/advantage_std": 0.10862193554639817, "signal/batch_coverage_0/centered_abs_mean": 0.14668529927730561, "signal/batch_coverage_0/group_std_mean": 0.18786006569862365, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0018335663713514804, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0018335663713514804, "signal/batch_coverage_1/centered_abs_mean": 0.14668529927730561, "signal/batch_coverage_1/group_std_mean": 0.18786006569862365, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0018335663713514804, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0018335663713514804, "signal/batch_coverage_5/centered_abs_mean": 0.15227998197078704, "signal/batch_coverage_5/group_std_mean": 0.1952932894229889, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.001903499779291451, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.001903499779291451, "signal/brier_reward/centered_abs_mean": 0.11385080218315125, "signal/brier_reward/group_std_mean": 0.14707624912261963, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.01138508040457964, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.01138508040457964, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.012812569923698901, "signal/confidence_uniqueness_reward/group_std_mean": 0.017841225117444993, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0012812570203095675, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012812570203095675, "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": 0.000567626953125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.000567626953125, "signal/frontier_aurc_reward/centered_abs_mean": 0.002319264691323042, "signal/frontier_aurc_reward/group_std_mean": 0.003852441068738699, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.8990809369133785e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.8990809369133785e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.15802746415138244, "signal/frontier_coverage_10/group_std_mean": 0.2027164340019226, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0019753432832658293, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0019753432832658293, "signal/frontier_coverage_15/centered_abs_mean": 0.1465783804655075, "signal/frontier_coverage_15/group_std_mean": 0.1882282167673111, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0018322297371923923, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0018322297371923923, "signal/frontier_coverage_20/centered_abs_mean": 0.09299429208040237, "signal/frontier_coverage_20/group_std_mean": 0.11972088217735291, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0011624286649748682, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0011624286649748682, "signal/frontier_coverage_25/centered_abs_mean": 0.05760718137025833, "signal/frontier_coverage_25/group_std_mean": 0.07376070767641067, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007200897671282291, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007200897671282291, "signal/frontier_ece_reward/centered_abs_mean": 0.005225432850420475, "signal/frontier_ece_reward/group_std_mean": 0.006732543837279082, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005225432571023703, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005225432571023703, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26823015213012696, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34410834312438965, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02682301476597786, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02682301476597786, "step": 275 }, { "calibration/aurc": 0.35420580524448686, "calibration/batch_distribution_entropy": 0.9825957823228452, "calibration/buffer_distribution_entropy": 0.9986545666292408, "calibration/confidence_entropy": 0.4869230028182316, "calibration/coverage@0%": 0.009776327054794521, "calibration/coverage@1%": 0.009776327054794521, "calibration/coverage@10%": 0.032463154354207435, "calibration/coverage@15%": 0.12470187133072405, "calibration/coverage@20%": 0.18253883317025438, "calibration/coverage@25%": 0.35604130993150684, "calibration/coverage@30%": 0.47016955112524456, "calibration/coverage@5%": 0.026203216731898238, "calibration/ece": 0.15174858499089836, "calibration/mean_confidence": 0.5042888624193199, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 541.0, "completions/max_terminated_length": 541.0, "completions/mean_length": 204.1130859375, "completions/mean_terminated_length": 204.21332092285155, "completions/min_length": 42.4, "completions/min_terminated_length": 105.0, "epoch": 0.896, "grad_norm": 0.0014212472597137094, "learning_rate": 1e-06, "loss": -0.0003, "num_tokens": 954865505.0, "reward": 0.9436590671539307, "reward_std": 0.07857488244771957, "rewards/accuracy_reward": 0.54228515625, "rewards/batch_coverage_0": 0.3803312540054321, "rewards/batch_coverage_1": 0.3803312540054321, "rewards/batch_coverage_5": 0.398857969045639, "rewards/brier_reward": 0.7943593502044678, "rewards/confidence_uniqueness_reward": 0.9523796916007996, "rewards/format_reward": 0.99951171875, "rewards/frontier_aurc_reward": -0.0028298380551859737, "rewards/frontier_coverage_10": 0.11004131585359574, "rewards/frontier_coverage_15": 0.10444882810115814, "rewards/frontier_coverage_20": 0.07032015025615693, "rewards/frontier_coverage_25": 0.05014690682291985, "rewards/frontier_ece_reward": 0.0024532068520784376, "rewards/frontier_entropy_batch_reward": -0.20804197788238527, "signal/accuracy_reward/centered_abs_mean": 0.080194091796875, "signal/accuracy_reward/group_std_mean": 0.10977384746074677, "signal/accuracy_reward/group_zero_std_frac": 0.665625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0400970458984375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0400970458984375, "signal/advantage_abs_mean": 0.05979133695363999, "signal/advantage_pre_scale_abs_mean": 0.05979133695363999, "signal/advantage_pre_scale_std": 0.09781416058540345, "signal/advantage_std": 0.09781416058540345, "signal/batch_coverage_0/centered_abs_mean": 0.1435524821281433, "signal/batch_coverage_0/group_std_mean": 0.1840268313884735, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0017944060498848557, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0017944060498848557, "signal/batch_coverage_1/centered_abs_mean": 0.1435524821281433, "signal/batch_coverage_1/group_std_mean": 0.1840268313884735, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0017944060498848557, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0017944060498848557, "signal/batch_coverage_5/centered_abs_mean": 0.14894903898239137, "signal/batch_coverage_5/group_std_mean": 0.19109267592430115, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0018618630012497306, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0018618630012497306, "signal/brier_reward/centered_abs_mean": 0.1110924944281578, "signal/brier_reward/group_std_mean": 0.14494978189468383, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011109249293804168, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011109249293804168, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.01262026559561491, "signal/confidence_uniqueness_reward/group_std_mean": 0.0172889219596982, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0012620265362784267, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012620265362784267, "signal/format_reward/centered_abs_mean": 0.000946044921875, "signal/format_reward/group_std_mean": 0.002762135770171881, "signal/format_reward/group_zero_std_frac": 0.984375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0004730224609375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0004730224609375, "signal/frontier_aurc_reward/centered_abs_mean": 0.002343117049895227, "signal/frontier_aurc_reward/group_std_mean": 0.003864579740911722, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.9288963560247794e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.9288963560247794e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.15725082755088807, "signal/frontier_coverage_10/group_std_mean": 0.20240573585033417, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0019656353862956165, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0019656353862956165, "signal/frontier_coverage_15/centered_abs_mean": 0.14419232904911042, "signal/frontier_coverage_15/group_std_mean": 0.18539913594722748, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0018024041084572673, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0018024041084572673, "signal/frontier_coverage_20/centered_abs_mean": 0.09211180359125137, "signal/frontier_coverage_20/group_std_mean": 0.11852122098207474, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0011513975681737066, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0011513975681737066, "signal/frontier_coverage_25/centered_abs_mean": 0.05666951462626457, "signal/frontier_coverage_25/group_std_mean": 0.07293299734592437, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007083689561113715, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007083689561113715, "signal/frontier_ece_reward/centered_abs_mean": 0.005120085831731558, "signal/frontier_ece_reward/group_std_mean": 0.006608256697654724, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005120086017996073, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005120086017996073, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2719715416431427, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34647042155265806, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.027197154238820076, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027197154238820076, "step": 280 }, { "calibration/aurc": 0.3638614821604429, "calibration/batch_distribution_entropy": 0.9738889334751939, "calibration/buffer_distribution_entropy": 0.9985142121400192, "calibration/confidence_entropy": 0.48338144297106495, "calibration/coverage@0%": 0.016015625, "calibration/coverage@1%": 0.016015625, "calibration/coverage@10%": 0.09453125, "calibration/coverage@15%": 0.160546875, "calibration/coverage@20%": 0.303515625, "calibration/coverage@25%": 0.392578125, "calibration/coverage@30%": 0.44921875, "calibration/coverage@5%": 0.02734375, "calibration/ece": 0.15531850117505117, "calibration/mean_confidence": 0.5144720905998142, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 708.2, "completions/max_terminated_length": 708.2, "completions/mean_length": 208.47900390625, "completions/mean_terminated_length": 208.60187377929688, "completions/min_length": 23.4, "completions/min_terminated_length": 112.2, "epoch": 0.912, "grad_norm": 0.0016397674335166812, "learning_rate": 1e-06, "loss": -0.0007, "num_tokens": 972051626.0, "reward": 0.9432343602180481, "reward_std": 0.08461738377809525, "rewards/accuracy_reward": 0.5439453125, "rewards/batch_coverage_0": 0.3605897545814514, "rewards/batch_coverage_1": 0.3605897545814514, "rewards/batch_coverage_5": 0.3883225917816162, "rewards/brier_reward": 0.7949118971824646, "rewards/confidence_uniqueness_reward": 0.9518064975738525, "rewards/format_reward": 0.9994140625, "rewards/frontier_aurc_reward": -0.0029322515008971097, "rewards/frontier_coverage_10": 0.10165179260075093, "rewards/frontier_coverage_15": 0.09448765683919191, "rewards/frontier_coverage_20": 0.06744903400540352, "rewards/frontier_coverage_25": 0.051872897893190384, "rewards/frontier_ece_reward": 0.002977025369182229, "rewards/frontier_entropy_batch_reward": -0.211902517080307, "signal/accuracy_reward/centered_abs_mean": 0.086962890625, "signal/accuracy_reward/group_std_mean": 0.12032638341188431, "signal/accuracy_reward/group_zero_std_frac": 0.6375, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0434814453125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0434814453125, "signal/advantage_abs_mean": 0.06350576281547546, "signal/advantage_pre_scale_abs_mean": 0.06350576281547546, "signal/advantage_pre_scale_std": 0.10434354841709137, "signal/advantage_std": 0.10434354841709137, "signal/batch_coverage_0/centered_abs_mean": 0.15041546821594237, "signal/batch_coverage_0/group_std_mean": 0.19042237102985382, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0018801933620125055, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0018801933620125055, "signal/batch_coverage_1/centered_abs_mean": 0.15041546821594237, "signal/batch_coverage_1/group_std_mean": 0.19042237102985382, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0018801933620125055, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0018801933620125055, "signal/batch_coverage_5/centered_abs_mean": 0.15842511355876923, "signal/batch_coverage_5/group_std_mean": 0.20219461023807525, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.001980313891544938, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.001980313891544938, "signal/brier_reward/centered_abs_mean": 0.11727479547262191, "signal/brier_reward/group_std_mean": 0.1518254667520523, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011727479845285415, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011727479845285415, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.012682436965405941, "signal/confidence_uniqueness_reward/group_std_mean": 0.017683419212698937, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0012682436732575296, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012682436732575296, "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": 0.000567626953125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.000567626953125, "signal/frontier_aurc_reward/centered_abs_mean": 0.002684690454043448, "signal/frontier_aurc_reward/group_std_mean": 0.0043406285345554355, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.355863264005165e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.355863264005165e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1571253329515457, "signal/frontier_coverage_10/group_std_mean": 0.2012400895357132, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0019640665967017414, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0019640665967017414, "signal/frontier_coverage_15/centered_abs_mean": 0.14200886487960815, "signal/frontier_coverage_15/group_std_mean": 0.1820073276758194, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0017751108156517149, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0017751108156517149, "signal/frontier_coverage_20/centered_abs_mean": 0.09170998632907867, "signal/frontier_coverage_20/group_std_mean": 0.11796575039625168, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0011463748523965478, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0011463748523965478, "signal/frontier_coverage_25/centered_abs_mean": 0.05969014316797257, "signal/frontier_coverage_25/group_std_mean": 0.0763975977897644, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007461267872713506, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007461267872713506, "signal/frontier_ece_reward/centered_abs_mean": 0.005277018249034882, "signal/frontier_ece_reward/group_std_mean": 0.006764029897749424, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.000527701823739335, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.000527701823739335, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.2682758390903473, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3394309163093567, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.026827583834528924, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.026827583834528924, "step": 285 }, { "calibration/aurc": 0.4198958428350924, "calibration/batch_distribution_entropy": 0.9854211341398911, "calibration/buffer_distribution_entropy": 0.9986680725885083, "calibration/confidence_entropy": 0.5089633649066212, "calibration/coverage@0%": 0.00783316965484824, "calibration/coverage@1%": 0.00783316965484824, "calibration/coverage@10%": 0.009009640243083535, "calibration/coverage@15%": 0.025466426724799505, "calibration/coverage@20%": 0.02781400488277503, "calibration/coverage@25%": 0.19330075352058634, "calibration/coverage@30%": 0.25749106363435786, "calibration/coverage@5%": 0.00783316965484824, "calibration/ece": 0.12966488647715124, "calibration/mean_confidence": 0.48708973508204867, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 605.6, "completions/max_terminated_length": 605.6, "completions/mean_length": 209.16123046875, "completions/mean_terminated_length": 209.28462829589844, "completions/min_length": 44.0, "completions/min_terminated_length": 95.6, "epoch": 0.928, "grad_norm": 0.0007559643127024174, "learning_rate": 1e-06, "loss": -0.0004, "num_tokens": 989220253.0, "reward": 0.9345414400100708, "reward_std": 0.081772480905056, "rewards/accuracy_reward": 0.52763671875, "rewards/batch_coverage_0": 0.3691330850124359, "rewards/batch_coverage_1": 0.3691330850124359, "rewards/batch_coverage_5": 0.38621372580528257, "rewards/brier_reward": 0.7843051671981811, "rewards/confidence_uniqueness_reward": 0.9518962979316712, "rewards/format_reward": 0.99931640625, "rewards/frontier_aurc_reward": -0.0029707553796470164, "rewards/frontier_coverage_10": 0.10881325751543044, "rewards/frontier_coverage_15": 0.10206395983695984, "rewards/frontier_coverage_20": 0.0704283982515335, "rewards/frontier_coverage_25": 0.05073222741484642, "rewards/frontier_ece_reward": 0.002549092145636678, "rewards/frontier_entropy_batch_reward": -0.2097952753305435, "signal/accuracy_reward/centered_abs_mean": 0.083795166015625, "signal/accuracy_reward/group_std_mean": 0.11404231339693069, "signal/accuracy_reward/group_zero_std_frac": 0.65625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0418975830078125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0418975830078125, "signal/advantage_abs_mean": 0.06221255958080292, "signal/advantage_pre_scale_abs_mean": 0.06221255958080292, "signal/advantage_pre_scale_std": 0.10113046169281006, "signal/advantage_std": 0.10113046169281006, "signal/batch_coverage_0/centered_abs_mean": 0.15219251811504364, "signal/batch_coverage_0/group_std_mean": 0.19208297729492188, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0019024065230041742, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0019024065230041742, "signal/batch_coverage_1/centered_abs_mean": 0.15219251811504364, "signal/batch_coverage_1/group_std_mean": 0.19208297729492188, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0019024065230041742, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0019024065230041742, "signal/batch_coverage_5/centered_abs_mean": 0.1581542283296585, "signal/batch_coverage_5/group_std_mean": 0.19989595115184783, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0019769279286265375, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0019769279286265375, "signal/brier_reward/centered_abs_mean": 0.11704501807689667, "signal/brier_reward/group_std_mean": 0.1495725303888321, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011704502068459988, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011704502068459988, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013593058660626411, "signal/confidence_uniqueness_reward/group_std_mean": 0.019005635380744935, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013593058800324797, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013593058800324797, "signal/format_reward/centered_abs_mean": 0.001324462890625, "signal/format_reward/group_std_mean": 0.003866990143433213, "signal/format_reward/group_zero_std_frac": 0.978125, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0006622314453125, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0006622314453125, "signal/frontier_aurc_reward/centered_abs_mean": 0.0025076627265661956, "signal/frontier_aurc_reward/group_std_mean": 0.0039647513069212435, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.134578437311575e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.134578437311575e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.16001522541046143, "signal/frontier_coverage_10/group_std_mean": 0.20408071279525758, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0020001903641968964, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0020001903641968964, "signal/frontier_coverage_15/centered_abs_mean": 0.1480545699596405, "signal/frontier_coverage_15/group_std_mean": 0.1891225755214691, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.001850682171061635, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.001850682171061635, "signal/frontier_coverage_20/centered_abs_mean": 0.09294760525226593, "signal/frontier_coverage_20/group_std_mean": 0.1197780579328537, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0011618450749665498, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0011618450749665498, "signal/frontier_coverage_25/centered_abs_mean": 0.05976380407810211, "signal/frontier_coverage_25/group_std_mean": 0.07655199617147446, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007470475626178086, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007470475626178086, "signal/frontier_ece_reward/centered_abs_mean": 0.005129948537796736, "signal/frontier_ece_reward/group_std_mean": 0.006620830483734607, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005129948898684233, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005129948898684233, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.27609426975250245, "signal/frontier_entropy_batch_reward/group_std_mean": 0.35209437608718874, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02760942615568638, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02760942615568638, "step": 290 }, { "calibration/aurc": 0.25085033565264336, "calibration/batch_distribution_entropy": 0.9842469931771731, "calibration/buffer_distribution_entropy": 0.9988941903741084, "calibration/confidence_entropy": 0.5025199505601536, "calibration/coverage@0%": 0.050390625, "calibration/coverage@1%": 0.050390625, "calibration/coverage@10%": 0.1921875, "calibration/coverage@15%": 0.31875, "calibration/coverage@20%": 0.455859375, "calibration/coverage@25%": 0.564453125, "calibration/coverage@30%": 0.654296875, "calibration/coverage@5%": 0.1046875, "calibration/ece": 0.0978781378944602, "calibration/mean_confidence": 0.4690214396482693, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 530.6, "completions/max_terminated_length": 530.6, "completions/mean_length": 215.825, "completions/mean_terminated_length": 215.90928344726564, "completions/min_length": 23.4, "completions/min_terminated_length": 111.8, "epoch": 0.944, "grad_norm": 0.0008739962941035628, "learning_rate": 1e-06, "loss": -0.0002, "num_tokens": 1006405725.0, "reward": 0.9356146097183228, "reward_std": 0.08910652995109558, "rewards/accuracy_reward": 0.5318359375, "rewards/batch_coverage_0": 0.3425533711910248, "rewards/batch_coverage_1": 0.3425533711910248, "rewards/batch_coverage_5": 0.36451648473739623, "rewards/brier_reward": 0.786684799194336, "rewards/confidence_uniqueness_reward": 0.9519042730331421, "rewards/format_reward": 0.999609375, "rewards/frontier_aurc_reward": -0.0028218948049470783, "rewards/frontier_coverage_10": 0.1091730386018753, "rewards/frontier_coverage_15": 0.09916338846087455, "rewards/frontier_coverage_20": 0.07232377529144288, "rewards/frontier_coverage_25": 0.049502771347761154, "rewards/frontier_ece_reward": 0.0025140484096482394, "rewards/frontier_entropy_batch_reward": -0.21430420279502868, "signal/accuracy_reward/centered_abs_mean": 0.1032958984375, "signal/accuracy_reward/group_std_mean": 0.1377536654472351, "signal/accuracy_reward/group_zero_std_frac": 0.596875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.05164794921875, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.05164794921875, "signal/advantage_abs_mean": 0.06843964532017707, "signal/advantage_pre_scale_abs_mean": 0.06843964532017707, "signal/advantage_pre_scale_std": 0.11038383245468139, "signal/advantage_std": 0.11038383245468139, "signal/batch_coverage_0/centered_abs_mean": 0.14654119908809662, "signal/batch_coverage_0/group_std_mean": 0.18539315164089204, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0018317650305107237, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0018317650305107237, "signal/batch_coverage_1/centered_abs_mean": 0.14654119908809662, "signal/batch_coverage_1/group_std_mean": 0.18539315164089204, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0018317650305107237, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0018317650305107237, "signal/batch_coverage_5/centered_abs_mean": 0.15288062393665314, "signal/batch_coverage_5/group_std_mean": 0.19395535588264465, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0019110078224912285, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0019110078224912285, "signal/brier_reward/centered_abs_mean": 0.11119691431522369, "signal/brier_reward/group_std_mean": 0.14237308800220488, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011119692213833332, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011119692213833332, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.012922956421971322, "signal/confidence_uniqueness_reward/group_std_mean": 0.01735894661396742, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0012922955676913262, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012922955676913262, "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": 0.00037841796875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00037841796875, "signal/frontier_aurc_reward/centered_abs_mean": 0.002141226688399911, "signal/frontier_aurc_reward/group_std_mean": 0.0033662895672023295, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.676533222256694e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.676533222256694e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.16631161868572236, "signal/frontier_coverage_10/group_std_mean": 0.2133659988641739, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0020788952242583036, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0020788952242583036, "signal/frontier_coverage_15/centered_abs_mean": 0.15309363901615142, "signal/frontier_coverage_15/group_std_mean": 0.19652132093906402, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0019136705435812474, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0019136705435812474, "signal/frontier_coverage_20/centered_abs_mean": 0.09737784415483475, "signal/frontier_coverage_20/group_std_mean": 0.12504442036151886, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0012172231217846274, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0012172231217846274, "signal/frontier_coverage_25/centered_abs_mean": 0.05944311022758484, "signal/frontier_coverage_25/group_std_mean": 0.07615808546543121, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007430389057844877, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007430389057844877, "signal/frontier_ece_reward/centered_abs_mean": 0.005109536647796631, "signal/frontier_ece_reward/group_std_mean": 0.006575622782111168, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005109536694362759, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005109536694362759, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.275804203748703, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3487266957759857, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.027580421417951584, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.027580421417951584, "step": 295 }, { "calibration/aurc": 0.3229438581171732, "calibration/batch_distribution_entropy": 0.9770622901786709, "calibration/buffer_distribution_entropy": 0.9989475163774502, "calibration/confidence_entropy": 0.46726227972999607, "calibration/coverage@0%": 0.019921875, "calibration/coverage@1%": 0.019921875, "calibration/coverage@10%": 0.1234375, "calibration/coverage@15%": 0.240234375, "calibration/coverage@20%": 0.303515625, "calibration/coverage@25%": 0.372265625, "calibration/coverage@30%": 0.428125, "calibration/coverage@5%": 0.0265625, "calibration/ece": 0.14295204049453908, "calibration/mean_confidence": 0.5535608828710752, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0009765625, "completions/max_length": 919.2, "completions/max_terminated_length": 919.2, "completions/mean_length": 222.2328125, "completions/mean_terminated_length": 222.45050659179688, "completions/min_length": 0.0, "completions/min_terminated_length": 117.0, "epoch": 0.96, "grad_norm": 0.000724839570466429, "learning_rate": 1e-06, "loss": -0.0006, "num_tokens": 1023621709.0, "reward": 0.9392164587974549, "reward_std": 0.07907238602638245, "rewards/accuracy_reward": 0.532421875, "rewards/batch_coverage_0": 0.3975462198257446, "rewards/batch_coverage_1": 0.3975462198257446, "rewards/batch_coverage_5": 0.41426302790641784, "rewards/brier_reward": 0.8033321619033813, "rewards/confidence_uniqueness_reward": 0.9501454949378967, "rewards/format_reward": 0.9990234375, "rewards/frontier_aurc_reward": -0.0034745996817946434, "rewards/frontier_coverage_10": 0.1234698086977005, "rewards/frontier_coverage_15": 0.11319096833467483, "rewards/frontier_coverage_20": 0.07962790504097939, "rewards/frontier_coverage_25": 0.057156264781951904, "rewards/frontier_ece_reward": 0.003401698311790824, "rewards/frontier_entropy_batch_reward": -0.21935696601867677, "signal/accuracy_reward/centered_abs_mean": 0.0761962890625, "signal/accuracy_reward/group_std_mean": 0.10855842977762223, "signal/accuracy_reward/group_zero_std_frac": 0.6625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.03809814453125, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.03809814453125, "signal/advantage_abs_mean": 0.05825046822428703, "signal/advantage_pre_scale_abs_mean": 0.05825046822428703, "signal/advantage_pre_scale_std": 0.09837974160909653, "signal/advantage_std": 0.09837974160909653, "signal/batch_coverage_0/centered_abs_mean": 0.14885650277137757, "signal/batch_coverage_0/group_std_mean": 0.19004371464252473, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.0018607063684612513, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.0018607063684612513, "signal/batch_coverage_1/centered_abs_mean": 0.14885650277137757, "signal/batch_coverage_1/group_std_mean": 0.19004371464252473, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.0018607063684612513, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.0018607063684612513, "signal/batch_coverage_5/centered_abs_mean": 0.15346541702747346, "signal/batch_coverage_5/group_std_mean": 0.19588360488414763, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.001918317750096321, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.001918317750096321, "signal/brier_reward/centered_abs_mean": 0.10768177956342698, "signal/brier_reward/group_std_mean": 0.14222208857536317, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.010768178105354308, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.010768178105354308, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.01446004156023264, "signal/confidence_uniqueness_reward/group_std_mean": 0.020395297929644583, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0014460041420534253, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0014460041420534253, "signal/format_reward/centered_abs_mean": 0.00186767578125, "signal/format_reward/group_std_mean": 0.004851654730737209, "signal/format_reward/group_zero_std_frac": 0.975, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.000933837890625, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.000933837890625, "signal/frontier_aurc_reward/centered_abs_mean": 0.0029538881964981554, "signal/frontier_aurc_reward/group_std_mean": 0.004771159123629331, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.6923602601746096e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.6923602601746096e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.1441126823425293, "signal/frontier_coverage_10/group_std_mean": 0.18875311613082885, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0018014084780588746, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0018014084780588746, "signal/frontier_coverage_15/centered_abs_mean": 0.13292936384677886, "signal/frontier_coverage_15/group_std_mean": 0.17427212595939637, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0016616170294582843, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0016616170294582843, "signal/frontier_coverage_20/centered_abs_mean": 0.08532071262598037, "signal/frontier_coverage_20/group_std_mean": 0.11228447258472443, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0010665089124813677, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0010665089124813677, "signal/frontier_coverage_25/centered_abs_mean": 0.0570922888815403, "signal/frontier_coverage_25/group_std_mean": 0.07393175810575485, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0007136536296457052, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0007136536296457052, "signal/frontier_ece_reward/centered_abs_mean": 0.005352666322141886, "signal/frontier_ece_reward/group_std_mean": 0.006874622870236635, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005352666368708015, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005352666368708015, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.26995482444763186, "signal/frontier_entropy_batch_reward/group_std_mean": 0.34402047991752627, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02699548415839672, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02699548415839672, "step": 300 }, { "epoch": 0.96, "eval_calibration/aurc": 0.4261856012734042, "eval_calibration/batch_distribution_entropy": 0.9150904952300255, "eval_calibration/buffer_distribution_entropy": 0.9987117008124493, "eval_calibration/confidence_entropy": 0.4750955585082007, "eval_calibration/coverage@0%": 0.046875, "eval_calibration/coverage@1%": 0.046875, "eval_calibration/coverage@10%": 0.046875, "eval_calibration/coverage@15%": 0.203125, "eval_calibration/coverage@20%": 0.2265625, "eval_calibration/coverage@25%": 0.25, "eval_calibration/coverage@30%": 0.328125, "eval_calibration/coverage@5%": 0.046875, "eval_calibration/ece": 0.20744989956221405, "eval_calibration/mean_confidence": 0.5127207124231831, "eval_completions/clipped_ratio": 0.001953125, "eval_completions/max_length": 377.25, "eval_completions/max_terminated_length": 377.25, "eval_completions/mean_length": 224.55205917358398, "eval_completions/mean_terminated_length": 224.9837760925293, "eval_completions/min_length": 98.0, "eval_completions/min_terminated_length": 131.75, "eval_loss": 0.0, "eval_num_tokens": 1023621709.0, "eval_reward": 0.7943826168775558, "eval_reward_std": 0.23918436467647552, "eval_rewards/accuracy_reward": 0.431640625, "eval_rewards/batch_coverage_0": 0.14696063846349716, "eval_rewards/batch_coverage_1": 0.14696063846349716, "eval_rewards/batch_coverage_5": 0.14696063846349716, "eval_rewards/brier_reward": 0.7773311734199524, "eval_rewards/confidence_uniqueness_reward": 0.8986344039440155, "eval_rewards/format_reward": 0.998046875, "eval_rewards/frontier_aurc_reward": -0.004537686298135668, "eval_rewards/frontier_coverage_10": 0.16706368699669838, "eval_rewards/frontier_coverage_15": 0.15455888211727142, "eval_rewards/frontier_coverage_20": 0.099946154281497, "eval_rewards/frontier_coverage_25": 0.0556830707937479, "eval_rewards/frontier_ece_reward": 0.0032707307836972177, "eval_rewards/frontier_entropy_batch_reward": -0.998046875, "eval_runtime": 28.798, "eval_samples_per_second": 17.362, "eval_signal/accuracy_reward/centered_abs_mean": 0.4781494140625, "eval_signal/accuracy_reward/group_std_mean": 0.49658260494470596, "eval_signal/accuracy_reward/group_zero_std_frac": 0.0, "eval_signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.23907470703125, "eval_signal/accuracy_reward/weight": 0.5, "eval_signal/accuracy_reward/weighted_centered_abs_mean": 0.23907470703125, "eval_signal/advantage_abs_mean": 0.2249186635017395, "eval_signal/advantage_pre_scale_abs_mean": 0.2249186635017395, "eval_signal/advantage_pre_scale_std": 0.23667898401618004, "eval_signal/advantage_std": 0.23667898401618004, "eval_signal/batch_coverage_0/centered_abs_mean": 0.29229119420051575, "eval_signal/batch_coverage_0/group_std_mean": 0.3650955706834793, "eval_signal/batch_coverage_0/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.003653639927506447, "eval_signal/batch_coverage_0/weight": 0.012500000186264515, "eval_signal/batch_coverage_0/weighted_centered_abs_mean": 0.003653639927506447, "eval_signal/batch_coverage_1/centered_abs_mean": 0.29229119420051575, "eval_signal/batch_coverage_1/group_std_mean": 0.3650955706834793, "eval_signal/batch_coverage_1/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.003653639927506447, "eval_signal/batch_coverage_1/weight": 0.012500000186264515, "eval_signal/batch_coverage_1/weighted_centered_abs_mean": 0.003653639927506447, "eval_signal/batch_coverage_5/centered_abs_mean": 0.29229119420051575, "eval_signal/batch_coverage_5/group_std_mean": 0.3650955706834793, "eval_signal/batch_coverage_5/group_zero_std_frac": 0.0, "eval_signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.003653639927506447, "eval_signal/batch_coverage_5/weight": 0.012500000186264515, "eval_signal/batch_coverage_5/weighted_centered_abs_mean": 0.003653639927506447, "eval_signal/brier_reward/centered_abs_mean": 0.2109338641166687, "eval_signal/brier_reward/group_std_mean": 0.2596461661159992, "eval_signal/brier_reward/group_zero_std_frac": 0.0, "eval_signal/brier_reward/scaled_weighted_centered_abs_mean": 0.02109338715672493, "eval_signal/brier_reward/weight": 0.10000000149011612, "eval_signal/brier_reward/weighted_centered_abs_mean": 0.02109338715672493, "eval_signal/confidence_uniqueness_reward/centered_abs_mean": 0.04196511581540108, "eval_signal/confidence_uniqueness_reward/group_std_mean": 0.05477655213326216, "eval_signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "eval_signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.004196511465124786, "eval_signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "eval_signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.004196511465124786, "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": 0.00189208984375, "eval_signal/format_reward/weight": 0.5, "eval_signal/format_reward/weighted_centered_abs_mean": 0.00189208984375, "eval_signal/frontier_aurc_reward/centered_abs_mean": 0.006113792187534273, "eval_signal/frontier_aurc_reward/group_std_mean": 0.012002985225990415, "eval_signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 7.64224014346837e-05, "eval_signal/frontier_aurc_reward/weight": 0.012500000186264515, "eval_signal/frontier_aurc_reward/weighted_centered_abs_mean": 7.64224014346837e-05, "eval_signal/frontier_coverage_10/centered_abs_mean": 0.3313248082995415, "eval_signal/frontier_coverage_10/group_std_mean": 0.41143108159303665, "eval_signal/frontier_coverage_10/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0041415601735934615, "eval_signal/frontier_coverage_10/weight": 0.012500000186264515, "eval_signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0041415601735934615, "eval_signal/frontier_coverage_15/centered_abs_mean": 0.3040120229125023, "eval_signal/frontier_coverage_15/group_std_mean": 0.3795413374900818, "eval_signal/frontier_coverage_15/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0038001504144631326, "eval_signal/frontier_coverage_15/weight": 0.012500000186264515, "eval_signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0038001504144631326, "eval_signal/frontier_coverage_20/centered_abs_mean": 0.18152179569005966, "eval_signal/frontier_coverage_20/group_std_mean": 0.2333090789616108, "eval_signal/frontier_coverage_20/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.002269022457767278, "eval_signal/frontier_coverage_20/weight": 0.012500000186264515, "eval_signal/frontier_coverage_20/weighted_centered_abs_mean": 0.002269022457767278, "eval_signal/frontier_coverage_25/centered_abs_mean": 0.09430776908993721, "eval_signal/frontier_coverage_25/group_std_mean": 0.12001598253846169, "eval_signal/frontier_coverage_25/group_zero_std_frac": 0.0, "eval_signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0011788471601903439, "eval_signal/frontier_coverage_25/weight": 0.012500000186264515, "eval_signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0011788471601903439, "eval_signal/frontier_ece_reward/centered_abs_mean": 0.007458990206941962, "eval_signal/frontier_ece_reward/group_std_mean": 0.009198537096381187, "eval_signal/frontier_ece_reward/group_zero_std_frac": 0.0, "eval_signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0007458990294253454, "eval_signal/frontier_ece_reward/weight": 0.10000000149011612, "eval_signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0007458990294253454, "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": 0.0003784179862122983, "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": 0.139, "step": 300 }, { "calibration/aurc": 0.2606589610327268, "calibration/batch_distribution_entropy": 0.9707845482539895, "calibration/buffer_distribution_entropy": 0.9986028947198576, "calibration/confidence_entropy": 0.5132883851241694, "calibration/coverage@0%": 0.046484375, "calibration/coverage@1%": 0.046484375, "calibration/coverage@10%": 0.26875, "calibration/coverage@15%": 0.3859375, "calibration/coverage@20%": 0.484375, "calibration/coverage@25%": 0.577734375, "calibration/coverage@30%": 0.616015625, "calibration/coverage@5%": 0.134375, "calibration/ece": 0.14023523949778266, "calibration/mean_confidence": 0.524700101373557, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 688.0, "completions/max_terminated_length": 688.0, "completions/mean_length": 228.7904296875, "completions/mean_terminated_length": 228.94674377441407, "completions/min_length": 25.4, "completions/min_terminated_length": 122.8, "epoch": 0.976, "grad_norm": 0.0009241602383553982, "learning_rate": 1e-06, "loss": -0.0005, "num_tokens": 1040825643.0, "reward": 0.9480137228965759, "reward_std": 0.08262738287448883, "rewards/accuracy_reward": 0.55556640625, "rewards/batch_coverage_0": 0.37107858061790466, "rewards/batch_coverage_1": 0.37107858061790466, "rewards/batch_coverage_5": 0.3869703412055969, "rewards/brier_reward": 0.7959351420402527, "rewards/confidence_uniqueness_reward": 0.9514917135238647, "rewards/format_reward": 0.99931640625, "rewards/frontier_aurc_reward": -0.003173557505942881, "rewards/frontier_coverage_10": 0.09997835960239172, "rewards/frontier_coverage_15": 0.09356479216367006, "rewards/frontier_coverage_20": 0.07112255636602641, "rewards/frontier_coverage_25": 0.05038532391190529, "rewards/frontier_ece_reward": 0.002544776737340726, "rewards/frontier_entropy_batch_reward": -0.2243742197751999, "signal/accuracy_reward/centered_abs_mean": 0.088360595703125, "signal/accuracy_reward/group_std_mean": 0.11695246323943138, "signal/accuracy_reward/group_zero_std_frac": 0.66875, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0441802978515625, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0441802978515625, "signal/advantage_abs_mean": 0.06376132741570473, "signal/advantage_pre_scale_abs_mean": 0.06376132741570473, "signal/advantage_pre_scale_std": 0.10275728553533554, "signal/advantage_std": 0.10275728553533554, "signal/batch_coverage_0/centered_abs_mean": 0.1470702588558197, "signal/batch_coverage_0/group_std_mean": 0.18703667521476747, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.001838378282263875, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.001838378282263875, "signal/batch_coverage_1/centered_abs_mean": 0.1470702588558197, "signal/batch_coverage_1/group_std_mean": 0.18703667521476747, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.001838378282263875, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.001838378282263875, "signal/batch_coverage_5/centered_abs_mean": 0.1504915565252304, "signal/batch_coverage_5/group_std_mean": 0.19151806235313415, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.0018811444984748959, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.0018811444984748959, "signal/brier_reward/centered_abs_mean": 0.10697132647037506, "signal/brier_reward/group_std_mean": 0.13873142302036284, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.010697133094072341, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.010697133094072341, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.013146862201392651, "signal/confidence_uniqueness_reward/group_std_mean": 0.017867789044976233, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0013146861921995879, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0013146861921995879, "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": 0.0006439208984375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0006439208984375, "signal/frontier_aurc_reward/centered_abs_mean": 0.002606235723942518, "signal/frontier_aurc_reward/group_std_mean": 0.0040924097411334515, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 3.257794633100275e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 3.257794633100275e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.15178255438804628, "signal/frontier_coverage_10/group_std_mean": 0.19580891728401184, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0018972819671034813, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0018972819671034813, "signal/frontier_coverage_15/centered_abs_mean": 0.1401112422347069, "signal/frontier_coverage_15/group_std_mean": 0.18112337589263916, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.001751390565186739, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.001751390565186739, "signal/frontier_coverage_20/centered_abs_mean": 0.08781458139419555, "signal/frontier_coverage_20/group_std_mean": 0.11425766199827195, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0010976822581142188, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0010976822581142188, "signal/frontier_coverage_25/centered_abs_mean": 0.05491496697068214, "signal/frontier_coverage_25/group_std_mean": 0.0706517368555069, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0006864371127448976, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0006864371127448976, "signal/frontier_ece_reward/centered_abs_mean": 0.005161408893764019, "signal/frontier_ece_reward/group_std_mean": 0.006664909049868583, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005161408917047083, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005161408917047083, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28244526982307433, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3530911445617676, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.02824452631175518, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.02824452631175518, "step": 305 }, { "calibration/aurc": 0.365374017494042, "calibration/batch_distribution_entropy": 0.977565572146543, "calibration/buffer_distribution_entropy": 0.9985787677455027, "calibration/confidence_entropy": 0.495913005234196, "calibration/coverage@0%": 0.016418504901960784, "calibration/coverage@1%": 0.016418504901960784, "calibration/coverage@10%": 0.04769761029411765, "calibration/coverage@15%": 0.08723958333333333, "calibration/coverage@20%": 0.13295802696078432, "calibration/coverage@25%": 0.2291176470588235, "calibration/coverage@30%": 0.4467463235294117, "calibration/coverage@5%": 0.017590379901960784, "calibration/ece": 0.14137153924889873, "calibration/mean_confidence": 0.43372769498577224, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 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": 680.0, "completions/max_terminated_length": 680.0, "completions/mean_length": 231.09736328125, "completions/mean_terminated_length": 231.21168518066406, "completions/min_length": 46.8, "completions/min_terminated_length": 117.8, "epoch": 0.992, "grad_norm": 0.0008104052394628525, "learning_rate": 1e-06, "loss": -0.0001, "num_tokens": 1058320560.0, "reward": 0.9328658699989318, "reward_std": 0.07802858650684356, "rewards/accuracy_reward": 0.5228515625, "rewards/batch_coverage_0": 0.38593215942382814, "rewards/batch_coverage_1": 0.38593215942382814, "rewards/batch_coverage_5": 0.3958631753921509, "rewards/brier_reward": 0.798362135887146, "rewards/confidence_uniqueness_reward": 0.9510131716728211, "rewards/format_reward": 0.99951171875, "rewards/frontier_aurc_reward": -0.002598634781315923, "rewards/frontier_coverage_10": 0.12723698765039443, "rewards/frontier_coverage_15": 0.11959595084190369, "rewards/frontier_coverage_20": 0.08388267457485199, "rewards/frontier_coverage_25": 0.05381896197795868, "rewards/frontier_ece_reward": 0.002247580257244408, "rewards/frontier_entropy_batch_reward": -0.22848860323429107, "signal/accuracy_reward/centered_abs_mean": 0.080419921875, "signal/accuracy_reward/group_std_mean": 0.10615092366933823, "signal/accuracy_reward/group_zero_std_frac": 0.7, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0402099609375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0402099609375, "signal/advantage_abs_mean": 0.06034851744771004, "signal/advantage_pre_scale_abs_mean": 0.06034851744771004, "signal/advantage_pre_scale_std": 0.09754381328821182, "signal/advantage_std": 0.09754381328821182, "signal/batch_coverage_0/centered_abs_mean": 0.1409256249666214, "signal/batch_coverage_0/group_std_mean": 0.1784323364496231, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.001761570363305509, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.001761570363305509, "signal/batch_coverage_1/centered_abs_mean": 0.1409256249666214, "signal/batch_coverage_1/group_std_mean": 0.1784323364496231, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.001761570363305509, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.001761570363305509, "signal/batch_coverage_5/centered_abs_mean": 0.14382491409778594, "signal/batch_coverage_5/group_std_mean": 0.18232801258563996, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.001797811477445066, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.001797811477445066, "signal/brier_reward/centered_abs_mean": 0.10617546737194061, "signal/brier_reward/group_std_mean": 0.13660653978586196, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.010617546737194061, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.010617546737194061, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.01346453782171011, "signal/confidence_uniqueness_reward/group_std_mean": 0.01773944068700075, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.001346453814767301, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.001346453814767301, "signal/format_reward/centered_abs_mean": 0.000921630859375, "signal/format_reward/group_std_mean": 0.0020895189139992, "signal/format_reward/group_zero_std_frac": 0.990625, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.0004608154296875, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.0004608154296875, "signal/frontier_aurc_reward/centered_abs_mean": 0.0019295686157420277, "signal/frontier_aurc_reward/group_std_mean": 0.0030131676234304906, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.4119608497130685e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.4119608497130685e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.15644589066505432, "signal/frontier_coverage_10/group_std_mean": 0.2001277267932892, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0019555736798793077, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0019555736798793077, "signal/frontier_coverage_15/centered_abs_mean": 0.14599400460720063, "signal/frontier_coverage_15/group_std_mean": 0.18689751327037812, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.001824925164692104, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.001824925164692104, "signal/frontier_coverage_20/centered_abs_mean": 0.09083109050989151, "signal/frontier_coverage_20/group_std_mean": 0.11699100136756897, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.0011353886453434824, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.0011353886453434824, "signal/frontier_coverage_25/centered_abs_mean": 0.054006976634263994, "signal/frontier_coverage_25/group_std_mean": 0.06911520063877105, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0006750872009433806, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0006750872009433806, "signal/frontier_ece_reward/centered_abs_mean": 0.004973601177334785, "signal/frontier_ece_reward/group_std_mean": 0.0063941096886992455, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.000497360120061785, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.000497360120061785, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28425076603889465, "signal/frontier_entropy_batch_reward/group_std_mean": 0.356691712141037, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.028425076603889467, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028425076603889467, "step": 310 }, { "calibration/aurc": 0.27088680212276284, "calibration/batch_distribution_entropy": 0.9847864550257454, "calibration/buffer_distribution_entropy": 0.9987949730158547, "calibration/confidence_entropy": 0.4944987664160725, "calibration/coverage@0%": 0.0107421875, "calibration/coverage@1%": 0.0107421875, "calibration/coverage@10%": 0.056640625, "calibration/coverage@15%": 0.21484375, "calibration/coverage@20%": 0.302734375, "calibration/coverage@25%": 0.40625, "calibration/coverage@30%": 0.6611328125, "calibration/coverage@5%": 0.0107421875, "calibration/ece": 0.12396722675023958, "calibration/mean_confidence": 0.5493359833581515, "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0005630181760203912, "completions/max_length": 514.5, "completions/max_terminated_length": 514.5, "completions/mean_length": 235.92743682861328, "completions/mean_terminated_length": 236.06040954589844, "completions/min_length": 0.0, "completions/min_terminated_length": 126.5, "epoch": 0.9984, "num_tokens": 1065291702.0, "reward": 0.9447338283061981, "reward_std": 0.0845753438770771, "rewards/accuracy_reward": 0.5546875, "rewards/batch_coverage_0": 0.3516007363796234, "rewards/batch_coverage_1": 0.3516007363796234, "rewards/batch_coverage_5": 0.36148694157600403, "rewards/brier_reward": 0.7678117156028748, "rewards/confidence_uniqueness_reward": 0.9530024528503418, "rewards/format_reward": 0.999267578125, "rewards/frontier_aurc_reward": -0.0026400991482660174, "rewards/frontier_coverage_10": 0.06620664149522781, "rewards/frontier_coverage_15": 0.0585860013961792, "rewards/frontier_coverage_20": 0.04189038276672363, "rewards/frontier_coverage_25": 0.03577941656112671, "rewards/frontier_ece_reward": 0.0014942202833481133, "rewards/frontier_entropy_batch_reward": -0.20280925929546356, "signal/accuracy_reward/centered_abs_mean": 0.093475341796875, "signal/accuracy_reward/group_std_mean": 0.12033192440867424, "signal/accuracy_reward/group_zero_std_frac": 0.65625, "signal/accuracy_reward/scaled_weighted_centered_abs_mean": 0.0467376708984375, "signal/accuracy_reward/weight": 0.5, "signal/accuracy_reward/weighted_centered_abs_mean": 0.0467376708984375, "signal/advantage_abs_mean": 0.06638789363205433, "signal/advantage_pre_scale_abs_mean": 0.06638789363205433, "signal/advantage_pre_scale_std": 0.10692764818668365, "signal/advantage_std": 0.10692764818668365, "signal/batch_coverage_0/centered_abs_mean": 0.15826985239982605, "signal/batch_coverage_0/group_std_mean": 0.2002842053771019, "signal/batch_coverage_0/group_zero_std_frac": 0.0, "signal/batch_coverage_0/scaled_weighted_centered_abs_mean": 0.001978373038582504, "signal/batch_coverage_0/weight": 0.012500000186264515, "signal/batch_coverage_0/weighted_centered_abs_mean": 0.001978373038582504, "signal/batch_coverage_1/centered_abs_mean": 0.15826985239982605, "signal/batch_coverage_1/group_std_mean": 0.2002842053771019, "signal/batch_coverage_1/group_zero_std_frac": 0.0, "signal/batch_coverage_1/scaled_weighted_centered_abs_mean": 0.001978373038582504, "signal/batch_coverage_1/weight": 0.012500000186264515, "signal/batch_coverage_1/weighted_centered_abs_mean": 0.001978373038582504, "signal/batch_coverage_5/centered_abs_mean": 0.16011271625757217, "signal/batch_coverage_5/group_std_mean": 0.2030055671930313, "signal/batch_coverage_5/group_zero_std_frac": 0.0, "signal/batch_coverage_5/scaled_weighted_centered_abs_mean": 0.002001408953219652, "signal/batch_coverage_5/weight": 0.012500000186264515, "signal/batch_coverage_5/weighted_centered_abs_mean": 0.002001408953219652, "signal/brier_reward/centered_abs_mean": 0.11534750834107399, "signal/brier_reward/group_std_mean": 0.14829222857952118, "signal/brier_reward/group_zero_std_frac": 0.0, "signal/brier_reward/scaled_weighted_centered_abs_mean": 0.011534750927239656, "signal/brier_reward/weight": 0.10000000149011612, "signal/brier_reward/weighted_centered_abs_mean": 0.011534750927239656, "signal/confidence_uniqueness_reward/centered_abs_mean": 0.012511179782450199, "signal/confidence_uniqueness_reward/group_std_mean": 0.017198790796101093, "signal/confidence_uniqueness_reward/group_zero_std_frac": 0.0, "signal/confidence_uniqueness_reward/scaled_weighted_centered_abs_mean": 0.0012511180248111486, "signal/confidence_uniqueness_reward/weight": 0.10000000149011612, "signal/confidence_uniqueness_reward/weighted_centered_abs_mean": 0.0012511180248111486, "signal/format_reward/centered_abs_mean": 0.0013885498046875, "signal/format_reward/group_std_mean": 0.003302432596683502, "signal/format_reward/group_zero_std_frac": 0.984375, "signal/format_reward/scaled_weighted_centered_abs_mean": 0.00069427490234375, "signal/format_reward/weight": 0.5, "signal/format_reward/weighted_centered_abs_mean": 0.00069427490234375, "signal/frontier_aurc_reward/centered_abs_mean": 0.0021804602001793683, "signal/frontier_aurc_reward/group_std_mean": 0.0039160531014204025, "signal/frontier_aurc_reward/group_zero_std_frac": 0.0, "signal/frontier_aurc_reward/scaled_weighted_centered_abs_mean": 2.7255752684141044e-05, "signal/frontier_aurc_reward/weight": 0.012500000186264515, "signal/frontier_aurc_reward/weighted_centered_abs_mean": 2.7255752684141044e-05, "signal/frontier_coverage_10/centered_abs_mean": 0.15913717448711395, "signal/frontier_coverage_10/group_std_mean": 0.20392150431871414, "signal/frontier_coverage_10/group_zero_std_frac": 0.0, "signal/frontier_coverage_10/scaled_weighted_centered_abs_mean": 0.0019892147975042462, "signal/frontier_coverage_10/weight": 0.012500000186264515, "signal/frontier_coverage_10/weighted_centered_abs_mean": 0.0019892147975042462, "signal/frontier_coverage_15/centered_abs_mean": 0.1482202708721161, "signal/frontier_coverage_15/group_std_mean": 0.18995162844657898, "signal/frontier_coverage_15/group_zero_std_frac": 0.0, "signal/frontier_coverage_15/scaled_weighted_centered_abs_mean": 0.0018527533393353224, "signal/frontier_coverage_15/weight": 0.012500000186264515, "signal/frontier_coverage_15/weighted_centered_abs_mean": 0.0018527533393353224, "signal/frontier_coverage_20/centered_abs_mean": 0.08981956541538239, "signal/frontier_coverage_20/group_std_mean": 0.11610593646764755, "signal/frontier_coverage_20/group_zero_std_frac": 0.0, "signal/frontier_coverage_20/scaled_weighted_centered_abs_mean": 0.001122744637541473, "signal/frontier_coverage_20/weight": 0.012500000186264515, "signal/frontier_coverage_20/weighted_centered_abs_mean": 0.001122744637541473, "signal/frontier_coverage_25/centered_abs_mean": 0.05227947235107422, "signal/frontier_coverage_25/group_std_mean": 0.06755021214485168, "signal/frontier_coverage_25/group_zero_std_frac": 0.0, "signal/frontier_coverage_25/scaled_weighted_centered_abs_mean": 0.0006534934509545565, "signal/frontier_coverage_25/weight": 0.012500000186264515, "signal/frontier_coverage_25/weighted_centered_abs_mean": 0.0006534934509545565, "signal/frontier_ece_reward/centered_abs_mean": 0.0052431917283684015, "signal/frontier_ece_reward/group_std_mean": 0.006774711888283491, "signal/frontier_ece_reward/group_zero_std_frac": 0.0, "signal/frontier_ece_reward/scaled_weighted_centered_abs_mean": 0.0005243191844783723, "signal/frontier_ece_reward/weight": 0.10000000149011612, "signal/frontier_ece_reward/weighted_centered_abs_mean": 0.0005243191844783723, "signal/frontier_entropy_batch_reward/centered_abs_mean": 0.28239084780216217, "signal/frontier_entropy_batch_reward/group_std_mean": 0.3569559305906296, "signal/frontier_entropy_batch_reward/group_zero_std_frac": 0.0, "signal/frontier_entropy_batch_reward/scaled_weighted_centered_abs_mean": 0.028239086270332336, "signal/frontier_entropy_batch_reward/weight": 0.10000000149011612, "signal/frontier_entropy_batch_reward/weighted_centered_abs_mean": 0.028239086270332336, "step": 312, "total_flos": 0.0, "train_loss": -0.0010778324898498026, "train_runtime": 62009.7571, "train_samples_per_second": 0.323, "train_steps_per_second": 0.005 } ], "logging_steps": 5, "max_steps": 312, "num_input_tokens_seen": 1065291702, "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 }